{ "architectures": [ "MiniMaxM3SparseForConditionalGeneration" ], "auto_map": { "AutoConfig": "configuration_minimax_m3_vl.MiniMaxM3VLConfig" }, "dtype": "bfloat16", "image_grid_pinpoints": "[(336, 336), (336, 672), (336, 1008), (336, 1344), (336, 1680), (336, 2016), (672, 336), (672, 672), (672, 1008), (672, 1344), (672, 1680), (672, 2016), (1008, 336), (1008, 672), (1008, 1008), (1008, 1344), (1008, 1680), (1008, 2016), (1344, 336), (1344, 672), (1344, 1008), (1344, 1344), (1344, 1680), (1344, 2016), (1680, 336), (1680, 672), (1680, 1008), (1680, 1344), (1680, 1680), (1680, 2016), (2016, 336), (2016, 672), (2016, 1008), (2016, 1344), (2016, 1680), (2016, 2016)]", "image_seq_length": 576, "image_token_index": 200025, "img_token_compression_config": { "image_token_compression_method": "patch_merge", "spatial_merge_size": 2, "temporal_patch_size": 2 }, "merged_hidden_size": 24576, "model_type": "minimax_m3_vl", "multimodal_projector_bias": true, "num_reward_heads": 0, "process_image_mode": "dynamic_res", "projector_hidden_act": "gelu", "projector_hidden_size": 6144, "quantization_config": { "autoround_version": "0.13.0", "batch_size": 1, "bits": 16, "block_name_to_quantize": "model.language_model.layers", "data_type": "int", "extra_config": { ".*model\\.language_model\\.layers\\.[3-59]\\.mlp\\.gate.*": { "data_type": "float" }, "model.language_model.layers.0.mlp.down_proj": { "bits": 4 }, "model.language_model.layers.0.mlp.gate_up_proj": { "bits": 4 }, "model.language_model.layers.0.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.0.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.0.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.0.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.1.mlp.down_proj": { "bits": 4 }, "model.language_model.layers.1.mlp.gate_up_proj": { "bits": 4 }, "model.language_model.layers.1.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.1.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.1.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.1.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.10.mlp.experts.0.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.0.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.0.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.1.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.1.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.1.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.10.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.10.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.10.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.100.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.100.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.100.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.101.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.101.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.101.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.102.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.102.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.102.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.103.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.103.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.103.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.104.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.104.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.104.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.105.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.105.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.105.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.106.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.106.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.106.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.107.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.107.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.107.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.108.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.108.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.108.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.109.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.109.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.109.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.11.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.11.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.11.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.110.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.110.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.110.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.111.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.111.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.111.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.112.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.112.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.112.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.113.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.113.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.113.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.114.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.114.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.114.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.115.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.115.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.115.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.116.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.116.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.116.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.117.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.117.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.117.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.118.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.118.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.118.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.119.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.119.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.119.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.12.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.12.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.12.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.120.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.120.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.120.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.121.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.121.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.121.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.122.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.122.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.122.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.123.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.123.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.123.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.124.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.124.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.124.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.125.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.125.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.125.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.126.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.126.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.126.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.127.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.127.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.127.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.13.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.13.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.13.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.14.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.14.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.14.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.15.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.15.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.15.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.16.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.16.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.16.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.17.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.17.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.17.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.18.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.18.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.18.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.19.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.19.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.19.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.2.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.2.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.2.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.20.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.20.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.20.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.21.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.21.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.21.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.22.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.22.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.22.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.23.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.23.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.23.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.24.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.24.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.24.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.25.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.25.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.25.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.26.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.26.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.26.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.27.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.27.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.27.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.28.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.28.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.28.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.29.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.29.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.29.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.3.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.3.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.3.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.30.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.30.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.30.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.31.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.31.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.31.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.32.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.32.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.32.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.33.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.33.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.33.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.34.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.34.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.34.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.35.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.35.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.35.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.36.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.36.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.36.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.37.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.37.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.37.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.38.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.38.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.38.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.39.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.39.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.39.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.4.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.4.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.4.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.40.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.40.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.40.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.41.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.41.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.41.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.42.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.42.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.42.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.43.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.43.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.43.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.44.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.44.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.44.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.45.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.45.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.45.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.46.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.46.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.46.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.47.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.47.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.47.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.48.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.48.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.48.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.49.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.49.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.49.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.5.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.5.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.5.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.50.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.50.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.50.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.51.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.51.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.51.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.52.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.52.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.52.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.53.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.53.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.53.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.54.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.54.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.54.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.55.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.55.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.55.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.56.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.56.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.56.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.57.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.57.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.57.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.58.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.58.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.58.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.59.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.59.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.59.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.6.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.6.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.6.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.60.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.60.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.60.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.61.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.61.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.61.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.62.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.62.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.62.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.63.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.63.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.63.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.64.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.64.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.64.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.65.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.65.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.65.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.66.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.66.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.66.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.67.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.67.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.67.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.68.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.68.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.68.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.69.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.69.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.69.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.7.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.7.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.7.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.70.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.70.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.70.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.71.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.71.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.71.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.72.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.72.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.72.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.73.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.73.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.73.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.74.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.74.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.74.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.75.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.75.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.75.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.76.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.76.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.76.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.77.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.77.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.77.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.78.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.78.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.78.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.79.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.79.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.79.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.8.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.8.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.8.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.80.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.80.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.80.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.81.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.81.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.81.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.82.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.82.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.82.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.83.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.83.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.83.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.84.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.84.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.84.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.85.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.85.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.85.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.86.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.86.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.86.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.87.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.87.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.87.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.88.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.88.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.88.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.89.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.89.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.89.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.9.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.9.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.9.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.90.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.90.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.90.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.91.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.91.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.91.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.92.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.92.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.92.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.93.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.93.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.93.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.94.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.94.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.94.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.95.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.95.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.95.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.96.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.96.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.96.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.97.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.97.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.97.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.98.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.98.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.98.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.99.down_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.99.gate_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.experts.99.up_proj": { "bits": 2 }, "model.language_model.layers.10.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.10.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.10.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.10.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.10.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.10.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.11.mlp.experts.0.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.0.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.0.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.1.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.1.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.1.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.10.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.10.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.10.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.100.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.100.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.100.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.101.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.101.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.101.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.102.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.102.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.102.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.103.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.103.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.103.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.104.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.104.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.104.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.105.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.105.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.105.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.106.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.106.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.106.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.107.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.107.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.107.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.108.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.108.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.108.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.109.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.109.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.109.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.11.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.11.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.11.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.110.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.110.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.110.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.111.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.111.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.111.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.112.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.112.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.112.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.113.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.113.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.113.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.114.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.114.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.114.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.115.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.115.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.115.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.116.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.116.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.116.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.117.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.117.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.117.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.118.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.118.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.118.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.119.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.119.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.119.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.12.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.12.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.12.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.120.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.120.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.120.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.121.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.121.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.121.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.122.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.122.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.122.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.123.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.123.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.123.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.124.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.124.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.124.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.125.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.125.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.125.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.126.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.126.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.126.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.127.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.127.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.127.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.13.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.13.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.13.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.14.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.14.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.14.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.15.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.15.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.15.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.16.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.16.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.16.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.17.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.17.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.17.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.18.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.18.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.18.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.19.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.19.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.19.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.2.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.2.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.2.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.20.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.20.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.20.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.21.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.21.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.21.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.22.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.22.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.22.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.23.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.23.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.23.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.24.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.24.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.24.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.25.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.25.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.25.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.26.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.26.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.26.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.27.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.27.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.27.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.28.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.28.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.28.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.29.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.29.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.29.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.3.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.3.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.3.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.30.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.30.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.30.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.31.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.31.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.31.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.32.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.32.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.32.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.33.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.33.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.33.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.34.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.34.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.34.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.35.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.35.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.35.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.36.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.36.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.36.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.37.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.37.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.37.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.38.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.38.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.38.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.39.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.39.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.39.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.4.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.4.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.4.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.40.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.40.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.40.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.41.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.41.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.41.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.42.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.42.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.42.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.43.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.43.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.43.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.44.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.44.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.44.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.45.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.45.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.45.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.46.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.46.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.46.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.47.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.47.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.47.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.48.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.48.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.48.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.49.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.49.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.49.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.5.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.5.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.5.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.50.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.50.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.50.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.51.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.51.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.51.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.52.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.52.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.52.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.53.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.53.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.53.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.54.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.54.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.54.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.55.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.55.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.55.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.56.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.56.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.56.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.57.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.57.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.57.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.58.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.58.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.58.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.59.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.59.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.59.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.6.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.6.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.6.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.60.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.60.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.60.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.61.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.61.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.61.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.62.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.62.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.62.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.63.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.63.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.63.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.64.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.64.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.64.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.65.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.65.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.65.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.66.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.66.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.66.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.67.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.67.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.67.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.68.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.68.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.68.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.69.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.69.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.69.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.7.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.7.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.7.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.70.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.70.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.70.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.71.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.71.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.71.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.72.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.72.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.72.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.73.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.73.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.73.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.74.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.74.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.74.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.75.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.75.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.75.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.76.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.76.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.76.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.77.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.77.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.77.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.78.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.78.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.78.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.79.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.79.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.79.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.8.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.8.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.8.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.80.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.80.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.80.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.81.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.81.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.81.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.82.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.82.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.82.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.83.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.83.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.83.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.84.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.84.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.84.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.85.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.85.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.85.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.86.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.86.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.86.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.87.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.87.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.87.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.88.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.88.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.88.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.89.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.89.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.89.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.9.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.9.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.9.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.90.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.90.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.90.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.91.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.91.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.91.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.92.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.92.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.92.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.93.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.93.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.93.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.94.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.94.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.94.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.95.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.95.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.95.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.96.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.96.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.96.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.97.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.97.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.97.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.98.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.98.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.98.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.99.down_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.99.gate_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.experts.99.up_proj": { "bits": 2 }, "model.language_model.layers.11.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.11.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.11.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.11.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.11.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.11.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.12.mlp.experts.0.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.0.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.0.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.1.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.1.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.1.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.10.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.10.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.10.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.100.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.100.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.100.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.101.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.101.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.101.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.102.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.102.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.102.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.103.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.103.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.103.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.104.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.104.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.104.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.105.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.105.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.105.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.106.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.106.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.106.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.107.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.107.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.107.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.108.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.108.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.108.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.109.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.109.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.109.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.11.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.11.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.11.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.110.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.110.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.110.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.111.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.111.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.111.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.112.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.112.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.112.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.113.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.113.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.113.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.114.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.114.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.114.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.115.down_proj": { "bits": 3 }, "model.language_model.layers.12.mlp.experts.115.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.115.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.116.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.116.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.116.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.117.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.117.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.117.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.118.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.118.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.118.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.119.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.119.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.119.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.12.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.12.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.12.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.120.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.120.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.120.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.121.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.121.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.121.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.122.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.122.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.122.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.123.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.123.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.123.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.124.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.124.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.124.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.125.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.125.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.125.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.126.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.126.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.126.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.127.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.127.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.127.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.13.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.13.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.13.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.14.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.14.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.14.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.15.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.15.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.15.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.16.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.16.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.16.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.17.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.17.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.17.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.18.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.18.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.18.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.19.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.19.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.19.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.2.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.2.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.2.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.20.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.20.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.20.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.21.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.21.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.21.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.22.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.22.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.22.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.23.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.23.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.23.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.24.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.24.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.24.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.25.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.25.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.25.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.26.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.26.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.26.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.27.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.27.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.27.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.28.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.28.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.28.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.29.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.29.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.29.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.3.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.3.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.3.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.30.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.30.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.30.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.31.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.31.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.31.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.32.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.32.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.32.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.33.down_proj": { "bits": 3 }, "model.language_model.layers.12.mlp.experts.33.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.33.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.34.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.34.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.34.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.35.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.35.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.35.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.36.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.36.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.36.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.37.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.37.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.37.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.38.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.38.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.38.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.39.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.39.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.39.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.4.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.4.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.4.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.40.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.40.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.40.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.41.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.41.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.41.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.42.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.42.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.42.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.43.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.43.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.43.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.44.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.44.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.44.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.45.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.45.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.45.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.46.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.46.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.46.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.47.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.47.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.47.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.48.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.48.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.48.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.49.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.49.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.49.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.5.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.5.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.5.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.50.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.50.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.50.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.51.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.51.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.51.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.52.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.52.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.52.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.53.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.53.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.53.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.54.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.54.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.54.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.55.down_proj": { "bits": 3 }, "model.language_model.layers.12.mlp.experts.55.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.55.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.56.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.56.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.56.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.57.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.57.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.57.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.58.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.58.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.58.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.59.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.59.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.59.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.6.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.6.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.6.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.60.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.60.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.60.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.61.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.61.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.61.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.62.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.62.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.62.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.63.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.63.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.63.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.64.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.64.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.64.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.65.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.65.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.65.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.66.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.66.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.66.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.67.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.67.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.67.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.68.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.68.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.68.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.69.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.69.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.69.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.7.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.7.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.7.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.70.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.70.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.70.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.71.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.71.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.71.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.72.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.72.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.72.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.73.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.73.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.73.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.74.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.74.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.74.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.75.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.75.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.75.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.76.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.76.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.76.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.77.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.77.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.77.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.78.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.78.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.78.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.79.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.79.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.79.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.8.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.8.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.8.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.80.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.80.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.80.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.81.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.81.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.81.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.82.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.82.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.82.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.83.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.83.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.83.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.84.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.84.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.84.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.85.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.85.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.85.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.86.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.86.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.86.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.87.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.87.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.87.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.88.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.88.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.88.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.89.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.89.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.89.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.9.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.9.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.9.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.90.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.90.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.90.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.91.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.91.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.91.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.92.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.92.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.92.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.93.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.93.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.93.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.94.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.94.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.94.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.95.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.95.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.95.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.96.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.96.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.96.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.97.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.97.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.97.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.98.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.98.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.98.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.99.down_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.99.gate_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.experts.99.up_proj": { "bits": 2 }, "model.language_model.layers.12.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.12.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.12.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.12.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.12.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.12.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.13.mlp.experts.0.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.0.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.0.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.1.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.1.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.1.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.10.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.10.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.10.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.100.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.100.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.100.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.101.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.101.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.101.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.102.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.102.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.102.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.103.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.103.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.103.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.104.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.104.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.104.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.105.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.105.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.105.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.106.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.106.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.106.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.107.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.107.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.107.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.108.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.108.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.108.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.109.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.109.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.109.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.11.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.11.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.11.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.110.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.110.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.110.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.111.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.111.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.111.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.112.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.112.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.112.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.113.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.113.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.113.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.114.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.114.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.114.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.115.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.115.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.115.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.116.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.116.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.116.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.117.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.117.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.117.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.118.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.118.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.118.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.119.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.119.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.119.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.12.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.12.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.12.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.120.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.120.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.120.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.121.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.121.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.121.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.122.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.122.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.122.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.123.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.123.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.123.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.124.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.124.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.124.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.125.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.125.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.125.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.126.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.126.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.126.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.127.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.127.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.127.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.13.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.13.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.13.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.14.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.14.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.14.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.15.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.15.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.15.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.16.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.16.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.16.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.17.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.17.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.17.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.18.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.18.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.18.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.19.down_proj": { "bits": 3 }, "model.language_model.layers.13.mlp.experts.19.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.19.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.2.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.2.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.2.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.20.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.20.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.20.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.21.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.21.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.21.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.22.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.22.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.22.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.23.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.23.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.23.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.24.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.24.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.24.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.25.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.25.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.25.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.26.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.26.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.26.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.27.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.27.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.27.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.28.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.28.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.28.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.29.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.29.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.29.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.3.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.3.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.3.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.30.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.30.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.30.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.31.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.31.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.31.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.32.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.32.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.32.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.33.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.33.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.33.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.34.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.34.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.34.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.35.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.35.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.35.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.36.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.36.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.36.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.37.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.37.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.37.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.38.down_proj": { "bits": 3 }, "model.language_model.layers.13.mlp.experts.38.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.38.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.39.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.39.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.39.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.4.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.4.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.4.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.40.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.40.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.40.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.41.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.41.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.41.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.42.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.42.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.42.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.43.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.43.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.43.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.44.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.44.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.44.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.45.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.45.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.45.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.46.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.46.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.46.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.47.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.47.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.47.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.48.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.48.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.48.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.49.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.49.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.49.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.5.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.5.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.5.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.50.down_proj": { "bits": 3 }, "model.language_model.layers.13.mlp.experts.50.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.50.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.51.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.51.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.51.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.52.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.52.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.52.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.53.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.53.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.53.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.54.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.54.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.54.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.55.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.55.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.55.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.56.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.56.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.56.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.57.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.57.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.57.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.58.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.58.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.58.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.59.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.59.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.59.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.6.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.6.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.6.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.60.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.60.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.60.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.61.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.61.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.61.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.62.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.62.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.62.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.63.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.63.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.63.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.64.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.64.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.64.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.65.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.65.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.65.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.66.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.66.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.66.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.67.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.67.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.67.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.68.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.68.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.68.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.69.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.69.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.69.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.7.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.7.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.7.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.70.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.70.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.70.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.71.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.71.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.71.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.72.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.72.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.72.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.73.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.73.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.73.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.74.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.74.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.74.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.75.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.75.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.75.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.76.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.76.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.76.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.77.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.77.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.77.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.78.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.78.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.78.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.79.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.79.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.79.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.8.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.8.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.8.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.80.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.80.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.80.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.81.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.81.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.81.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.82.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.82.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.82.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.83.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.83.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.83.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.84.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.84.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.84.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.85.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.85.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.85.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.86.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.86.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.86.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.87.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.87.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.87.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.88.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.88.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.88.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.89.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.89.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.89.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.9.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.9.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.9.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.90.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.90.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.90.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.91.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.91.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.91.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.92.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.92.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.92.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.93.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.93.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.93.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.94.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.94.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.94.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.95.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.95.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.95.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.96.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.96.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.96.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.97.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.97.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.97.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.98.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.98.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.98.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.99.down_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.99.gate_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.experts.99.up_proj": { "bits": 2 }, "model.language_model.layers.13.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.13.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.13.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.13.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.13.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.13.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.14.mlp.experts.0.down_proj": { "bits": 3 }, "model.language_model.layers.14.mlp.experts.0.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.0.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.1.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.1.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.1.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.10.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.10.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.10.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.100.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.100.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.100.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.101.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.101.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.101.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.102.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.102.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.102.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.103.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.103.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.103.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.104.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.104.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.104.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.105.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.105.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.105.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.106.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.106.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.106.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.107.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.107.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.107.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.108.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.108.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.108.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.109.down_proj": { "bits": 3 }, "model.language_model.layers.14.mlp.experts.109.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.109.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.11.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.11.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.11.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.110.down_proj": { "bits": 3 }, "model.language_model.layers.14.mlp.experts.110.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.110.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.111.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.111.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.111.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.112.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.112.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.112.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.113.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.113.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.113.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.114.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.114.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.114.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.115.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.115.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.115.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.116.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.116.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.116.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.117.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.117.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.117.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.118.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.118.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.118.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.119.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.119.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.119.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.12.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.12.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.12.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.120.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.120.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.120.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.121.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.121.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.121.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.122.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.122.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.122.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.123.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.123.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.123.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.124.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.124.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.124.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.125.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.125.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.125.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.126.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.126.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.126.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.127.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.127.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.127.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.13.down_proj": { "bits": 3 }, "model.language_model.layers.14.mlp.experts.13.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.13.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.14.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.14.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.14.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.15.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.15.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.15.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.16.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.16.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.16.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.17.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.17.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.17.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.18.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.18.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.18.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.19.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.19.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.19.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.2.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.2.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.2.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.20.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.20.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.20.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.21.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.21.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.21.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.22.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.22.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.22.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.23.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.23.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.23.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.24.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.24.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.24.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.25.down_proj": { "bits": 3 }, "model.language_model.layers.14.mlp.experts.25.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.25.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.26.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.26.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.26.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.27.down_proj": { "bits": 3 }, "model.language_model.layers.14.mlp.experts.27.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.27.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.28.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.28.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.28.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.29.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.29.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.29.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.3.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.3.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.3.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.30.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.30.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.30.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.31.down_proj": { "bits": 3 }, "model.language_model.layers.14.mlp.experts.31.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.31.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.32.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.32.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.32.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.33.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.33.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.33.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.34.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.34.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.34.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.35.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.35.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.35.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.36.down_proj": { "bits": 3 }, "model.language_model.layers.14.mlp.experts.36.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.36.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.37.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.37.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.37.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.38.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.38.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.38.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.39.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.39.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.39.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.4.down_proj": { "bits": 3 }, "model.language_model.layers.14.mlp.experts.4.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.4.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.40.down_proj": { "bits": 3 }, "model.language_model.layers.14.mlp.experts.40.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.40.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.41.down_proj": { "bits": 3 }, "model.language_model.layers.14.mlp.experts.41.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.41.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.42.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.42.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.42.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.43.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.43.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.43.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.44.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.44.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.44.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.45.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.45.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.45.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.46.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.46.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.46.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.47.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.47.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.47.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.48.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.48.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.48.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.49.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.49.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.49.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.5.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.5.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.5.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.50.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.50.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.50.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.51.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.51.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.51.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.52.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.52.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.52.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.53.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.53.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.53.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.54.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.54.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.54.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.55.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.55.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.55.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.56.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.56.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.56.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.57.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.57.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.57.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.58.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.58.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.58.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.59.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.59.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.59.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.6.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.6.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.6.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.60.down_proj": { "bits": 3 }, "model.language_model.layers.14.mlp.experts.60.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.60.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.61.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.61.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.61.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.62.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.62.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.62.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.63.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.63.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.63.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.64.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.64.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.64.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.65.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.65.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.65.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.66.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.66.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.66.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.67.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.67.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.67.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.68.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.68.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.68.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.69.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.69.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.69.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.7.down_proj": { "bits": 3 }, "model.language_model.layers.14.mlp.experts.7.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.7.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.70.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.70.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.70.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.71.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.71.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.71.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.72.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.72.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.72.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.73.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.73.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.73.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.74.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.74.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.74.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.75.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.75.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.75.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.76.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.76.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.76.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.77.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.77.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.77.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.78.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.78.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.78.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.79.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.79.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.79.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.8.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.8.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.8.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.80.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.80.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.80.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.81.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.81.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.81.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.82.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.82.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.82.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.83.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.83.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.83.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.84.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.84.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.84.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.85.down_proj": { "bits": 3 }, "model.language_model.layers.14.mlp.experts.85.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.85.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.86.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.86.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.86.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.87.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.87.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.87.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.88.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.88.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.88.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.89.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.89.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.89.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.9.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.9.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.9.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.90.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.90.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.90.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.91.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.91.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.91.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.92.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.92.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.92.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.93.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.93.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.93.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.94.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.94.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.94.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.95.down_proj": { "bits": 3 }, "model.language_model.layers.14.mlp.experts.95.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.95.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.96.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.96.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.96.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.97.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.97.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.97.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.98.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.98.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.98.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.99.down_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.99.gate_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.experts.99.up_proj": { "bits": 2 }, "model.language_model.layers.14.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.14.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.14.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.14.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.14.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.14.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.15.mlp.experts.0.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.0.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.0.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.1.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.1.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.1.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.10.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.10.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.10.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.100.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.100.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.100.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.101.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.101.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.101.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.102.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.102.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.102.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.103.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.103.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.103.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.104.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.104.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.104.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.105.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.105.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.105.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.106.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.106.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.106.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.107.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.107.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.107.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.108.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.108.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.108.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.109.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.109.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.109.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.11.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.11.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.11.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.110.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.110.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.110.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.111.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.111.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.111.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.112.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.112.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.112.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.113.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.113.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.113.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.114.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.114.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.114.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.115.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.115.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.115.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.116.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.116.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.116.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.117.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.117.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.117.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.118.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.118.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.118.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.119.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.119.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.119.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.12.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.12.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.12.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.120.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.120.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.120.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.121.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.121.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.121.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.122.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.122.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.122.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.123.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.123.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.123.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.124.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.124.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.124.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.125.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.125.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.125.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.126.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.126.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.126.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.127.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.127.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.127.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.13.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.13.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.13.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.14.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.14.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.14.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.15.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.15.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.15.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.16.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.16.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.16.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.17.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.17.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.17.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.18.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.18.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.18.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.19.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.19.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.19.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.2.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.2.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.2.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.20.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.20.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.20.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.21.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.21.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.21.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.22.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.22.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.22.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.23.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.23.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.23.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.24.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.24.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.24.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.25.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.25.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.25.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.26.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.26.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.26.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.27.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.27.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.27.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.28.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.28.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.28.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.29.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.29.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.29.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.3.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.3.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.3.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.30.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.30.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.30.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.31.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.31.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.31.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.32.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.32.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.32.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.33.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.33.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.33.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.34.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.34.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.34.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.35.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.35.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.35.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.36.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.36.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.36.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.37.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.37.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.37.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.38.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.38.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.38.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.39.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.39.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.39.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.4.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.4.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.4.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.40.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.40.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.40.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.41.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.41.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.41.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.42.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.42.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.42.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.43.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.43.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.43.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.44.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.44.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.44.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.45.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.45.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.45.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.46.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.46.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.46.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.47.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.47.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.47.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.48.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.48.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.48.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.49.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.49.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.49.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.5.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.5.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.5.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.50.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.50.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.50.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.51.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.51.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.51.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.52.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.52.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.52.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.53.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.53.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.53.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.54.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.54.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.54.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.55.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.55.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.55.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.56.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.56.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.56.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.57.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.57.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.57.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.58.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.58.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.58.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.59.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.59.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.59.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.6.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.6.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.6.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.60.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.60.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.60.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.61.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.61.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.61.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.62.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.62.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.62.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.63.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.63.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.63.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.64.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.64.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.64.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.65.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.65.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.65.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.66.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.66.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.66.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.67.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.67.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.67.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.68.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.68.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.68.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.69.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.69.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.69.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.7.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.7.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.7.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.70.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.70.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.70.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.71.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.71.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.71.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.72.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.72.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.72.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.73.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.73.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.73.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.74.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.74.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.74.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.75.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.75.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.75.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.76.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.76.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.76.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.77.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.77.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.77.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.78.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.78.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.78.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.79.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.79.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.79.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.8.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.8.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.8.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.80.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.80.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.80.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.81.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.81.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.81.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.82.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.82.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.82.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.83.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.83.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.83.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.84.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.84.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.84.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.85.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.85.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.85.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.86.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.86.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.86.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.87.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.87.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.87.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.88.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.88.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.88.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.89.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.89.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.89.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.9.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.9.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.9.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.90.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.90.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.90.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.91.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.91.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.91.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.92.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.92.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.92.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.93.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.93.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.93.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.94.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.94.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.94.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.95.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.95.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.95.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.96.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.96.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.96.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.97.down_proj": { "bits": 3 }, "model.language_model.layers.15.mlp.experts.97.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.97.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.98.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.98.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.98.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.99.down_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.99.gate_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.experts.99.up_proj": { "bits": 2 }, "model.language_model.layers.15.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.15.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.15.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.15.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.15.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.15.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.16.mlp.experts.0.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.0.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.0.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.1.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.1.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.1.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.10.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.10.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.10.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.100.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.100.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.100.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.101.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.101.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.101.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.102.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.102.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.102.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.103.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.103.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.103.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.104.down_proj": { "bits": 4 }, "model.language_model.layers.16.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.105.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.105.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.105.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.106.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.106.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.106.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.107.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.107.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.107.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.108.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.108.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.108.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.109.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.109.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.109.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.11.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.11.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.11.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.110.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.110.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.110.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.111.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.111.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.111.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.112.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.112.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.112.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.113.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.113.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.113.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.114.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.114.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.114.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.115.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.115.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.115.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.116.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.116.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.116.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.117.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.117.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.117.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.118.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.118.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.118.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.119.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.119.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.119.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.12.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.12.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.12.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.120.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.120.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.120.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.121.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.121.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.121.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.122.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.122.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.122.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.123.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.123.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.123.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.124.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.124.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.124.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.125.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.125.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.125.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.126.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.126.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.126.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.127.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.127.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.127.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.13.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.13.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.13.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.14.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.14.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.14.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.15.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.15.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.15.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.16.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.16.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.16.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.17.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.17.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.17.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.18.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.18.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.18.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.19.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.19.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.19.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.2.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.2.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.2.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.20.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.20.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.20.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.21.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.21.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.21.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.22.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.22.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.22.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.23.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.23.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.23.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.24.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.24.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.24.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.25.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.25.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.25.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.26.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.26.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.26.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.27.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.27.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.27.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.28.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.28.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.28.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.29.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.29.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.29.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.3.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.3.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.3.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.30.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.30.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.30.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.31.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.31.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.31.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.32.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.32.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.32.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.33.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.33.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.33.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.34.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.34.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.34.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.35.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.35.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.35.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.36.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.36.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.36.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.37.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.37.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.37.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.38.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.38.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.38.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.39.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.39.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.39.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.4.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.4.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.4.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.40.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.40.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.40.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.41.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.41.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.41.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.42.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.42.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.42.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.43.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.43.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.43.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.44.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.44.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.44.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.45.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.45.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.45.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.46.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.46.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.46.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.47.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.47.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.47.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.48.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.48.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.48.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.49.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.49.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.49.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.5.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.5.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.5.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.50.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.50.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.50.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.51.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.51.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.51.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.52.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.52.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.52.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.53.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.53.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.53.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.54.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.54.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.54.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.55.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.55.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.55.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.56.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.56.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.56.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.57.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.57.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.57.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.58.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.58.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.58.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.59.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.59.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.59.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.6.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.6.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.6.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.60.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.60.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.60.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.61.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.61.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.61.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.62.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.62.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.62.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.63.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.63.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.63.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.64.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.64.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.64.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.65.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.65.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.65.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.66.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.66.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.66.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.67.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.67.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.67.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.68.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.68.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.68.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.69.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.69.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.69.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.7.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.7.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.7.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.70.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.70.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.70.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.71.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.71.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.71.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.72.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.72.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.72.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.73.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.73.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.73.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.74.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.74.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.74.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.75.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.75.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.75.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.76.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.76.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.76.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.77.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.77.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.77.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.78.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.78.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.78.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.79.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.79.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.79.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.8.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.8.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.8.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.80.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.80.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.80.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.81.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.81.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.81.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.82.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.82.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.82.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.83.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.83.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.83.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.84.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.84.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.84.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.85.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.85.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.85.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.86.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.86.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.86.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.87.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.87.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.87.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.88.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.88.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.88.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.89.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.89.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.89.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.9.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.9.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.9.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.90.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.90.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.91.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.91.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.91.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.92.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.92.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.92.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.93.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.93.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.93.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.94.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.94.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.94.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.95.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.95.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.95.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.96.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.96.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.96.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.97.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.97.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.97.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.98.down_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.98.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.98.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.99.down_proj": { "bits": 3 }, "model.language_model.layers.16.mlp.experts.99.gate_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.experts.99.up_proj": { "bits": 2 }, "model.language_model.layers.16.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.16.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.16.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.16.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.16.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.16.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.17.mlp.experts.0.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.0.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.0.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.1.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.1.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.1.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.10.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.10.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.10.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.100.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.100.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.100.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.101.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.101.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.101.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.102.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.102.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.102.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.103.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.103.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.103.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.104.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.104.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.104.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.105.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.105.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.105.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.106.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.106.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.106.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.107.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.107.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.107.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.108.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.108.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.108.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.109.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.109.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.109.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.11.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.11.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.11.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.110.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.110.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.110.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.111.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.111.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.111.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.112.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.112.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.112.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.113.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.113.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.113.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.114.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.114.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.114.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.115.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.115.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.115.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.116.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.116.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.116.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.117.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.117.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.117.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.118.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.119.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.119.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.119.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.12.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.12.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.12.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.120.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.120.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.120.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.121.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.121.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.121.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.122.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.122.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.122.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.123.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.123.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.123.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.124.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.124.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.124.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.125.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.125.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.125.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.126.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.126.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.126.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.127.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.127.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.127.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.13.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.14.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.14.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.14.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.15.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.15.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.15.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.16.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.16.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.16.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.17.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.17.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.17.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.18.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.18.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.18.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.19.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.19.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.19.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.2.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.2.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.2.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.20.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.20.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.20.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.21.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.21.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.21.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.22.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.22.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.22.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.23.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.23.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.23.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.24.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.24.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.24.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.25.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.25.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.25.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.26.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.26.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.26.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.27.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.27.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.27.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.28.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.28.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.28.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.29.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.29.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.29.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.3.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.3.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.3.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.30.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.30.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.30.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.31.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.31.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.31.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.32.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.32.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.32.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.33.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.33.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.33.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.34.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.34.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.34.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.35.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.35.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.35.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.36.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.36.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.36.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.37.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.37.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.37.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.38.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.38.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.38.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.39.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.39.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.39.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.4.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.4.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.4.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.40.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.40.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.40.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.41.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.41.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.41.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.42.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.42.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.42.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.43.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.43.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.43.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.44.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.44.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.44.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.45.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.45.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.45.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.46.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.46.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.46.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.47.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.47.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.47.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.48.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.48.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.48.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.49.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.49.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.49.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.5.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.5.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.5.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.50.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.50.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.50.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.51.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.51.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.51.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.52.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.52.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.52.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.53.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.53.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.53.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.54.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.54.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.54.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.55.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.55.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.55.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.56.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.56.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.56.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.57.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.57.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.57.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.58.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.58.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.58.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.59.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.59.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.59.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.6.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.6.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.6.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.60.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.60.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.60.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.61.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.61.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.61.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.62.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.62.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.62.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.63.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.63.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.63.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.64.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.64.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.64.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.65.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.65.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.65.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.66.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.66.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.66.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.67.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.67.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.67.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.68.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.68.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.68.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.69.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.69.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.69.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.7.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.7.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.7.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.70.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.70.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.70.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.71.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.71.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.71.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.72.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.72.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.72.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.73.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.73.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.73.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.74.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.74.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.74.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.75.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.75.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.75.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.76.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.76.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.76.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.77.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.77.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.77.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.78.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.78.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.78.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.79.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.79.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.79.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.8.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.8.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.8.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.80.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.80.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.80.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.81.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.81.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.81.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.82.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.82.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.82.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.83.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.83.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.83.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.84.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.84.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.84.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.17.mlp.experts.85.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.85.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.86.down_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.86.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.86.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.87.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.87.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.87.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.88.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.88.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.88.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.89.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.89.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.89.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.9.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.9.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.9.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.90.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.90.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.90.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.91.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.91.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.91.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.92.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.92.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.92.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.93.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.93.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.93.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.94.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.94.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.94.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.95.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.95.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.95.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.96.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.96.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.96.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.97.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.97.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.97.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.98.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.98.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.98.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.99.down_proj": { "bits": 3 }, "model.language_model.layers.17.mlp.experts.99.gate_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.experts.99.up_proj": { "bits": 2 }, "model.language_model.layers.17.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.17.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.17.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.17.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.17.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.17.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.18.mlp.experts.0.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.0.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.0.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.1.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.1.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.1.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.10.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.10.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.10.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.100.down_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.100.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.18.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.102.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.102.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.102.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.103.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.103.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.103.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.104.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.104.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.104.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.105.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.105.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.105.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.106.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.106.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.106.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.107.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.107.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.108.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.108.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.109.down_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.109.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.11.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.11.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.11.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.110.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.110.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.110.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.111.down_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.111.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.111.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.112.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.112.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.112.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.113.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.113.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.113.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.114.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.114.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.115.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.115.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.115.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.18.mlp.experts.116.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.117.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.117.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.117.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.118.down_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.118.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.118.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.119.down_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.119.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.119.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.12.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.12.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.12.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.120.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.120.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.120.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.121.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.121.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.121.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.122.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.122.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.122.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.123.down_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.123.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.123.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.124.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.124.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.124.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.125.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.125.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.125.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.126.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.126.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.126.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.127.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.127.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.127.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.13.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.13.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.13.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.14.down_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.14.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.15.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.15.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.15.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.16.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.16.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.16.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.17.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.17.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.17.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.18.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.18.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.18.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.19.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.19.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.19.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.2.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.2.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.20.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.20.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.20.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.21.down_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.21.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.21.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.22.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.22.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.22.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.23.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.23.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.23.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.24.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.24.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.24.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.25.down_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.25.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.25.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.26.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.26.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.26.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.27.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.27.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.27.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.28.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.28.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.28.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.29.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.29.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.29.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.3.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.3.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.3.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.30.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.30.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.30.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.31.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.31.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.31.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.32.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.32.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.33.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.33.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.33.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.34.down_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.34.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.34.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.35.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.35.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.35.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.36.down_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.36.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.36.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.37.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.37.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.37.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.38.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.38.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.38.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.39.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.39.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.39.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.4.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.4.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.4.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.40.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.40.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.40.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.41.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.42.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.42.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.42.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.43.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.43.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.43.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.44.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.44.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.44.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.45.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.46.down_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.46.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.46.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.47.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.47.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.47.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.48.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.48.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.48.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.49.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.49.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.49.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.5.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.5.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.5.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.50.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.50.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.50.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.51.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.51.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.51.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.52.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.52.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.52.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.53.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.53.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.53.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.54.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.54.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.54.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.55.down_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.55.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.55.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.56.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.56.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.56.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.57.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.57.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.57.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.58.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.58.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.58.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.59.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.59.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.59.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.6.down_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.6.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.6.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.60.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.60.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.60.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.61.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.61.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.61.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.62.down_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.62.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.62.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.63.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.63.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.63.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.64.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.64.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.64.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.65.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.65.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.65.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.66.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.66.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.67.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.67.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.67.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.68.down_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.69.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.69.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.69.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.7.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.7.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.7.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.70.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.70.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.70.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.71.down_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.71.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.71.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.72.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.72.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.72.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.73.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.73.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.73.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.74.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.74.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.75.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.75.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.75.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.76.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.76.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.76.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.77.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.77.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.77.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.78.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.78.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.78.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.79.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.79.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.79.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.8.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.8.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.8.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.80.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.80.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.80.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.81.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.81.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.81.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.82.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.82.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.82.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.83.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.83.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.83.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.84.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.84.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.84.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.85.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.85.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.85.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.86.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.86.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.87.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.87.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.87.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.88.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.88.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.88.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.89.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.89.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.89.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.9.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.9.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.9.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.90.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.90.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.90.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.91.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.91.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.91.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.92.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.92.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.92.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.93.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.93.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.93.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.94.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.94.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.94.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.95.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.95.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.95.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.96.down_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.96.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.96.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.97.down_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.97.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.97.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.98.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.98.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.98.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.99.down_proj": { "bits": 3 }, "model.language_model.layers.18.mlp.experts.99.gate_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.experts.99.up_proj": { "bits": 2 }, "model.language_model.layers.18.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.18.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.18.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.18.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.18.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.18.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.19.mlp.experts.0.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.0.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.0.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.1.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.1.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.1.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.10.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.10.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.10.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.100.down_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.100.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.19.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.102.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.103.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.103.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.103.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.104.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.105.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.105.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.106.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.107.down_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.107.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.107.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.108.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.108.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.108.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.109.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.109.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.11.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.110.down_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.110.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.111.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.111.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.112.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.113.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.113.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.113.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.114.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.114.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.114.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.115.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.115.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.115.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.116.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.116.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.117.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.117.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.118.down_proj": { "bits": 4 }, "model.language_model.layers.19.mlp.experts.118.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.119.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.12.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.120.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.121.down_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.121.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.122.down_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.123.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.123.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.124.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.124.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.124.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.125.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.125.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.126.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.126.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.127.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.127.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.127.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.13.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.13.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.13.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.14.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.14.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.19.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.16.down_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.16.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.17.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.17.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.18.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.18.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.19.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.2.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.20.down_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.20.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.21.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.21.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.22.down_proj": { "bits": 4 }, "model.language_model.layers.19.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.23.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.23.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.23.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.24.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.24.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.25.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.25.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.26.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.26.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.27.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.28.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.19.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.19.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.30.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.30.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.30.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.31.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.31.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.19.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.33.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.34.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.35.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.35.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.36.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.36.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.36.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.37.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.37.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.38.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.38.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.38.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.39.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.39.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.39.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.4.down_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.4.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.4.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.40.down_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.40.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.41.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.41.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.41.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.42.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.43.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.43.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.43.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.44.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.44.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.44.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.45.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.45.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.46.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.46.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.46.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.47.down_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.47.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.47.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.48.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.49.down_proj": { "bits": 4 }, "model.language_model.layers.19.mlp.experts.49.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.5.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.50.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.50.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.50.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.51.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.51.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.51.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.52.down_proj": { "bits": 4 }, "model.language_model.layers.19.mlp.experts.52.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.52.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.53.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.53.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.53.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.54.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.54.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.54.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.55.down_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.55.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.56.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.56.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.57.down_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.57.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.57.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.58.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.58.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.59.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.6.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.6.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.6.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.60.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.61.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.61.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.62.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.62.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.63.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.63.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.63.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.64.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.64.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.65.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.65.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.65.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.66.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.66.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.67.down_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.67.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.68.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.68.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.68.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.69.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.69.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.7.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.70.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.70.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.71.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.71.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.72.down_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.72.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.73.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.74.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.74.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.19.mlp.experts.75.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.76.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.77.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.77.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.78.down_proj": { "bits": 4 }, "model.language_model.layers.19.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.79.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.79.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.19.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.80.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.81.down_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.81.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.82.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.82.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.83.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.83.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.83.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.84.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.84.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.84.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.85.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.85.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.86.down_proj": { "bits": 4 }, "model.language_model.layers.19.mlp.experts.86.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.87.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.87.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.88.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.19.mlp.experts.89.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.9.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.90.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.90.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.90.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.91.down_proj": { "bits": 4 }, "model.language_model.layers.19.mlp.experts.91.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.92.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.92.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.93.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.93.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.93.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.94.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.94.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.95.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.95.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.19.mlp.experts.96.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.97.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.97.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.97.up_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.98.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.98.gate_proj": { "bits": 2 }, "model.language_model.layers.19.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.99.down_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.19.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.19.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.19.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.19.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.19.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.19.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.2.mlp.down_proj": { "bits": 4 }, "model.language_model.layers.2.mlp.gate_up_proj": { "bits": 4 }, "model.language_model.layers.2.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.2.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.2.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.2.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.1.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.10.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.100.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.101.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.103.down_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.104.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.105.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.106.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.107.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.108.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.108.gate_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.108.up_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.109.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.111.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.112.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.113.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.114.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.115.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.117.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.118.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.119.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.120.down_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.120.gate_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.120.up_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.121.down_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.123.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.124.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.125.gate_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.126.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.13.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.14.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.16.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.17.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.18.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.19.down_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.19.gate_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.2.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.20.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.21.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.22.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.23.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.25.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.26.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.27.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.28.down_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.28.gate_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.3.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.30.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.31.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.34.gate_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.35.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.36.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.37.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.38.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.39.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.4.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.40.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.41.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.42.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.43.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.44.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.45.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.46.down_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.46.gate_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.47.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.49.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.49.gate_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.5.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.50.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.50.gate_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.50.up_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.51.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.52.down_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.52.gate_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.53.down_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.54.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.55.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.56.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.57.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.58.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.58.gate_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.59.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.6.down_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.6.gate_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.60.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.62.gate_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.62.up_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.63.down_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.63.gate_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.63.up_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.64.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.66.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.67.down_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.67.gate_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.68.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.68.gate_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.70.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.71.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.72.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.73.down_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.74.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.76.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.77.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.78.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.8.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.80.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.81.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.82.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.83.down_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.84.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.85.gate_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.86.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.87.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.88.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.9.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.90.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.91.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.91.gate_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.92.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.92.gate_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.93.down_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.93.gate_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.95.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.96.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.97.down_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.98.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.99.down_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.experts.99.gate_proj": { "bits": 2 }, "model.language_model.layers.20.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.20.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.20.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.20.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.20.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.20.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.20.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.0.gate_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.1.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.10.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.100.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.101.down_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.103.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.104.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.105.down_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.105.gate_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.106.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.107.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.109.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.11.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.111.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.112.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.114.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.115.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.117.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.118.down_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.118.gate_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.118.up_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.119.down_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.120.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.121.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.122.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.123.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.124.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.124.gate_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.125.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.126.down_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.127.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.13.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.14.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.15.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.16.down_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.17.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.19.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.2.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.20.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.21.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.22.down_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.22.gate_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.22.up_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.23.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.26.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.27.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.30.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.31.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.31.gate_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.33.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.36.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.37.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.38.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.40.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.41.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.41.gate_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.42.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.44.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.46.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.47.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.49.down_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.49.gate_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.5.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.50.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.51.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.52.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.54.gate_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.55.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.57.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.58.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.59.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.6.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.60.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.65.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.67.down_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.68.down_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.7.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.7.gate_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.70.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.71.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.72.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.72.gate_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.73.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.74.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.76.down_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.76.gate_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.77.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.78.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.82.down_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.86.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.87.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.87.gate_proj": { "bits": 2 }, "model.language_model.layers.21.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.88.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.9.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.90.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.91.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.92.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.93.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.95.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.97.down_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.21.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.21.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.21.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.21.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.21.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.21.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.1.down_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.10.down_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.100.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.101.down_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.101.gate_proj": { "bits": 2 }, "model.language_model.layers.22.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.104.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.105.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.106.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.107.down_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.111.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.113.down_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.114.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.115.down_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.117.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.118.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.119.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.120.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.120.gate_proj": { "bits": 2 }, "model.language_model.layers.22.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.121.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.123.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.124.down_proj": { "bits": 2 }, "model.language_model.layers.22.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.125.down_proj": { "bits": 2 }, "model.language_model.layers.22.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.126.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.13.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.14.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.16.down_proj": { "bits": 2 }, "model.language_model.layers.22.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.17.down_proj": { "bits": 2 }, "model.language_model.layers.22.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.18.down_proj": { "bits": 2 }, "model.language_model.layers.22.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.19.down_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.2.down_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.20.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.21.down_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.22.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.26.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.27.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.28.down_proj": { "bits": 2 }, "model.language_model.layers.22.mlp.experts.28.gate_proj": { "bits": 2 }, "model.language_model.layers.22.mlp.experts.28.up_proj": { "bits": 2 }, "model.language_model.layers.22.mlp.experts.29.down_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.30.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.31.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.35.down_proj": { "bits": 2 }, "model.language_model.layers.22.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.36.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.37.down_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.38.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.4.down_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.40.down_proj": { "bits": 2 }, "model.language_model.layers.22.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.41.down_proj": { "bits": 2 }, "model.language_model.layers.22.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.42.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.44.down_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.44.gate_proj": { "bits": 2 }, "model.language_model.layers.22.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.46.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.47.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.49.down_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.50.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.51.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.52.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.55.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.56.down_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.57.down_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.58.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.59.down_proj": { "bits": 2 }, "model.language_model.layers.22.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.6.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.60.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.64.down_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.67.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.68.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.71.down_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.72.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.73.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.74.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.76.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.77.down_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.78.down_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.80.down_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.82.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.83.down_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.86.down_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.87.down_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.88.down_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.9.down_proj": { "bits": 2 }, "model.language_model.layers.22.mlp.experts.9.gate_proj": { "bits": 2 }, "model.language_model.layers.22.mlp.experts.9.up_proj": { "bits": 2 }, "model.language_model.layers.22.mlp.experts.90.down_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.91.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.92.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.93.down_proj": { "bits": 2 }, "model.language_model.layers.22.mlp.experts.93.gate_proj": { "bits": 2 }, "model.language_model.layers.22.mlp.experts.93.up_proj": { "bits": 2 }, "model.language_model.layers.22.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.95.down_proj": { "bits": 2 }, "model.language_model.layers.22.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.97.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.22.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.22.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.22.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.22.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.22.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.22.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.10.down_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.100.down_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.104.down_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.105.down_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.106.down_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.107.down_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.11.down_proj": { "bits": 2 }, "model.language_model.layers.23.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.111.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.114.down_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.115.down_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.117.down_proj": { "bits": 2 }, "model.language_model.layers.23.mlp.experts.117.gate_proj": { "bits": 2 }, "model.language_model.layers.23.mlp.experts.117.up_proj": { "bits": 2 }, "model.language_model.layers.23.mlp.experts.118.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.119.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.120.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.121.down_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.122.down_proj": { "bits": 2 }, "model.language_model.layers.23.mlp.experts.122.gate_proj": { "bits": 2 }, "model.language_model.layers.23.mlp.experts.122.up_proj": { "bits": 2 }, "model.language_model.layers.23.mlp.experts.123.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.124.down_proj": { "bits": 2 }, "model.language_model.layers.23.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.126.down_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.127.down_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.13.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.14.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.16.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.17.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.18.down_proj": { "bits": 2 }, "model.language_model.layers.23.mlp.experts.18.gate_proj": { "bits": 2 }, "model.language_model.layers.23.mlp.experts.18.up_proj": { "bits": 2 }, "model.language_model.layers.23.mlp.experts.19.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.2.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.20.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.21.down_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.22.down_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.26.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.27.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.29.down_proj": { "bits": 2 }, "model.language_model.layers.23.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.30.down_proj": { "bits": 2 }, "model.language_model.layers.23.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.31.down_proj": { "bits": 2 }, "model.language_model.layers.23.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.32.down_proj": { "bits": 2 }, "model.language_model.layers.23.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.36.down_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.37.down_proj": { "bits": 2 }, "model.language_model.layers.23.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.38.down_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.40.down_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.42.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.44.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.46.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.47.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.49.down_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.50.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.51.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.52.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.53.down_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.55.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.57.down_proj": { "bits": 2 }, "model.language_model.layers.23.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.58.down_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.59.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.6.down_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.60.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.61.down_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.67.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.68.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.71.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.72.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.73.down_proj": { "bits": 2 }, "model.language_model.layers.23.mlp.experts.73.gate_proj": { "bits": 2 }, "model.language_model.layers.23.mlp.experts.73.up_proj": { "bits": 2 }, "model.language_model.layers.23.mlp.experts.74.down_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.76.down_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.77.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.78.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.8.down_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.81.down_proj": { "bits": 2 }, "model.language_model.layers.23.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.82.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.85.down_proj": { "bits": 2 }, "model.language_model.layers.23.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.86.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.87.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.88.down_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.9.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.90.down_proj": { "bits": 2 }, "model.language_model.layers.23.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.91.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.92.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.93.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.95.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.96.down_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.97.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.23.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.23.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.23.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.23.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.23.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.23.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.0.down_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.10.down_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.100.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.102.down_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.103.down_proj": { "bits": 2 }, "model.language_model.layers.24.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.104.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.105.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.106.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.107.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.111.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.112.down_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.114.down_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.115.down_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.117.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.118.down_proj": { "bits": 2 }, "model.language_model.layers.24.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.119.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.12.down_proj": { "bits": 2 }, "model.language_model.layers.24.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.120.down_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.121.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.123.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.124.down_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.126.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.13.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.14.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.16.down_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.17.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.19.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.2.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.20.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.21.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.22.down_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.22.gate_proj": { "bits": 2 }, "model.language_model.layers.24.mlp.experts.22.up_proj": { "bits": 2 }, "model.language_model.layers.24.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.26.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.27.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.3.down_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.30.down_proj": { "bits": 2 }, "model.language_model.layers.24.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.31.down_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.36.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.37.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.38.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.40.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.42.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.44.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.45.down_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.46.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.47.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.49.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.50.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.51.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.52.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.55.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.57.down_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.58.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.59.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.6.down_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.60.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.61.down_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.67.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.68.down_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.7.down_proj": { "bits": 2 }, "model.language_model.layers.24.mlp.experts.7.gate_proj": { "bits": 2 }, "model.language_model.layers.24.mlp.experts.7.up_proj": { "bits": 2 }, "model.language_model.layers.24.mlp.experts.70.down_proj": { "bits": 2 }, "model.language_model.layers.24.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.71.down_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.72.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.73.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.74.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.76.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.77.down_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.78.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.8.down_proj": { "bits": 2 }, "model.language_model.layers.24.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.81.down_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.82.down_proj": { "bits": 2 }, "model.language_model.layers.24.mlp.experts.82.gate_proj": { "bits": 2 }, "model.language_model.layers.24.mlp.experts.82.up_proj": { "bits": 2 }, "model.language_model.layers.24.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.86.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.87.down_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.88.down_proj": { "bits": 2 }, "model.language_model.layers.24.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.9.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.90.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.91.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.92.down_proj": { "bits": 2 }, "model.language_model.layers.24.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.93.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.94.down_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.95.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.96.down_proj": { "bits": 2 }, "model.language_model.layers.24.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.97.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.24.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.24.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.24.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.24.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.24.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.24.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.0.down_proj": { "bits": 2 }, "model.language_model.layers.25.mlp.experts.0.gate_proj": { "bits": 2 }, "model.language_model.layers.25.mlp.experts.0.up_proj": { "bits": 2 }, "model.language_model.layers.25.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.10.down_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.100.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.104.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.105.down_proj": { "bits": 2 }, "model.language_model.layers.25.mlp.experts.105.gate_proj": { "bits": 2 }, "model.language_model.layers.25.mlp.experts.105.up_proj": { "bits": 2 }, "model.language_model.layers.25.mlp.experts.106.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.107.down_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.11.down_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.111.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.112.down_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.114.down_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.115.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.116.down_proj": { "bits": 2 }, "model.language_model.layers.25.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.117.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.118.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.119.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.12.down_proj": { "bits": 2 }, "model.language_model.layers.25.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.120.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.121.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.123.down_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.124.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.126.down_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.13.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.14.down_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.16.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.17.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.19.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.2.down_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.20.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.21.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.22.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.25.down_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.26.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.27.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.3.down_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.30.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.31.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.36.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.37.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.38.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.40.down_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.42.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.44.down_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.46.down_proj": { "bits": 2 }, "model.language_model.layers.25.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.47.down_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.49.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.5.down_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.50.down_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.51.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.52.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.55.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.57.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.58.down_proj": { "bits": 2 }, "model.language_model.layers.25.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.59.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.6.down_proj": { "bits": 2 }, "model.language_model.layers.25.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.60.down_proj": { "bits": 2 }, "model.language_model.layers.25.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.62.down_proj": { "bits": 2 }, "model.language_model.layers.25.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.67.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.68.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.70.down_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.71.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.72.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.73.down_proj": { "bits": 2 }, "model.language_model.layers.25.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.74.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.76.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.77.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.78.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.82.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.83.down_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.86.down_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.87.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.88.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.9.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.90.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.91.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.92.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.93.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.95.down_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.97.down_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.25.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.25.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.25.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.25.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.25.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.25.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.10.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.100.down_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.104.down_proj": { "bits": 2 }, "model.language_model.layers.26.mlp.experts.104.gate_proj": { "bits": 2 }, "model.language_model.layers.26.mlp.experts.104.up_proj": { "bits": 2 }, "model.language_model.layers.26.mlp.experts.105.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.106.down_proj": { "bits": 2 }, "model.language_model.layers.26.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.107.down_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.111.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.114.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.115.down_proj": { "bits": 2 }, "model.language_model.layers.26.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.116.down_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.117.down_proj": { "bits": 2 }, "model.language_model.layers.26.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.118.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.119.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.120.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.121.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.123.down_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.124.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.126.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.13.down_proj": { "bits": 2 }, "model.language_model.layers.26.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.14.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.16.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.17.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.19.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.2.down_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.20.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.21.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.22.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.25.down_proj": { "bits": 2 }, "model.language_model.layers.26.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.26.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.27.down_proj": { "bits": 2 }, "model.language_model.layers.26.mlp.experts.27.gate_proj": { "bits": 2 }, "model.language_model.layers.26.mlp.experts.27.up_proj": { "bits": 2 }, "model.language_model.layers.26.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.3.down_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.30.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.31.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.32.down_proj": { "bits": 2 }, "model.language_model.layers.26.mlp.experts.32.gate_proj": { "bits": 2 }, "model.language_model.layers.26.mlp.experts.32.up_proj": { "bits": 2 }, "model.language_model.layers.26.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.34.down_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.36.down_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.37.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.38.down_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.40.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.41.down_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.42.down_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.44.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.46.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.47.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.49.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.50.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.51.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.52.down_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.55.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.57.down_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.58.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.59.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.6.down_proj": { "bits": 2 }, "model.language_model.layers.26.mlp.experts.6.gate_proj": { "bits": 2 }, "model.language_model.layers.26.mlp.experts.6.up_proj": { "bits": 2 }, "model.language_model.layers.26.mlp.experts.60.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.61.down_proj": { "bits": 2 }, "model.language_model.layers.26.mlp.experts.61.gate_proj": { "bits": 2 }, "model.language_model.layers.26.mlp.experts.61.up_proj": { "bits": 2 }, "model.language_model.layers.26.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.67.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.68.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.69.down_proj": { "bits": 2 }, "model.language_model.layers.26.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.71.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.72.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.73.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.74.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.76.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.77.down_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.78.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.81.down_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.82.down_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.84.down_proj": { "bits": 2 }, "model.language_model.layers.26.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.86.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.87.down_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.88.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.9.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.90.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.91.down_proj": { "bits": 2 }, "model.language_model.layers.26.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.92.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.93.down_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.95.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.97.down_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.99.down_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.26.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.26.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.26.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.26.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.26.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.26.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.1.down_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.10.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.100.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.103.down_proj": { "bits": 8 }, "model.language_model.layers.27.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.104.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.105.down_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.106.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.107.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.11.down_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.111.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.114.down_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.115.down_proj": { "bits": 2 }, "model.language_model.layers.27.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.116.down_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.117.down_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.118.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.119.down_proj": { "bits": 2 }, "model.language_model.layers.27.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.120.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.120.gate_proj": { "bits": 2 }, "model.language_model.layers.27.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.121.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.122.down_proj": { "bits": 2 }, "model.language_model.layers.27.mlp.experts.122.gate_proj": { "bits": 2 }, "model.language_model.layers.27.mlp.experts.122.up_proj": { "bits": 2 }, "model.language_model.layers.27.mlp.experts.123.down_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.124.down_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.125.down_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.126.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.13.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.14.down_proj": { "bits": 2 }, "model.language_model.layers.27.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.16.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.17.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.19.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.2.down_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.20.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.21.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.22.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.25.down_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.26.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.27.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.30.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.31.down_proj": { "bits": 2 }, "model.language_model.layers.27.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.33.down_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.34.down_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.36.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.37.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.38.down_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.4.down_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.40.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.42.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.44.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.46.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.47.down_proj": { "bits": 2 }, "model.language_model.layers.27.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.49.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.50.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.51.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.52.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.55.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.57.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.58.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.59.down_proj": { "bits": 2 }, "model.language_model.layers.27.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.6.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.60.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.67.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.68.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.69.down_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.71.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.72.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.73.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.74.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.76.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.77.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.78.down_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.81.down_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.82.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.85.down_proj": { "bits": 2 }, "model.language_model.layers.27.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.86.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.87.down_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.88.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.9.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.90.down_proj": { "bits": 2 }, "model.language_model.layers.27.mlp.experts.90.gate_proj": { "bits": 2 }, "model.language_model.layers.27.mlp.experts.90.up_proj": { "bits": 2 }, "model.language_model.layers.27.mlp.experts.91.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.92.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.93.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.95.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.96.down_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.97.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.99.down_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.27.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.27.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.27.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.27.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.27.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.27.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.0.down_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.10.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.100.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.104.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.105.down_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.106.down_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.107.down_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.111.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.112.down_proj": { "bits": 2 }, "model.language_model.layers.28.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.114.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.115.down_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.117.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.118.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.119.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.120.down_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.121.down_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.123.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.124.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.125.down_proj": { "bits": 2 }, "model.language_model.layers.28.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.126.down_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.13.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.14.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.16.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.17.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.18.down_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.19.down_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.2.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.20.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.21.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.22.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.24.down_proj": { "bits": 2 }, "model.language_model.layers.28.mlp.experts.24.gate_proj": { "bits": 2 }, "model.language_model.layers.28.mlp.experts.24.up_proj": { "bits": 2 }, "model.language_model.layers.28.mlp.experts.25.down_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.26.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.27.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.30.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.31.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.35.down_proj": { "bits": 2 }, "model.language_model.layers.28.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.36.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.37.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.38.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.40.down_proj": { "bits": 2 }, "model.language_model.layers.28.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.42.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.44.down_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.46.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.47.down_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.49.down_proj": { "bits": 8 }, "model.language_model.layers.28.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.50.down_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.51.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.52.down_proj": { "bits": 2 }, "model.language_model.layers.28.mlp.experts.52.gate_proj": { "bits": 2 }, "model.language_model.layers.28.mlp.experts.52.up_proj": { "bits": 2 }, "model.language_model.layers.28.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.55.down_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.56.down_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.57.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.58.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.59.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.6.down_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.60.down_proj": { "bits": 2 }, "model.language_model.layers.28.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.61.down_proj": { "bits": 2 }, "model.language_model.layers.28.mlp.experts.61.gate_proj": { "bits": 2 }, "model.language_model.layers.28.mlp.experts.61.up_proj": { "bits": 2 }, "model.language_model.layers.28.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.65.down_proj": { "bits": 2 }, "model.language_model.layers.28.mlp.experts.65.gate_proj": { "bits": 2 }, "model.language_model.layers.28.mlp.experts.65.up_proj": { "bits": 2 }, "model.language_model.layers.28.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.67.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.68.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.71.down_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.72.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.73.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.74.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.76.down_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.77.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.78.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.80.down_proj": { "bits": 2 }, "model.language_model.layers.28.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.82.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.83.down_proj": { "bits": 2 }, "model.language_model.layers.28.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.86.down_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.87.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.88.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.9.down_proj": { "bits": 2 }, "model.language_model.layers.28.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.90.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.91.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.92.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.93.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.94.down_proj": { "bits": 2 }, "model.language_model.layers.28.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.95.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.97.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.28.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.28.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.28.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.28.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.28.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.28.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.10.down_proj": { "bits": 2 }, "model.language_model.layers.29.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.100.down_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.104.down_proj": { "bits": 2 }, "model.language_model.layers.29.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.105.down_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.106.down_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.107.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.109.down_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.111.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.114.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.115.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.117.down_proj": { "bits": 2 }, "model.language_model.layers.29.mlp.experts.117.gate_proj": { "bits": 2 }, "model.language_model.layers.29.mlp.experts.117.up_proj": { "bits": 2 }, "model.language_model.layers.29.mlp.experts.118.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.119.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.120.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.121.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.123.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.124.down_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.126.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.13.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.14.down_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.16.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.17.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.19.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.2.down_proj": { "bits": 8 }, "model.language_model.layers.29.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.20.down_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.21.down_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.22.down_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.26.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.27.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.28.down_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.30.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.31.down_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.36.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.37.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.38.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.40.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.41.down_proj": { "bits": 2 }, "model.language_model.layers.29.mlp.experts.41.gate_proj": { "bits": 2 }, "model.language_model.layers.29.mlp.experts.41.up_proj": { "bits": 2 }, "model.language_model.layers.29.mlp.experts.42.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.44.down_proj": { "bits": 2 }, "model.language_model.layers.29.mlp.experts.44.gate_proj": { "bits": 2 }, "model.language_model.layers.29.mlp.experts.44.up_proj": { "bits": 2 }, "model.language_model.layers.29.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.46.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.47.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.49.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.50.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.51.down_proj": { "bits": 2 }, "model.language_model.layers.29.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.52.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.53.down_proj": { "bits": 2 }, "model.language_model.layers.29.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.55.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.57.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.58.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.59.down_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.6.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.60.down_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.62.down_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.66.down_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.67.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.68.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.71.down_proj": { "bits": 2 }, "model.language_model.layers.29.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.72.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.73.down_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.74.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.76.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.77.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.78.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.82.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.86.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.87.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.88.down_proj": { "bits": 2 }, "model.language_model.layers.29.mlp.experts.88.gate_proj": { "bits": 2 }, "model.language_model.layers.29.mlp.experts.88.up_proj": { "bits": 2 }, "model.language_model.layers.29.mlp.experts.89.down_proj": { "bits": 2 }, "model.language_model.layers.29.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.9.down_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.90.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.91.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.92.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.93.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.95.down_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.97.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.29.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.29.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.29.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.29.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.29.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.29.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.3.mlp.experts.0.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.0.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.0.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.1.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.1.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.1.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.10.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.10.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.10.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.100.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.100.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.100.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.101.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.101.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.101.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.102.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.102.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.102.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.103.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.103.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.103.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.104.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.104.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.104.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.105.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.105.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.105.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.106.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.106.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.106.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.107.down_proj": { "bits": 3 }, "model.language_model.layers.3.mlp.experts.107.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.107.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.108.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.108.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.108.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.109.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.109.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.109.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.11.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.11.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.11.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.110.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.110.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.110.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.111.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.111.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.111.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.112.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.112.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.112.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.113.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.113.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.113.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.114.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.114.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.114.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.115.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.115.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.115.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.116.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.116.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.116.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.117.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.117.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.117.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.118.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.118.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.118.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.119.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.119.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.119.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.12.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.12.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.12.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.120.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.120.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.120.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.121.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.121.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.121.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.122.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.122.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.122.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.123.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.123.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.123.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.124.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.124.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.124.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.125.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.125.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.125.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.126.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.126.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.126.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.127.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.127.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.127.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.13.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.13.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.13.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.14.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.14.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.14.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.15.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.15.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.15.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.16.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.16.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.16.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.17.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.17.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.17.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.18.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.18.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.18.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.19.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.19.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.19.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.2.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.2.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.2.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.20.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.20.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.20.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.21.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.21.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.21.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.22.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.22.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.22.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.23.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.23.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.23.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.24.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.24.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.24.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.25.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.25.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.25.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.26.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.26.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.26.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.27.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.27.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.27.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.28.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.28.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.28.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.29.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.29.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.29.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.3.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.3.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.3.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.30.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.30.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.30.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.31.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.31.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.31.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.32.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.32.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.32.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.33.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.33.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.33.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.34.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.34.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.34.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.35.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.35.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.35.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.36.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.36.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.36.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.37.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.37.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.37.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.38.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.38.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.38.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.39.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.39.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.39.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.4.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.4.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.4.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.40.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.40.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.40.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.41.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.41.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.41.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.42.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.42.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.42.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.43.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.43.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.43.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.44.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.44.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.44.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.45.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.45.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.45.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.46.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.46.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.46.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.47.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.47.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.47.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.48.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.48.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.48.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.49.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.49.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.49.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.5.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.5.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.5.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.50.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.50.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.50.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.51.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.51.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.51.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.52.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.52.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.52.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.53.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.53.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.53.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.54.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.54.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.54.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.55.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.55.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.55.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.56.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.56.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.56.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.57.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.57.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.57.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.58.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.58.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.58.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.59.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.59.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.59.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.6.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.6.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.6.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.60.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.60.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.60.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.61.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.61.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.61.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.62.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.62.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.62.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.63.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.63.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.63.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.64.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.64.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.64.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.65.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.65.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.65.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.66.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.66.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.66.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.67.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.67.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.67.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.68.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.68.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.68.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.69.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.69.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.69.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.7.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.7.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.7.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.70.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.70.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.70.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.71.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.71.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.71.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.72.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.72.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.72.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.73.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.73.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.73.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.74.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.74.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.74.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.75.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.75.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.75.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.76.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.76.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.76.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.77.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.77.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.77.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.78.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.78.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.78.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.79.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.79.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.79.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.8.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.8.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.8.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.80.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.80.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.80.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.81.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.81.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.81.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.82.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.82.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.82.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.83.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.83.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.83.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.84.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.84.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.84.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.85.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.85.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.85.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.86.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.86.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.86.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.87.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.87.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.87.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.88.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.88.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.88.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.89.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.89.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.89.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.9.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.9.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.9.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.90.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.90.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.90.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.91.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.91.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.91.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.92.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.92.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.92.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.93.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.93.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.93.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.94.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.94.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.94.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.95.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.95.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.95.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.96.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.96.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.96.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.97.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.97.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.97.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.98.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.98.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.98.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.99.down_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.99.gate_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.experts.99.up_proj": { "bits": 2 }, "model.language_model.layers.3.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.3.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.3.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.3.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.3.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.3.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.10.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.100.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.104.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.105.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.106.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.107.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.111.down_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.114.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.115.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.117.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.118.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.119.down_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.120.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.121.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.123.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.124.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.125.down_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.126.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.13.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.14.down_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.15.down_proj": { "bits": 2 }, "model.language_model.layers.30.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.16.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.17.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.19.down_proj": { "bits": 2 }, "model.language_model.layers.30.mlp.experts.19.gate_proj": { "bits": 2 }, "model.language_model.layers.30.mlp.experts.19.up_proj": { "bits": 2 }, "model.language_model.layers.30.mlp.experts.2.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.20.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.21.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.22.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.26.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.27.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.30.down_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.31.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.34.down_proj": { "bits": 2 }, "model.language_model.layers.30.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.36.down_proj": { "bits": 2 }, "model.language_model.layers.30.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.37.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.38.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.40.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.42.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.44.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.46.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.47.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.49.down_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.50.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.51.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.52.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.54.down_proj": { "bits": 2 }, "model.language_model.layers.30.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.55.down_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.57.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.58.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.59.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.6.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.60.down_proj": { "bits": 2 }, "model.language_model.layers.30.mlp.experts.60.gate_proj": { "bits": 2 }, "model.language_model.layers.30.mlp.experts.60.up_proj": { "bits": 2 }, "model.language_model.layers.30.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.67.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.68.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.71.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.72.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.73.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.74.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.76.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.77.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.78.down_proj": { "bits": 8 }, "model.language_model.layers.30.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.79.down_proj": { "bits": 2 }, "model.language_model.layers.30.mlp.experts.79.gate_proj": { "bits": 2 }, "model.language_model.layers.30.mlp.experts.79.up_proj": { "bits": 2 }, "model.language_model.layers.30.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.82.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.84.down_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.85.down_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.86.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.87.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.88.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.9.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.90.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.91.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.92.down_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.93.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.94.down_proj": { "bits": 2 }, "model.language_model.layers.30.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.95.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.97.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.30.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.30.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.30.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.30.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.30.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.30.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.10.down_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.100.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.104.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.105.down_proj": { "bits": 2 }, "model.language_model.layers.31.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.106.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.107.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.110.down_proj": { "bits": 2 }, "model.language_model.layers.31.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.111.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.114.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.115.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.117.down_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.118.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.119.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.12.down_proj": { "bits": 2 }, "model.language_model.layers.31.mlp.experts.12.gate_proj": { "bits": 2 }, "model.language_model.layers.31.mlp.experts.12.up_proj": { "bits": 2 }, "model.language_model.layers.31.mlp.experts.120.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.121.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.123.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.124.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.126.down_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.13.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.14.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.16.down_proj": { "bits": 2 }, "model.language_model.layers.31.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.17.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.19.down_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.2.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.20.down_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.21.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.22.down_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.26.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.27.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.30.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.31.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.36.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.37.down_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.38.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.40.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.42.down_proj": { "bits": 2 }, "model.language_model.layers.31.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.44.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.45.down_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.46.down_proj": { "bits": 8 }, "model.language_model.layers.31.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.47.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.49.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.50.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.51.down_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.52.down_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.55.down_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.57.down_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.58.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.59.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.6.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.60.down_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.63.down_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.67.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.68.down_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.70.down_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.71.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.72.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.73.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.74.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.76.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.77.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.78.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.82.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.85.down_proj": { "bits": 2 }, "model.language_model.layers.31.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.86.down_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.87.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.88.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.9.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.90.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.91.down_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.92.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.93.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.95.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.97.down_proj": { "bits": 2 }, "model.language_model.layers.31.mlp.experts.97.gate_proj": { "bits": 2 }, "model.language_model.layers.31.mlp.experts.97.up_proj": { "bits": 2 }, "model.language_model.layers.31.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.31.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.31.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.31.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.31.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.31.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.31.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.10.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.100.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.102.down_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.103.down_proj": { "bits": 2 }, "model.language_model.layers.32.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.104.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.105.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.106.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.107.down_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.111.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.112.down_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.114.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.115.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.117.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.118.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.119.down_proj": { "bits": 8 }, "model.language_model.layers.32.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.120.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.121.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.123.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.124.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.126.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.13.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.14.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.16.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.17.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.18.down_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.19.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.2.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.20.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.21.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.22.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.24.down_proj": { "bits": 2 }, "model.language_model.layers.32.mlp.experts.24.gate_proj": { "bits": 2 }, "model.language_model.layers.32.mlp.experts.24.up_proj": { "bits": 2 }, "model.language_model.layers.32.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.26.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.27.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.3.down_proj": { "bits": 2 }, "model.language_model.layers.32.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.30.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.31.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.36.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.37.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.38.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.40.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.42.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.44.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.46.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.47.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.49.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.50.down_proj": { "bits": 2 }, "model.language_model.layers.32.mlp.experts.50.gate_proj": { "bits": 2 }, "model.language_model.layers.32.mlp.experts.50.up_proj": { "bits": 2 }, "model.language_model.layers.32.mlp.experts.51.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.52.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.55.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.57.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.58.down_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.59.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.6.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.60.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.65.down_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.67.down_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.68.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.69.down_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.71.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.72.down_proj": { "bits": 2 }, "model.language_model.layers.32.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.73.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.74.down_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.76.down_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.77.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.78.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.82.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.84.down_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.86.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.87.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.88.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.89.down_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.9.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.90.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.91.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.92.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.93.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.95.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.97.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.32.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.32.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.32.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.32.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.32.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.32.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.0.down_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.10.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.100.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.104.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.105.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.106.down_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.107.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.110.down_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.111.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.114.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.115.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.116.down_proj": { "bits": 2 }, "model.language_model.layers.33.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.117.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.118.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.119.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.120.down_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.121.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.123.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.124.down_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.126.down_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.13.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.14.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.16.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.17.down_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.19.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.2.down_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.20.down_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.21.down_proj": { "bits": 2 }, "model.language_model.layers.33.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.22.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.24.down_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.26.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.27.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.30.down_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.31.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.32.down_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.36.down_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.37.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.38.down_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.40.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.42.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.44.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.46.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.47.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.48.down_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.49.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.50.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.51.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.52.down_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.55.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.57.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.58.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.59.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.6.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.60.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.63.down_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.64.down_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.65.down_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.67.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.68.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.71.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.72.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.73.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.74.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.76.down_proj": { "bits": 2 }, "model.language_model.layers.33.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.77.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.78.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.8.down_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.82.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.86.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.87.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.88.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.9.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.90.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.91.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.92.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.93.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.95.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.97.down_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.33.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.33.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.33.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.33.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.33.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.33.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.10.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.100.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.104.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.105.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.106.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.107.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.111.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.113.down_proj": { "bits": 2 }, "model.language_model.layers.34.mlp.experts.113.gate_proj": { "bits": 2 }, "model.language_model.layers.34.mlp.experts.113.up_proj": { "bits": 2 }, "model.language_model.layers.34.mlp.experts.114.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.115.down_proj": { "bits": 2 }, "model.language_model.layers.34.mlp.experts.115.gate_proj": { "bits": 2 }, "model.language_model.layers.34.mlp.experts.115.up_proj": { "bits": 2 }, "model.language_model.layers.34.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.117.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.118.down_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.119.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.120.down_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.121.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.123.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.124.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.125.down_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.126.down_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.13.down_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.14.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.16.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.17.down_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.19.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.2.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.20.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.21.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.22.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.26.down_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.27.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.30.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.31.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.36.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.37.down_proj": { "bits": 2 }, "model.language_model.layers.34.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.38.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.40.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.41.down_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.42.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.43.down_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.44.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.46.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.47.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.49.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.5.down_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.50.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.51.down_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.52.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.53.down_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.55.down_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.57.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.58.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.59.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.6.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.60.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.62.down_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.64.down_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.67.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.68.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.71.down_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.72.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.73.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.74.down_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.76.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.77.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.78.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.79.down_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.80.down_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.82.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.85.down_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.86.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.87.down_proj": { "bits": 2 }, "model.language_model.layers.34.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.88.down_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.9.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.90.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.91.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.92.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.93.down_proj": { "bits": 2 }, "model.language_model.layers.34.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.95.down_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.97.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.34.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.34.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.34.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.34.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.34.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.34.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.0.down_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.10.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.100.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.104.down_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.105.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.106.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.107.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.110.down_proj": { "bits": 8 }, "model.language_model.layers.35.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.111.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.114.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.115.down_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.117.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.118.down_proj": { "bits": 2 }, "model.language_model.layers.35.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.119.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.120.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.121.down_proj": { "bits": 8 }, "model.language_model.layers.35.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.123.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.124.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.126.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.13.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.14.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.16.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.17.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.19.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.2.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.20.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.21.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.22.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.26.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.27.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.28.down_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.3.down_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.30.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.31.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.36.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.37.down_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.38.down_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.40.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.41.down_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.42.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.44.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.46.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.47.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.49.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.5.down_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.50.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.51.down_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.52.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.53.down_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.55.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.57.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.58.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.59.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.6.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.60.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.62.down_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.64.down_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.65.down_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.67.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.68.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.71.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.72.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.73.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.74.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.76.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.77.down_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.78.down_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.82.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.86.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.87.down_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.88.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.9.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.90.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.91.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.92.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.93.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.95.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.97.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.35.mlp.experts.99.down_proj": { "bits": 2 }, "model.language_model.layers.35.mlp.experts.99.gate_proj": { "bits": 2 }, "model.language_model.layers.35.mlp.experts.99.up_proj": { "bits": 2 }, "model.language_model.layers.35.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.35.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.35.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.35.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.35.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.35.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.10.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.100.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.104.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.105.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.106.down_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.107.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.111.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.114.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.115.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.116.down_proj": { "bits": 2 }, "model.language_model.layers.36.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.117.down_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.118.down_proj": { "bits": 8 }, "model.language_model.layers.36.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.119.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.120.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.121.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.123.down_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.124.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.126.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.13.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.14.down_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.16.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.17.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.19.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.2.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.20.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.21.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.22.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.26.down_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.27.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.30.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.31.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.36.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.37.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.38.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.40.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.42.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.44.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.46.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.47.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.49.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.50.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.51.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.52.down_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.55.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.57.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.58.down_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.59.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.6.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.60.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.67.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.68.down_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.71.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.72.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.73.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.74.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.76.down_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.77.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.78.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.82.down_proj": { "bits": 2 }, "model.language_model.layers.36.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.86.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.87.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.88.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.9.down_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.90.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.91.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.92.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.93.down_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.95.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.97.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.99.down_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.36.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.36.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.36.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.36.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.36.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.36.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.10.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.100.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.102.down_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.104.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.105.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.106.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.107.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.111.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.114.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.115.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.117.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.118.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.119.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.120.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.121.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.123.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.124.down_proj": { "bits": 2 }, "model.language_model.layers.37.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.126.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.127.down_proj": { "bits": 2 }, "model.language_model.layers.37.mlp.experts.127.gate_proj": { "bits": 2 }, "model.language_model.layers.37.mlp.experts.127.up_proj": { "bits": 2 }, "model.language_model.layers.37.mlp.experts.13.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.14.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.16.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.17.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.19.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.2.down_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.20.down_proj": { "bits": 2 }, "model.language_model.layers.37.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.21.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.22.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.26.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.27.down_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.29.down_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.30.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.31.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.36.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.37.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.38.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.40.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.42.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.44.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.46.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.47.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.49.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.50.down_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.51.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.52.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.55.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.57.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.58.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.59.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.6.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.60.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.67.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.68.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.71.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.72.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.73.down_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.74.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.76.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.77.down_proj": { "bits": 2 }, "model.language_model.layers.37.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.78.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.82.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.86.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.87.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.88.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.9.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.90.down_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.91.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.92.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.93.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.95.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.97.down_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.37.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.37.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.37.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.37.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.37.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.37.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.0.down_proj": { "bits": 2 }, "model.language_model.layers.38.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.1.down_proj": { "bits": 2 }, "model.language_model.layers.38.mlp.experts.1.gate_proj": { "bits": 2 }, "model.language_model.layers.38.mlp.experts.1.up_proj": { "bits": 2 }, "model.language_model.layers.38.mlp.experts.10.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.100.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.104.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.105.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.106.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.107.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.111.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.114.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.115.down_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.117.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.118.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.119.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.120.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.121.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.123.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.124.down_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.126.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.13.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.14.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.15.down_proj": { "bits": 2 }, "model.language_model.layers.38.mlp.experts.15.gate_proj": { "bits": 2 }, "model.language_model.layers.38.mlp.experts.15.up_proj": { "bits": 2 }, "model.language_model.layers.38.mlp.experts.16.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.17.down_proj": { "bits": 2 }, "model.language_model.layers.38.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.19.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.2.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.20.down_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.21.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.22.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.23.down_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.26.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.27.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.30.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.31.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.33.down_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.36.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.37.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.38.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.40.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.42.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.44.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.46.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.47.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.49.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.50.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.51.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.52.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.55.down_proj": { "bits": 2 }, "model.language_model.layers.38.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.56.down_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.57.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.58.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.59.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.6.down_proj": { "bits": 2 }, "model.language_model.layers.38.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.60.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.63.down_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.67.down_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.68.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.70.down_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.71.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.72.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.73.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.74.down_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.76.down_proj": { "bits": 2 }, "model.language_model.layers.38.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.77.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.78.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.82.down_proj": { "bits": 2 }, "model.language_model.layers.38.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.85.down_proj": { "bits": 2 }, "model.language_model.layers.38.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.86.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.87.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.88.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.9.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.90.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.91.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.92.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.93.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.95.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.97.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.38.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.38.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.38.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.38.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.38.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.38.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.10.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.100.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.102.down_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.104.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.105.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.106.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.107.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.111.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.114.down_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.115.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.117.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.118.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.119.down_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.120.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.121.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.123.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.124.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.126.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.13.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.14.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.16.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.17.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.19.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.2.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.20.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.21.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.22.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.24.down_proj": { "bits": 2 }, "model.language_model.layers.39.mlp.experts.24.gate_proj": { "bits": 2 }, "model.language_model.layers.39.mlp.experts.24.up_proj": { "bits": 2 }, "model.language_model.layers.39.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.26.down_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.27.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.30.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.31.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.36.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.37.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.38.down_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.39.down_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.40.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.42.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.44.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.46.down_proj": { "bits": 2 }, "model.language_model.layers.39.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.47.down_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.49.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.50.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.51.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.52.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.55.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.56.down_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.57.down_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.58.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.59.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.6.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.60.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.63.down_proj": { "bits": 2 }, "model.language_model.layers.39.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.67.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.68.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.71.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.72.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.73.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.74.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.76.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.77.down_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.78.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.82.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.86.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.87.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.88.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.9.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.90.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.91.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.92.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.93.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.95.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.97.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.39.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.39.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.39.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.39.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.39.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.39.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.4.mlp.experts.0.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.0.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.0.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.1.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.1.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.1.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.10.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.10.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.10.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.100.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.100.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.100.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.101.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.101.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.101.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.102.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.102.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.102.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.103.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.103.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.103.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.104.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.104.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.104.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.105.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.105.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.105.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.106.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.106.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.106.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.107.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.107.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.107.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.108.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.108.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.108.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.109.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.109.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.109.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.11.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.11.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.11.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.110.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.110.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.110.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.111.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.111.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.111.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.112.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.112.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.112.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.113.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.113.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.113.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.114.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.114.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.114.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.115.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.115.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.115.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.116.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.116.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.116.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.117.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.117.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.117.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.118.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.118.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.118.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.119.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.119.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.119.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.12.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.12.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.12.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.120.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.120.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.120.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.121.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.121.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.121.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.122.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.122.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.122.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.123.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.123.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.123.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.124.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.124.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.124.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.125.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.125.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.125.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.126.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.126.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.126.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.127.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.127.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.127.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.13.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.13.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.13.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.14.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.14.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.14.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.15.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.15.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.15.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.16.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.16.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.16.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.17.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.17.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.17.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.18.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.18.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.18.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.19.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.19.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.19.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.2.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.2.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.2.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.20.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.20.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.20.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.21.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.21.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.21.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.22.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.22.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.22.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.23.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.23.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.23.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.24.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.24.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.24.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.25.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.25.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.25.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.26.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.26.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.26.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.27.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.27.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.27.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.28.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.28.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.28.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.29.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.29.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.29.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.3.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.3.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.3.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.30.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.30.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.30.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.31.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.31.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.31.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.32.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.32.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.32.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.33.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.33.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.33.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.34.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.34.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.34.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.35.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.35.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.35.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.36.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.36.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.36.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.37.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.37.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.37.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.38.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.38.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.38.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.39.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.39.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.39.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.4.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.4.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.4.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.40.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.40.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.40.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.41.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.41.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.41.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.42.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.42.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.42.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.43.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.43.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.43.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.44.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.44.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.44.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.45.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.45.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.45.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.46.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.46.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.46.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.47.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.47.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.47.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.48.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.48.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.48.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.49.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.49.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.49.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.5.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.5.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.5.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.50.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.50.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.50.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.51.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.51.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.51.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.52.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.52.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.52.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.53.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.53.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.53.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.54.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.54.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.54.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.55.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.55.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.55.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.56.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.56.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.56.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.57.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.57.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.57.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.58.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.58.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.58.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.59.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.59.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.59.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.6.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.6.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.6.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.60.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.60.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.60.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.61.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.61.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.61.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.62.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.62.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.62.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.63.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.63.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.63.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.64.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.64.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.64.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.65.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.65.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.65.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.66.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.66.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.66.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.67.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.67.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.67.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.68.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.68.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.68.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.69.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.69.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.69.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.7.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.7.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.7.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.70.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.70.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.70.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.71.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.71.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.71.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.72.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.72.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.72.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.73.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.73.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.73.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.74.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.74.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.74.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.75.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.75.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.75.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.76.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.76.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.76.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.77.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.77.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.77.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.78.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.78.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.78.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.79.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.79.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.79.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.8.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.8.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.8.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.80.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.80.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.80.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.81.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.81.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.81.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.82.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.82.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.82.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.83.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.83.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.83.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.84.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.84.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.84.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.85.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.85.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.85.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.86.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.86.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.86.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.87.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.87.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.87.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.88.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.88.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.88.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.89.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.89.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.89.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.9.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.9.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.9.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.90.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.90.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.90.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.91.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.91.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.91.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.92.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.92.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.92.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.93.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.93.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.93.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.94.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.94.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.94.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.95.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.95.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.95.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.96.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.96.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.96.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.97.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.97.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.97.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.98.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.98.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.98.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.99.down_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.99.gate_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.experts.99.up_proj": { "bits": 2 }, "model.language_model.layers.4.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.4.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.4.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.4.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.4.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.4.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.10.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.100.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.104.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.105.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.106.down_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.107.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.111.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.112.down_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.114.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.115.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.117.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.118.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.119.down_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.120.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.121.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.123.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.124.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.126.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.13.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.14.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.16.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.17.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.18.down_proj": { "bits": 2 }, "model.language_model.layers.40.mlp.experts.18.gate_proj": { "bits": 2 }, "model.language_model.layers.40.mlp.experts.18.up_proj": { "bits": 2 }, "model.language_model.layers.40.mlp.experts.19.down_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.2.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.20.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.21.down_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.22.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.23.down_proj": { "bits": 2 }, "model.language_model.layers.40.mlp.experts.23.gate_proj": { "bits": 2 }, "model.language_model.layers.40.mlp.experts.23.up_proj": { "bits": 2 }, "model.language_model.layers.40.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.26.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.27.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.30.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.31.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.36.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.37.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.38.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.40.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.42.down_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.44.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.46.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.47.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.49.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.50.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.51.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.52.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.53.down_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.55.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.56.down_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.57.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.58.down_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.59.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.6.down_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.60.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.65.down_proj": { "bits": 2 }, "model.language_model.layers.40.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.67.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.68.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.71.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.72.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.73.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.74.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.76.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.77.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.78.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.82.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.86.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.87.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.88.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.9.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.90.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.91.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.92.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.93.down_proj": { "bits": 8 }, "model.language_model.layers.40.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.95.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.97.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.40.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.40.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.40.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.40.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.40.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.40.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.10.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.100.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.103.down_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.104.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.105.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.106.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.107.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.109.down_proj": { "bits": 2 }, "model.language_model.layers.41.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.111.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.114.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.115.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.117.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.118.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.119.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.120.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.121.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.123.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.124.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.126.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.127.down_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.13.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.14.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.16.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.17.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.19.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.2.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.20.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.21.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.22.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.26.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.27.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.28.down_proj": { "bits": 2 }, "model.language_model.layers.41.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.30.down_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.31.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.36.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.37.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.38.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.40.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.42.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.43.down_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.44.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.46.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.47.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.49.down_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.50.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.51.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.52.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.55.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.57.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.58.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.59.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.6.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.60.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.67.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.68.down_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.71.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.72.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.73.down_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.74.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.76.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.77.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.78.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.82.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.86.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.87.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.88.down_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.9.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.90.down_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.91.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.92.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.93.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.95.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.97.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.41.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.41.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.41.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.41.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.41.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.41.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.10.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.100.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.104.down_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.105.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.106.down_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.107.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.109.down_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.11.down_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.111.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.114.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.115.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.117.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.118.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.119.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.120.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.121.down_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.123.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.124.down_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.126.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.13.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.14.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.16.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.17.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.19.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.2.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.20.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.21.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.22.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.24.down_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.26.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.27.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.30.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.31.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.36.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.37.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.38.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.40.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.42.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.44.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.46.down_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.47.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.49.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.50.down_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.51.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.52.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.55.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.57.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.58.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.59.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.6.down_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.60.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.67.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.68.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.71.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.72.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.73.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.74.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.76.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.77.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.78.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.82.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.86.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.87.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.88.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.9.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.90.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.91.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.92.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.93.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.94.down_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.95.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.97.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.42.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.42.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.42.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.42.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.42.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.42.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.0.down_proj": { "bits": 8 }, "model.language_model.layers.43.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.10.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.100.down_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.104.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.105.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.106.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.107.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.111.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.114.down_proj": { "bits": 2 }, "model.language_model.layers.43.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.115.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.117.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.118.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.119.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.120.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.121.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.123.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.124.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.126.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.13.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.14.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.16.down_proj": { "bits": 8 }, "model.language_model.layers.43.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.17.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.19.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.2.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.20.down_proj": { "bits": 2 }, "model.language_model.layers.43.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.21.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.22.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.26.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.27.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.30.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.31.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.36.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.37.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.38.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.40.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.42.down_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.44.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.46.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.47.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.49.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.50.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.51.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.52.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.55.down_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.57.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.58.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.59.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.6.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.60.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.67.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.68.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.71.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.72.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.73.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.74.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.76.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.77.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.78.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.82.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.86.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.87.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.88.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.9.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.90.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.91.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.92.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.93.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.95.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.97.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.43.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.43.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.43.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.43.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.43.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.43.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.10.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.100.down_proj": { "bits": 8 }, "model.language_model.layers.44.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.104.down_proj": { "bits": 8 }, "model.language_model.layers.44.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.105.down_proj": { "bits": 8 }, "model.language_model.layers.44.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.106.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.107.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.111.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.114.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.115.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.116.down_proj": { "bits": 8 }, "model.language_model.layers.44.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.117.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.118.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.119.down_proj": { "bits": 2 }, "model.language_model.layers.44.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.120.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.121.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.123.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.124.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.126.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.13.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.14.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.16.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.17.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.19.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.2.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.20.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.21.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.22.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.26.down_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.27.down_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.29.down_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.30.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.31.down_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.36.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.37.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.38.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.40.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.42.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.43.up_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.44.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.46.down_proj": { "bits": 8 }, "model.language_model.layers.44.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.46.up_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.47.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.49.down_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.50.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.51.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.52.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.55.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.57.down_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.58.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.59.down_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.6.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.60.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.64.down_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.67.down_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.68.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.71.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.72.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.73.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.74.down_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.76.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.77.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.78.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.82.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.86.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.87.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.88.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.9.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.90.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.91.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.92.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.93.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.95.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.97.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.44.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.44.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.44.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.44.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.44.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.44.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.10.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.100.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.104.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.105.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.106.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.107.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.111.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.114.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.115.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.117.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.118.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.119.down_proj": { "bits": 8 }, "model.language_model.layers.45.mlp.experts.119.gate_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.119.up_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.12.up_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.120.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.121.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.123.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.124.down_proj": { "bits": 8 }, "model.language_model.layers.45.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.126.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.13.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.14.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.16.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.17.down_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.19.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.19.up_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.2.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.20.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.21.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.22.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.26.down_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.27.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.29.down_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.30.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.31.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.35.down_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.36.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.37.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.38.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.40.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.42.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.44.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.46.down_proj": { "bits": 8 }, "model.language_model.layers.45.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.47.down_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.49.down_proj": { "bits": 8 }, "model.language_model.layers.45.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.50.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.51.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.52.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.54.down_proj": { "bits": 8 }, "model.language_model.layers.45.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.55.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.56.down_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.57.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.58.down_proj": { "bits": 8 }, "model.language_model.layers.45.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.59.down_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.6.down_proj": { "bits": 8 }, "model.language_model.layers.45.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.60.down_proj": { "bits": 8 }, "model.language_model.layers.45.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.62.down_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.67.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.68.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.71.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.72.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.73.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.74.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.76.down_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.77.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.78.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.8.down_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.82.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.86.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.87.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.88.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.9.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.90.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.91.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.92.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.93.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.94.down_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.95.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.97.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.45.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.45.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.45.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.45.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.45.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.45.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.10.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.100.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.104.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.105.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.106.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.107.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.111.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.114.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.115.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.117.down_proj": { "bits": 8 }, "model.language_model.layers.46.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.117.up_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.118.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.119.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.120.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.121.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.123.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.124.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.126.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.13.down_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.14.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.16.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.17.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.19.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.2.down_proj": { "bits": 8 }, "model.language_model.layers.46.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.20.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.21.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.22.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.23.up_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.26.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.27.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.3.down_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.30.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.31.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.36.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.37.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.38.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.40.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.42.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.44.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.46.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.47.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.49.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.50.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.51.down_proj": { "bits": 8 }, "model.language_model.layers.46.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.51.up_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.52.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.54.down_proj": { "bits": 8 }, "model.language_model.layers.46.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.55.down_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.57.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.58.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.59.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.6.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.60.down_proj": { "bits": 8 }, "model.language_model.layers.46.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.61.down_proj": { "bits": 8 }, "model.language_model.layers.46.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.61.up_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.67.down_proj": { "bits": 8 }, "model.language_model.layers.46.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.67.up_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.68.down_proj": { "bits": 8 }, "model.language_model.layers.46.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.70.down_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.71.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.71.up_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.72.down_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.73.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.74.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.76.down_proj": { "bits": 8 }, "model.language_model.layers.46.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.77.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.78.down_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.79.down_proj": { "bits": 8 }, "model.language_model.layers.46.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.82.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.83.down_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.86.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.87.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.88.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.89.up_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.9.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.90.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.91.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.92.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.93.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.95.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.97.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.46.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.46.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.46.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.46.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.46.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.46.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.0.down_proj": { "bits": 8 }, "model.language_model.layers.47.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.0.up_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.10.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.100.down_proj": { "bits": 8 }, "model.language_model.layers.47.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.104.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.105.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.106.down_proj": { "bits": 8 }, "model.language_model.layers.47.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.106.up_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.107.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.110.down_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.111.down_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.114.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.114.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.115.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.117.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.118.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.119.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.120.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.121.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.123.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.124.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.126.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.127.up_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.13.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.14.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.14.up_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.16.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.17.down_proj": { "bits": 8 }, "model.language_model.layers.47.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.19.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.2.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.20.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.21.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.22.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.26.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.27.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.28.down_proj": { "bits": 8 }, "model.language_model.layers.47.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.30.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.31.down_proj": { "bits": 8 }, "model.language_model.layers.47.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.32.up_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.36.down_proj": { "bits": 8 }, "model.language_model.layers.47.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.37.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.38.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.39.down_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.4.down_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.40.down_proj": { "bits": 8 }, "model.language_model.layers.47.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.40.up_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.41.up_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.42.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.42.up_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.44.down_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.46.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.47.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.49.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.50.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.51.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.52.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.55.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.57.down_proj": { "bits": 8 }, "model.language_model.layers.47.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.58.down_proj": { "bits": 8 }, "model.language_model.layers.47.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.59.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.6.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.60.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.61.down_proj": { "bits": 8 }, "model.language_model.layers.47.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.61.up_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.63.up_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.67.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.68.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.71.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.72.down_proj": { "bits": 8 }, "model.language_model.layers.47.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.73.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.74.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.75.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.76.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.77.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.78.down_proj": { "bits": 8 }, "model.language_model.layers.47.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.8.down_proj": { "bits": 8 }, "model.language_model.layers.47.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.82.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.82.up_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.85.down_proj": { "bits": 8 }, "model.language_model.layers.47.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.86.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.87.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.87.up_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.88.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.9.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.9.up_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.90.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.90.up_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.91.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.92.down_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.93.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.95.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.97.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.47.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.47.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.47.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.47.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.47.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.47.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.0.up_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.10.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.10.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.100.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.100.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.104.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.104.up_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.105.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.105.up_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.106.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.106.up_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.107.down_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.110.down_proj": { "bits": 8 }, "model.language_model.layers.48.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.111.down_proj": { "bits": 2 }, "model.language_model.layers.48.mlp.experts.111.gate_proj": { "bits": 2 }, "model.language_model.layers.48.mlp.experts.111.up_proj": { "bits": 2 }, "model.language_model.layers.48.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.114.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.114.up_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.115.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.116.down_proj": { "bits": 8 }, "model.language_model.layers.48.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.116.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.117.down_proj": { "bits": 8 }, "model.language_model.layers.48.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.118.down_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.119.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.120.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.121.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.122.down_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.123.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.124.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.125.up_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.126.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.13.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.14.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.16.down_proj": { "bits": 8 }, "model.language_model.layers.48.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.17.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.18.down_proj": { "bits": 8 }, "model.language_model.layers.48.mlp.experts.18.gate_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.18.up_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.19.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.2.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.20.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.21.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.22.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.23.down_proj": { "bits": 8 }, "model.language_model.layers.48.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.23.up_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.26.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.27.down_proj": { "bits": 8 }, "model.language_model.layers.48.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.29.up_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.30.down_proj": { "bits": 2 }, "model.language_model.layers.48.mlp.experts.30.gate_proj": { "bits": 2 }, "model.language_model.layers.48.mlp.experts.30.up_proj": { "bits": 2 }, "model.language_model.layers.48.mlp.experts.31.down_proj": { "bits": 8 }, "model.language_model.layers.48.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.31.up_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.32.down_proj": { "bits": 8 }, "model.language_model.layers.48.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.36.down_proj": { "bits": 8 }, "model.language_model.layers.48.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.36.up_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.37.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.38.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.40.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.40.up_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.41.up_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.42.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.44.down_proj": { "bits": 8 }, "model.language_model.layers.48.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.45.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.46.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.47.down_proj": { "bits": 8 }, "model.language_model.layers.48.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.47.up_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.49.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.5.up_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.50.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.50.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.51.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.52.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.53.gate_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.53.up_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.55.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.57.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.58.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.59.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.6.down_proj": { "bits": 8 }, "model.language_model.layers.48.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.60.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.63.up_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.64.down_proj": { "bits": 2 }, "model.language_model.layers.48.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.66.down_proj": { "bits": 8 }, "model.language_model.layers.48.mlp.experts.66.gate_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.66.up_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.67.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.68.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.71.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.72.down_proj": { "bits": 8 }, "model.language_model.layers.48.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.73.down_proj": { "bits": 8 }, "model.language_model.layers.48.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.74.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.75.up_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.76.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.77.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.78.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.79.down_proj": { "bits": 8 }, "model.language_model.layers.48.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.79.up_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.8.down_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.80.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.82.down_proj": { "bits": 8 }, "model.language_model.layers.48.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.82.up_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.85.up_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.86.down_proj": { "bits": 8 }, "model.language_model.layers.48.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.87.down_proj": { "bits": 8 }, "model.language_model.layers.48.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.87.up_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.88.down_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.9.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.9.up_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.90.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.90.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.91.down_proj": { "bits": 8 }, "model.language_model.layers.48.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.92.down_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.93.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.95.down_proj": { "bits": 8 }, "model.language_model.layers.48.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.95.up_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.97.down_proj": { "bits": 8 }, "model.language_model.layers.48.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.48.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.48.mlp.shared_experts.down_proj": { "bits": 8 }, "model.language_model.layers.48.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.48.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.48.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.48.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.48.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.10.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.10.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.100.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.100.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.101.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.101.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.104.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.104.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.105.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.106.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.106.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.107.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.107.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.110.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.111.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.111.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.114.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.114.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.115.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.116.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.117.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.118.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.118.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.119.down_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.120.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.120.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.121.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.121.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.123.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.124.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.124.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.125.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.126.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.126.gate_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.126.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.13.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.13.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.14.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.15.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.16.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.16.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.17.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.19.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.19.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.2.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.2.gate_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.2.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.20.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.20.gate_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.20.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.21.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.22.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.25.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.25.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.26.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.27.down_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.29.down_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.30.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.30.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.31.down_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.32.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.34.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.35.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.36.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.36.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.37.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.37.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.38.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.4.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.4.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.40.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.42.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.42.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.44.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.45.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.46.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.47.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.48.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.48.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.49.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.49.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.50.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.50.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.51.down_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.52.down_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.53.gate_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.53.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.55.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.55.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.57.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.58.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.59.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.59.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.6.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.60.down_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.63.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.63.gate_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.63.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.64.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.65.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.66.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.67.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.67.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.68.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.69.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.7.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.7.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.71.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.71.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.72.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.73.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.73.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.74.down_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.74.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.75.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.76.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.76.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.77.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.77.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.78.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.80.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.82.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.85.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.86.down_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.87.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.87.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.88.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.89.down_proj": { "bits": 2 }, "model.language_model.layers.49.mlp.experts.89.gate_proj": { "bits": 2 }, "model.language_model.layers.49.mlp.experts.89.up_proj": { "bits": 2 }, "model.language_model.layers.49.mlp.experts.9.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.9.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.90.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.90.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.91.down_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.92.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.93.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.94.gate_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.94.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.95.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.95.up_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.97.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.49.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.49.mlp.shared_experts.down_proj": { "bits": 8 }, "model.language_model.layers.49.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.49.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.49.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.49.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.49.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.5.mlp.experts.0.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.0.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.0.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.1.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.1.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.1.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.10.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.10.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.10.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.100.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.100.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.100.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.101.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.101.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.101.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.102.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.102.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.102.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.103.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.103.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.103.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.104.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.104.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.104.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.105.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.105.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.105.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.106.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.106.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.106.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.107.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.107.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.107.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.108.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.108.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.108.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.109.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.109.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.109.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.11.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.11.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.11.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.110.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.110.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.110.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.111.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.111.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.111.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.112.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.112.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.112.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.113.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.113.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.113.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.114.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.114.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.114.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.115.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.115.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.115.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.116.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.116.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.116.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.117.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.117.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.117.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.118.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.118.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.118.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.119.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.119.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.119.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.12.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.12.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.12.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.120.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.120.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.120.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.121.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.121.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.121.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.122.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.122.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.122.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.123.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.123.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.123.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.124.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.124.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.124.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.125.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.125.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.125.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.126.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.126.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.126.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.127.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.127.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.127.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.13.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.13.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.13.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.14.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.14.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.14.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.15.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.15.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.15.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.16.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.16.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.16.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.17.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.17.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.17.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.18.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.18.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.18.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.19.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.19.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.19.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.2.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.2.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.2.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.20.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.20.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.20.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.21.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.21.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.21.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.22.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.22.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.22.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.23.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.23.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.23.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.24.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.24.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.24.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.25.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.25.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.25.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.26.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.26.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.26.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.27.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.27.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.27.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.28.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.28.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.28.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.29.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.29.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.29.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.3.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.3.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.3.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.30.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.30.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.30.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.31.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.31.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.31.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.32.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.32.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.32.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.33.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.33.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.33.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.34.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.34.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.34.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.35.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.35.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.35.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.36.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.36.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.36.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.37.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.37.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.37.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.38.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.38.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.38.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.39.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.39.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.39.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.4.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.4.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.4.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.40.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.40.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.40.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.41.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.41.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.41.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.42.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.42.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.42.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.43.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.43.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.43.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.44.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.44.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.44.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.45.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.45.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.45.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.46.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.46.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.46.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.47.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.47.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.47.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.48.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.48.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.48.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.49.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.49.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.49.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.5.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.5.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.5.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.50.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.50.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.50.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.51.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.51.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.51.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.52.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.52.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.52.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.53.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.53.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.53.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.54.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.54.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.54.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.55.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.55.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.55.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.56.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.56.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.56.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.57.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.57.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.57.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.58.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.58.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.58.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.59.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.59.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.59.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.6.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.6.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.6.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.60.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.60.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.60.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.61.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.61.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.61.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.62.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.62.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.62.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.63.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.63.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.63.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.64.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.64.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.64.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.65.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.65.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.65.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.66.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.66.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.66.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.67.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.67.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.67.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.68.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.68.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.68.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.69.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.69.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.69.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.7.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.7.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.7.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.70.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.70.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.70.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.71.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.71.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.71.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.72.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.72.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.72.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.73.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.73.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.73.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.74.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.74.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.74.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.75.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.75.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.75.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.76.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.76.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.76.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.77.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.77.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.77.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.78.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.78.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.78.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.79.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.79.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.79.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.8.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.8.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.8.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.80.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.80.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.80.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.81.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.81.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.81.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.82.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.82.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.82.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.83.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.83.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.83.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.84.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.84.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.84.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.85.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.85.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.85.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.86.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.86.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.86.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.87.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.87.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.87.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.88.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.88.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.88.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.89.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.89.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.89.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.9.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.9.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.9.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.90.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.90.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.90.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.91.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.91.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.91.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.92.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.92.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.92.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.93.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.93.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.93.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.94.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.94.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.94.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.95.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.95.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.95.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.96.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.96.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.96.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.97.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.97.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.97.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.98.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.98.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.98.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.99.down_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.99.gate_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.experts.99.up_proj": { "bits": 2 }, "model.language_model.layers.5.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.5.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.5.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.5.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.5.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.5.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.10.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.10.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.100.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.100.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.101.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.101.gate_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.101.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.102.down_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.104.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.104.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.105.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.105.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.106.down_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.106.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.107.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.107.gate_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.107.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.108.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.109.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.109.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.11.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.110.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.111.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.112.gate_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.112.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.114.down_proj": { "bits": 2 }, "model.language_model.layers.50.mlp.experts.114.gate_proj": { "bits": 2 }, "model.language_model.layers.50.mlp.experts.114.up_proj": { "bits": 2 }, "model.language_model.layers.50.mlp.experts.115.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.115.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.116.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.117.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.117.gate_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.117.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.118.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.119.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.12.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.120.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.120.gate_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.120.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.121.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.122.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.123.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.123.gate_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.123.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.124.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.124.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.126.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.126.gate_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.126.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.127.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.127.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.13.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.14.down_proj": { "bits": 2 }, "model.language_model.layers.50.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.14.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.16.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.17.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.18.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.18.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.19.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.19.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.2.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.20.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.21.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.22.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.23.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.26.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.27.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.27.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.29.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.29.gate_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.29.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.3.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.30.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.30.gate_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.30.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.31.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.31.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.32.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.32.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.33.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.33.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.34.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.36.down_proj": { "bits": 2 }, "model.language_model.layers.50.mlp.experts.36.gate_proj": { "bits": 2 }, "model.language_model.layers.50.mlp.experts.36.up_proj": { "bits": 2 }, "model.language_model.layers.50.mlp.experts.37.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.38.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.38.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.39.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.4.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.40.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.40.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.41.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.41.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.42.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.44.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.44.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.45.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.45.gate_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.45.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.46.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.47.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.48.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.49.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.49.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.50.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.50.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.51.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.51.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.52.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.53.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.55.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.56.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.56.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.57.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.58.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.58.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.59.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.59.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.6.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.60.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.60.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.61.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.62.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.63.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.65.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.65.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.66.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.67.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.68.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.69.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.7.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.70.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.71.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.71.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.71.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.72.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.73.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.73.gate_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.73.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.74.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.74.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.75.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.76.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.76.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.77.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.77.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.78.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.80.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.82.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.82.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.83.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.84.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.86.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.87.down_proj": { "bits": 2 }, "model.language_model.layers.50.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.88.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.89.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.9.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.90.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.90.gate_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.90.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.91.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.91.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.92.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.92.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.93.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.93.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.95.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.95.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.96.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.97.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.97.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.50.mlp.experts.99.up_proj": { "bits": 4 }, "model.language_model.layers.50.mlp.shared_experts.down_proj": { "bits": 8 }, "model.language_model.layers.50.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.50.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.50.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.50.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.50.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.0.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.1.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.1.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.10.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.10.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.100.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.100.gate_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.100.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.101.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.102.down_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.102.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.103.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.103.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.104.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.104.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.105.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.105.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.106.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.106.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.107.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.107.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.108.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.108.gate_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.108.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.110.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.110.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.111.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.111.gate_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.111.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.112.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.113.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.113.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.114.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.114.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.115.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.116.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.116.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.117.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.117.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.118.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.119.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.120.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.120.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.121.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.121.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.122.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.122.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.123.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.123.gate_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.123.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.124.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.124.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.125.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.126.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.127.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.127.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.127.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.13.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.13.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.14.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.14.gate_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.14.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.15.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.15.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.16.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.16.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.17.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.17.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.19.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.19.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.2.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.2.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.20.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.20.gate_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.20.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.21.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.21.gate_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.21.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.22.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.22.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.23.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.23.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.24.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.25.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.25.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.26.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.26.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.27.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.27.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.29.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.3.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.30.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.31.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.31.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.32.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.32.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.33.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.33.gate_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.33.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.34.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.34.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.35.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.36.down_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.37.down_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.37.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.38.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.38.gate_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.38.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.39.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.39.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.4.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.4.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.40.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.40.gate_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.40.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.41.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.42.down_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.42.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.43.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.43.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.44.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.44.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.45.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.45.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.46.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.47.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.47.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.48.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.49.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.49.gate_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.49.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.5.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.50.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.50.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.51.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.51.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.52.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.52.gate_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.52.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.53.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.54.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.54.gate_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.54.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.55.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.55.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.56.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.57.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.58.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.59.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.59.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.6.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.6.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.60.down_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.61.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.61.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.62.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.64.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.65.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.66.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.66.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.67.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.67.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.68.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.68.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.69.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.7.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.70.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.70.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.71.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.71.gate_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.71.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.72.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.73.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.73.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.74.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.74.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.75.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.75.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.76.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.77.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.77.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.78.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.79.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.79.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.8.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.80.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.81.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.82.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.82.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.83.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.83.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.84.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.85.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.85.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.86.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.87.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.88.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.89.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.9.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.9.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.90.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.90.gate_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.90.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.91.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.91.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.92.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.92.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.93.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.93.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.94.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.95.down_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.96.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.96.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.97.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.97.gate_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.97.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.98.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.98.up_proj": { "bits": 3 }, "model.language_model.layers.51.mlp.experts.99.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.experts.99.gate_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.experts.99.up_proj": { "bits": 4 }, "model.language_model.layers.51.mlp.shared_experts.down_proj": { "bits": 8 }, "model.language_model.layers.51.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.51.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.51.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.51.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.51.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.0.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.0.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.1.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.1.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.10.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.10.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.100.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.100.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.101.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.101.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.102.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.102.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.102.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.103.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.104.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.104.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.104.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.105.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.105.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.105.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.106.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.106.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.106.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.107.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.107.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.108.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.109.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.109.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.11.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.110.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.110.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.111.down_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.112.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.112.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.113.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.113.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.113.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.114.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.114.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.114.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.115.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.115.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.116.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.116.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.117.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.117.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.118.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.119.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.119.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.12.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.12.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.12.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.120.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.120.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.121.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.121.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.122.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.123.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.123.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.124.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.124.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.124.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.125.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.125.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.126.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.126.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.126.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.127.down_proj": { "bits": 2 }, "model.language_model.layers.52.mlp.experts.127.gate_proj": { "bits": 2 }, "model.language_model.layers.52.mlp.experts.127.up_proj": { "bits": 2 }, "model.language_model.layers.52.mlp.experts.13.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.14.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.14.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.15.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.16.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.16.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.17.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.17.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.18.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.18.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.19.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.19.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.2.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.2.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.20.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.20.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.21.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.21.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.22.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.22.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.23.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.23.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.24.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.24.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.25.down_proj": { "bits": 2 }, "model.language_model.layers.52.mlp.experts.25.gate_proj": { "bits": 2 }, "model.language_model.layers.52.mlp.experts.25.up_proj": { "bits": 2 }, "model.language_model.layers.52.mlp.experts.26.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.26.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.27.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.27.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.27.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.28.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.29.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.3.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.3.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.3.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.30.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.30.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.30.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.31.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.31.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.32.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.32.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.33.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.33.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.34.down_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.35.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.36.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.37.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.37.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.38.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.38.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.38.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.39.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.39.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.39.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.4.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.40.down_proj": { "bits": 2 }, "model.language_model.layers.52.mlp.experts.40.gate_proj": { "bits": 2 }, "model.language_model.layers.52.mlp.experts.40.up_proj": { "bits": 2 }, "model.language_model.layers.52.mlp.experts.41.down_proj": { "bits": 2 }, "model.language_model.layers.52.mlp.experts.41.gate_proj": { "bits": 2 }, "model.language_model.layers.52.mlp.experts.41.up_proj": { "bits": 2 }, "model.language_model.layers.52.mlp.experts.42.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.42.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.43.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.43.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.44.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.44.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.45.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.45.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.46.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.46.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.47.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.47.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.48.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.48.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.49.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.49.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.5.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.5.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.50.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.50.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.50.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.51.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.51.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.52.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.52.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.53.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.53.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.54.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.54.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.54.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.55.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.55.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.56.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.57.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.58.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.58.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.59.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.59.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.6.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.6.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.60.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.60.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.61.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.61.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.61.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.62.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.63.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.63.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.65.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.65.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.65.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.66.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.66.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.67.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.67.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.68.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.68.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.69.down_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.69.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.69.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.7.down_proj": { "bits": 2 }, "model.language_model.layers.52.mlp.experts.7.gate_proj": { "bits": 2 }, "model.language_model.layers.52.mlp.experts.7.up_proj": { "bits": 2 }, "model.language_model.layers.52.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.70.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.71.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.71.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.71.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.72.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.72.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.73.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.73.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.73.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.74.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.74.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.75.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.75.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.75.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.76.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.76.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.76.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.77.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.77.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.78.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.79.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.79.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.79.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.8.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.8.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.80.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.80.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.81.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.81.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.82.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.82.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.83.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.83.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.84.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.84.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.85.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.85.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.86.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.87.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.87.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.88.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.88.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.88.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.89.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.89.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.89.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.9.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.9.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.90.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.90.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.91.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.91.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.91.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.92.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.92.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.93.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.93.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.94.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.94.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.95.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.95.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.95.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.96.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.96.gate_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.96.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.97.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.97.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.98.up_proj": { "bits": 4 }, "model.language_model.layers.52.mlp.experts.99.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.experts.99.up_proj": { "bits": 3 }, "model.language_model.layers.52.mlp.shared_experts.down_proj": { "bits": 8 }, "model.language_model.layers.52.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.52.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.52.self_attn.o_proj": { "bits": 8 }, "model.language_model.layers.52.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.52.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.0.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.0.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.1.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.1.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.1.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.10.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.10.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.10.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.100.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.100.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.101.down_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.101.up_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.102.down_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.102.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.103.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.103.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.104.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.104.up_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.105.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.105.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.105.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.106.down_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.106.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.107.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.108.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.108.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.108.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.109.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.109.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.11.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.110.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.110.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.111.down_proj": { "bits": 2 }, "model.language_model.layers.53.mlp.experts.111.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.111.up_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.112.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.112.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.112.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.113.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.113.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.113.up_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.114.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.114.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.115.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.115.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.115.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.116.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.116.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.117.down_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.117.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.118.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.118.gate_proj": { "bits": 2 }, "model.language_model.layers.53.mlp.experts.118.up_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.119.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.119.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.12.up_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.120.down_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.120.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.120.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.121.down_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.121.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.121.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.122.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.122.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.123.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.123.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.124.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.124.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.125.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.125.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.125.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.126.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.126.up_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.127.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.127.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.127.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.13.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.13.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.14.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.14.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.15.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.16.down_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.16.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.17.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.17.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.17.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.18.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.18.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.18.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.19.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.19.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.19.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.2.down_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.2.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.20.down_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.20.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.21.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.21.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.22.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.22.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.22.up_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.23.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.24.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.24.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.25.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.26.down_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.26.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.27.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.27.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.27.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.28.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.28.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.28.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.29.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.29.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.3.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.3.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.30.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.30.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.30.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.31.down_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.31.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.32.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.32.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.33.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.33.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.33.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.34.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.35.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.35.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.36.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.36.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.37.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.37.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.38.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.38.up_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.39.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.39.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.39.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.4.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.4.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.4.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.40.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.40.up_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.41.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.41.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.42.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.42.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.42.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.43.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.43.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.43.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.44.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.44.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.44.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.45.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.45.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.45.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.46.down_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.46.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.46.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.47.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.47.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.48.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.48.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.49.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.49.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.5.up_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.50.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.50.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.51.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.51.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.51.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.52.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.52.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.52.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.53.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.54.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.54.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.55.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.55.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.56.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.57.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.57.up_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.58.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.58.up_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.59.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.59.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.59.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.6.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.6.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.6.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.60.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.60.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.61.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.62.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.62.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.64.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.64.up_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.65.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.65.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.66.down_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.66.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.66.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.67.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.67.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.67.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.68.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.68.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.69.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.69.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.69.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.7.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.7.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.70.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.70.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.70.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.71.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.71.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.71.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.72.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.72.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.73.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.73.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.73.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.74.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.74.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.75.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.75.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.75.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.76.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.76.up_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.77.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.77.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.77.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.78.down_proj": { "bits": 2 }, "model.language_model.layers.53.mlp.experts.78.gate_proj": { "bits": 2 }, "model.language_model.layers.53.mlp.experts.78.up_proj": { "bits": 2 }, "model.language_model.layers.53.mlp.experts.79.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.79.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.79.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.8.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.8.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.80.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.80.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.81.down_proj": { "bits": 2 }, "model.language_model.layers.53.mlp.experts.81.gate_proj": { "bits": 2 }, "model.language_model.layers.53.mlp.experts.81.up_proj": { "bits": 2 }, "model.language_model.layers.53.mlp.experts.82.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.82.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.82.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.83.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.83.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.83.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.84.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.84.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.84.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.85.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.85.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.85.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.86.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.86.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.86.up_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.87.down_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.87.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.88.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.88.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.89.down_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.89.up_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.9.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.9.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.9.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.90.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.90.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.90.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.91.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.91.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.91.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.92.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.92.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.93.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.93.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.93.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.94.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.94.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.94.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.95.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.95.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.95.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.96.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.96.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.97.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.97.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.97.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.53.mlp.experts.98.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.99.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.experts.99.gate_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.experts.99.up_proj": { "bits": 4 }, "model.language_model.layers.53.mlp.shared_experts.down_proj": { "bits": 8 }, "model.language_model.layers.53.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.53.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.53.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.53.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.53.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.0.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.1.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.1.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.1.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.10.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.10.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.10.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.100.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.100.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.100.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.101.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.101.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.101.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.102.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.102.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.102.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.103.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.103.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.104.down_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.104.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.104.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.105.down_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.105.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.106.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.106.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.106.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.107.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.107.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.108.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.108.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.108.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.109.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.109.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.109.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.11.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.11.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.110.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.110.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.110.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.111.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.111.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.111.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.112.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.112.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.112.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.113.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.113.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.113.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.114.down_proj": { "bits": 2 }, "model.language_model.layers.54.mlp.experts.114.gate_proj": { "bits": 2 }, "model.language_model.layers.54.mlp.experts.114.up_proj": { "bits": 2 }, "model.language_model.layers.54.mlp.experts.115.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.115.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.115.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.116.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.117.down_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.117.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.118.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.118.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.118.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.119.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.119.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.12.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.12.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.120.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.120.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.120.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.121.down_proj": { "bits": 2 }, "model.language_model.layers.54.mlp.experts.121.gate_proj": { "bits": 2 }, "model.language_model.layers.54.mlp.experts.121.up_proj": { "bits": 2 }, "model.language_model.layers.54.mlp.experts.122.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.122.up_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.123.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.123.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.124.down_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.124.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.125.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.125.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.125.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.126.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.126.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.126.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.127.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.127.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.127.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.13.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.13.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.13.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.14.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.14.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.14.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.15.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.15.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.16.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.16.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.16.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.17.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.17.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.18.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.18.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.18.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.19.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.19.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.2.down_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.2.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.20.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.20.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.21.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.21.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.22.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.22.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.22.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.23.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.23.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.23.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.24.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.24.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.24.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.25.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.25.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.25.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.26.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.26.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.26.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.27.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.27.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.27.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.28.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.28.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.29.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.29.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.3.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.3.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.30.down_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.30.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.31.down_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.31.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.32.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.32.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.32.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.33.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.33.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.33.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.34.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.34.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.34.up_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.35.down_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.35.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.35.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.36.down_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.36.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.37.down_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.37.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.37.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.38.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.38.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.39.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.39.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.39.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.4.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.4.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.4.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.40.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.40.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.41.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.41.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.41.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.42.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.42.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.42.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.43.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.44.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.44.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.44.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.45.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.45.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.45.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.46.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.46.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.46.up_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.47.down_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.47.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.48.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.48.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.48.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.49.down_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.49.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.5.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.5.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.50.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.50.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.50.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.51.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.51.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.52.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.52.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.52.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.53.down_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.53.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.54.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.54.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.55.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.55.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.55.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.56.down_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.56.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.57.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.57.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.57.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.58.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.58.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.59.down_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.59.up_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.6.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.6.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.6.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.60.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.60.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.60.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.61.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.61.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.62.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.62.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.62.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.63.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.63.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.63.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.64.down_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.64.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.65.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.65.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.66.down_proj": { "bits": 2 }, "model.language_model.layers.54.mlp.experts.66.gate_proj": { "bits": 2 }, "model.language_model.layers.54.mlp.experts.66.up_proj": { "bits": 2 }, "model.language_model.layers.54.mlp.experts.67.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.67.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.68.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.68.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.69.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.69.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.69.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.7.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.7.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.7.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.70.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.70.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.71.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.71.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.71.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.72.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.72.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.73.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.73.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.73.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.74.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.74.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.75.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.75.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.76.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.76.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.76.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.77.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.77.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.78.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.78.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.79.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.79.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.79.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.8.down_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.8.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.80.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.80.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.80.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.81.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.81.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.81.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.82.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.82.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.82.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.83.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.83.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.84.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.84.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.84.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.85.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.85.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.85.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.86.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.86.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.86.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.87.down_proj": { "bits": 2 }, "model.language_model.layers.54.mlp.experts.87.gate_proj": { "bits": 2 }, "model.language_model.layers.54.mlp.experts.87.up_proj": { "bits": 2 }, "model.language_model.layers.54.mlp.experts.88.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.88.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.89.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.89.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.89.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.9.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.9.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.9.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.90.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.90.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.91.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.91.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.92.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.92.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.92.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.93.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.93.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.94.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.94.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.95.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.95.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.95.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.96.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.96.gate_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.96.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.97.down_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.97.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.98.down_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.98.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.experts.99.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.54.mlp.experts.99.up_proj": { "bits": 4 }, "model.language_model.layers.54.mlp.shared_experts.down_proj": { "bits": 8 }, "model.language_model.layers.54.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.54.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.54.self_attn.o_proj": { "bits": 8 }, "model.language_model.layers.54.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.54.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.0.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.0.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.0.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.1.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.1.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.1.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.10.down_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.10.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.10.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.100.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.100.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.100.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.101.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.101.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.101.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.102.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.102.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.103.down_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.103.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.103.up_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.104.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.104.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.104.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.105.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.105.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.105.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.106.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.106.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.106.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.107.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.107.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.107.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.108.down_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.108.up_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.109.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.109.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.11.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.11.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.11.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.110.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.110.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.111.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.111.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.111.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.112.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.112.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.112.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.113.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.113.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.113.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.114.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.114.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.114.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.115.down_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.115.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.115.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.116.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.116.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.116.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.117.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.117.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.118.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.118.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.118.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.119.down_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.119.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.12.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.12.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.12.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.120.down_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.120.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.121.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.121.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.121.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.122.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.122.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.122.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.123.down_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.123.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.124.down_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.124.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.124.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.125.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.125.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.125.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.126.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.126.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.127.down_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.127.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.127.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.13.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.13.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.14.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.14.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.14.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.15.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.15.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.15.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.16.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.16.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.16.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.17.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.17.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.18.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.18.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.19.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.19.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.19.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.2.down_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.2.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.20.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.20.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.21.down_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.21.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.21.up_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.22.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.22.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.22.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.23.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.23.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.23.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.24.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.25.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.25.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.25.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.26.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.26.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.26.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.27.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.27.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.27.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.28.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.28.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.28.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.29.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.29.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.29.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.3.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.3.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.3.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.30.down_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.30.up_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.31.down_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.31.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.31.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.32.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.32.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.32.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.33.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.33.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.33.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.34.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.34.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.34.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.35.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.35.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.35.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.36.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.36.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.36.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.37.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.37.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.37.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.38.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.38.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.38.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.39.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.39.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.4.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.4.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.40.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.40.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.40.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.41.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.41.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.41.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.42.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.42.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.42.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.43.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.43.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.43.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.44.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.44.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.44.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.45.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.45.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.45.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.46.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.46.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.46.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.47.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.47.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.48.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.48.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.48.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.49.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.49.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.49.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.5.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.5.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.50.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.50.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.50.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.51.down_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.51.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.51.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.52.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.52.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.52.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.53.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.53.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.53.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.54.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.54.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.54.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.55.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.55.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.55.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.56.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.56.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.57.down_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.57.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.57.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.58.down_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.58.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.58.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.59.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.59.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.59.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.6.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.6.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.6.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.60.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.60.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.61.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.61.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.62.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.62.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.62.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.63.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.63.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.63.up_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.64.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.64.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.64.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.65.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.65.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.65.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.66.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.66.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.66.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.67.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.67.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.68.down_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.68.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.68.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.69.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.69.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.7.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.7.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.7.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.70.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.70.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.70.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.71.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.71.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.71.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.72.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.72.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.72.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.73.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.73.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.73.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.74.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.74.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.75.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.75.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.75.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.76.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.76.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.76.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.77.down_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.77.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.78.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.78.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.78.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.79.down_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.79.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.79.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.8.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.8.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.80.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.80.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.81.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.81.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.81.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.82.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.82.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.82.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.83.down_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.83.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.83.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.84.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.84.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.84.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.85.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.85.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.85.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.86.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.86.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.86.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.87.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.87.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.87.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.88.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.88.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.89.down_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.89.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.9.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.9.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.9.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.90.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.90.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.91.down_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.91.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.91.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.92.down_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.92.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.92.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.93.down_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.93.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.94.down_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.94.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.94.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.95.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.95.gate_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.95.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.96.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.96.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.97.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.98.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.98.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.98.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.experts.99.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.55.mlp.experts.99.up_proj": { "bits": 4 }, "model.language_model.layers.55.mlp.shared_experts.down_proj": { "bits": 8 }, "model.language_model.layers.55.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.55.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.55.self_attn.o_proj": { "bits": 8 }, "model.language_model.layers.55.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.55.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.0.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.0.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.1.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.1.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.1.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.10.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.10.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.10.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.100.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.100.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.100.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.101.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.101.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.101.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.102.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.102.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.102.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.103.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.103.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.103.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.104.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.104.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.104.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.105.down_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.105.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.105.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.106.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.106.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.106.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.107.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.107.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.108.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.108.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.108.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.109.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.109.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.109.up_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.11.down_proj": { "bits": 2 }, "model.language_model.layers.56.mlp.experts.11.gate_proj": { "bits": 2 }, "model.language_model.layers.56.mlp.experts.11.up_proj": { "bits": 2 }, "model.language_model.layers.56.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.110.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.110.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.111.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.111.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.111.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.112.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.112.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.112.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.113.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.113.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.113.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.114.down_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.114.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.114.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.115.down_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.115.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.116.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.116.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.116.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.117.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.117.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.118.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.118.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.118.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.119.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.119.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.119.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.12.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.12.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.120.down_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.120.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.120.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.121.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.121.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.121.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.122.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.122.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.122.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.123.down_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.123.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.123.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.124.down_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.124.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.124.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.125.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.126.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.126.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.126.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.127.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.127.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.127.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.13.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.13.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.13.up_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.14.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.14.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.14.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.15.down_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.15.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.15.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.16.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.16.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.16.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.17.down_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.17.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.17.up_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.18.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.18.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.19.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.19.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.19.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.2.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.2.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.2.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.20.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.20.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.21.down_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.21.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.21.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.22.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.22.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.22.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.23.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.23.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.23.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.24.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.24.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.25.down_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.25.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.25.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.26.down_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.26.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.26.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.27.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.27.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.27.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.28.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.28.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.28.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.29.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.29.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.29.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.3.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.3.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.3.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.30.down_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.30.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.30.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.31.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.31.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.31.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.32.down_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.32.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.32.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.33.down_proj": { "bits": 2 }, "model.language_model.layers.56.mlp.experts.33.gate_proj": { "bits": 2 }, "model.language_model.layers.56.mlp.experts.33.up_proj": { "bits": 2 }, "model.language_model.layers.56.mlp.experts.34.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.34.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.34.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.35.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.35.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.35.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.36.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.36.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.36.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.37.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.37.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.37.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.38.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.38.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.38.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.39.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.39.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.39.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.4.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.4.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.4.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.40.down_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.40.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.40.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.41.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.41.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.41.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.42.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.42.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.42.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.43.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.43.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.43.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.44.down_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.44.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.44.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.45.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.45.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.45.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.46.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.46.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.46.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.47.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.47.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.47.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.48.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.48.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.48.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.49.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.49.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.49.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.5.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.5.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.5.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.50.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.50.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.50.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.51.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.51.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.51.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.52.down_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.52.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.52.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.53.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.53.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.53.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.54.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.54.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.54.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.55.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.55.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.55.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.56.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.56.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.56.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.57.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.57.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.57.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.58.down_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.58.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.58.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.59.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.59.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.59.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.6.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.6.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.6.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.60.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.60.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.60.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.61.down_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.61.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.61.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.62.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.62.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.62.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.63.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.63.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.63.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.64.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.64.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.64.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.65.down_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.65.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.66.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.66.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.66.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.67.down_proj": { "bits": 2 }, "model.language_model.layers.56.mlp.experts.67.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.67.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.68.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.68.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.68.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.69.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.69.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.69.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.7.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.7.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.70.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.70.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.70.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.71.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.71.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.71.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.72.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.72.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.72.up_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.73.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.73.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.73.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.74.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.74.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.75.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.75.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.75.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.76.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.76.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.76.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.77.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.77.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.78.down_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.78.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.79.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.79.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.79.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.8.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.8.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.8.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.80.down_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.80.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.81.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.81.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.82.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.82.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.82.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.83.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.83.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.83.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.84.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.84.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.85.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.85.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.85.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.86.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.86.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.86.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.87.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.87.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.87.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.88.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.88.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.89.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.89.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.89.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.9.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.9.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.9.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.90.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.90.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.90.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.91.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.91.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.91.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.92.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.92.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.92.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.93.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.93.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.93.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.94.down_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.94.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.94.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.95.down_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.95.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.95.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.96.down_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.96.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.96.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.97.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.97.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.97.up_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.98.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.98.gate_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.98.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.experts.99.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.experts.99.gate_proj": { "bits": 3 }, "model.language_model.layers.56.mlp.experts.99.up_proj": { "bits": 4 }, "model.language_model.layers.56.mlp.shared_experts.down_proj": { "bits": 8 }, "model.language_model.layers.56.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.56.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.56.self_attn.o_proj": { "bits": 8 }, "model.language_model.layers.56.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.56.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.0.down_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.0.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.0.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.1.down_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.1.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.1.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.10.down_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.10.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.10.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.100.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.100.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.100.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.101.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.101.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.101.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.102.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.102.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.102.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.103.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.103.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.103.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.104.down_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.104.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.104.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.105.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.105.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.105.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.106.down_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.106.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.106.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.107.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.107.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.107.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.108.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.108.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.108.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.109.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.109.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.109.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.11.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.11.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.110.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.110.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.111.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.111.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.111.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.112.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.112.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.112.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.113.down_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.113.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.113.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.114.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.114.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.114.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.115.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.115.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.115.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.116.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.116.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.117.down_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.117.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.117.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.118.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.118.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.118.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.119.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.119.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.119.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.12.down_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.12.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.12.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.120.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.120.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.120.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.121.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.121.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.121.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.122.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.122.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.122.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.123.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.123.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.123.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.124.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.124.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.124.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.125.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.125.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.125.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.126.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.126.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.126.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.127.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.127.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.127.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.13.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.13.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.13.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.14.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.14.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.14.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.15.down_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.15.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.15.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.16.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.16.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.16.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.17.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.17.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.17.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.18.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.18.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.18.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.19.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.19.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.19.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.2.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.2.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.2.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.20.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.20.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.20.up_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.21.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.21.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.21.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.22.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.22.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.22.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.23.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.23.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.23.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.24.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.24.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.24.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.25.down_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.25.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.25.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.26.down_proj": { "bits": 2 }, "model.language_model.layers.57.mlp.experts.26.gate_proj": { "bits": 2 }, "model.language_model.layers.57.mlp.experts.26.up_proj": { "bits": 2 }, "model.language_model.layers.57.mlp.experts.27.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.27.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.27.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.28.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.28.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.28.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.29.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.29.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.29.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.3.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.3.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.3.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.30.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.30.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.30.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.31.down_proj": { "bits": 2 }, "model.language_model.layers.57.mlp.experts.31.gate_proj": { "bits": 2 }, "model.language_model.layers.57.mlp.experts.31.up_proj": { "bits": 2 }, "model.language_model.layers.57.mlp.experts.32.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.32.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.32.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.33.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.33.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.33.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.34.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.34.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.34.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.35.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.35.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.35.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.36.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.36.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.36.up_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.37.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.37.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.37.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.38.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.38.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.38.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.39.down_proj": { "bits": 2 }, "model.language_model.layers.57.mlp.experts.39.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.39.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.4.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.4.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.40.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.40.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.40.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.41.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.41.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.41.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.42.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.42.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.42.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.43.down_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.43.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.43.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.44.down_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.44.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.44.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.45.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.45.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.45.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.46.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.46.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.46.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.47.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.47.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.47.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.48.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.48.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.49.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.49.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.49.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.5.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.5.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.5.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.50.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.50.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.50.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.51.down_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.51.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.51.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.52.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.52.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.52.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.53.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.53.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.53.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.54.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.54.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.54.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.55.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.55.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.55.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.56.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.56.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.57.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.57.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.57.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.58.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.58.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.58.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.59.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.59.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.59.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.6.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.6.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.6.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.60.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.60.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.60.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.61.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.61.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.61.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.62.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.62.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.62.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.63.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.63.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.63.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.64.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.64.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.64.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.65.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.65.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.65.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.66.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.66.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.66.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.67.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.67.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.67.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.68.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.68.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.68.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.69.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.69.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.69.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.7.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.7.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.7.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.70.down_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.70.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.70.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.71.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.71.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.71.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.72.down_proj": { "bits": 2 }, "model.language_model.layers.57.mlp.experts.72.gate_proj": { "bits": 2 }, "model.language_model.layers.57.mlp.experts.72.up_proj": { "bits": 2 }, "model.language_model.layers.57.mlp.experts.73.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.73.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.73.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.74.down_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.74.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.74.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.75.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.75.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.75.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.76.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.76.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.76.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.77.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.77.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.77.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.78.down_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.78.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.78.up_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.79.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.79.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.79.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.8.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.8.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.8.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.80.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.80.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.80.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.81.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.81.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.81.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.82.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.82.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.82.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.83.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.83.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.83.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.84.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.84.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.84.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.85.down_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.85.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.85.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.86.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.86.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.86.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.87.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.87.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.87.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.88.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.88.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.88.up_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.89.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.89.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.89.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.9.down_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.9.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.9.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.90.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.90.gate_proj": { "bits": 3 }, "model.language_model.layers.57.mlp.experts.90.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.91.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.91.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.91.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.92.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.92.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.92.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.93.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.93.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.93.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.94.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.94.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.94.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.95.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.95.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.95.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.96.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.96.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.96.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.97.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.97.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.97.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.98.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.98.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.98.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.99.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.experts.99.gate_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.experts.99.up_proj": { "bits": 4 }, "model.language_model.layers.57.mlp.shared_experts.down_proj": { "bits": 8 }, "model.language_model.layers.57.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.57.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.57.self_attn.o_proj": { "bits": 8 }, "model.language_model.layers.57.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.57.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.0.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.0.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.1.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.1.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.1.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.10.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.10.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.10.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.100.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.100.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.100.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.101.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.101.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.101.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.102.down_proj": { "bits": 3 }, "model.language_model.layers.58.mlp.experts.102.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.102.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.103.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.103.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.103.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.104.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.104.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.104.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.105.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.105.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.105.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.106.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.106.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.106.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.107.down_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.107.gate_proj": { "bits": 3 }, "model.language_model.layers.58.mlp.experts.107.up_proj": { "bits": 3 }, "model.language_model.layers.58.mlp.experts.108.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.108.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.108.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.109.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.109.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.109.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.11.down_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.11.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.11.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.110.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.110.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.110.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.111.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.111.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.111.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.112.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.112.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.112.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.113.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.113.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.113.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.114.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.114.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.114.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.115.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.115.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.115.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.116.down_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.116.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.116.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.117.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.117.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.117.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.118.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.118.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.118.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.119.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.119.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.119.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.12.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.12.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.12.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.120.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.120.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.120.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.121.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.121.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.121.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.122.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.122.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.122.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.123.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.123.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.123.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.124.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.124.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.124.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.125.down_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.125.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.125.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.126.down_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.126.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.126.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.127.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.127.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.127.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.13.down_proj": { "bits": 2 }, "model.language_model.layers.58.mlp.experts.13.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.13.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.14.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.14.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.14.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.15.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.15.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.15.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.16.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.16.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.16.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.17.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.17.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.17.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.18.down_proj": { "bits": 2 }, "model.language_model.layers.58.mlp.experts.18.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.18.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.19.down_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.19.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.19.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.2.down_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.2.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.2.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.20.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.20.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.20.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.21.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.21.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.21.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.22.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.22.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.22.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.23.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.23.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.23.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.24.down_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.24.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.24.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.25.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.25.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.25.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.26.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.26.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.26.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.27.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.27.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.27.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.28.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.28.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.28.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.29.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.29.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.29.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.3.down_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.3.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.3.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.30.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.30.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.30.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.31.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.31.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.31.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.32.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.32.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.32.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.33.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.33.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.33.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.34.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.34.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.34.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.35.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.35.gate_proj": { "bits": 3 }, "model.language_model.layers.58.mlp.experts.35.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.36.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.36.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.36.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.37.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.37.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.37.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.38.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.38.gate_proj": { "bits": 3 }, "model.language_model.layers.58.mlp.experts.38.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.39.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.39.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.39.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.4.down_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.4.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.4.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.40.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.40.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.40.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.41.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.41.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.42.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.42.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.42.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.43.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.43.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.43.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.44.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.44.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.44.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.45.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.45.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.45.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.46.down_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.46.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.46.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.47.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.47.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.47.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.48.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.48.gate_proj": { "bits": 3 }, "model.language_model.layers.58.mlp.experts.48.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.49.down_proj": { "bits": 3 }, "model.language_model.layers.58.mlp.experts.49.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.49.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.5.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.5.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.5.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.50.down_proj": { "bits": 3 }, "model.language_model.layers.58.mlp.experts.50.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.50.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.51.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.51.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.51.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.52.down_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.52.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.52.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.53.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.53.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.53.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.54.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.54.gate_proj": { "bits": 3 }, "model.language_model.layers.58.mlp.experts.54.up_proj": { "bits": 3 }, "model.language_model.layers.58.mlp.experts.55.down_proj": { "bits": 2 }, "model.language_model.layers.58.mlp.experts.55.gate_proj": { "bits": 2 }, "model.language_model.layers.58.mlp.experts.55.up_proj": { "bits": 2 }, "model.language_model.layers.58.mlp.experts.56.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.56.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.56.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.57.down_proj": { "bits": 3 }, "model.language_model.layers.58.mlp.experts.57.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.57.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.58.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.58.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.58.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.59.down_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.59.gate_proj": { "bits": 3 }, "model.language_model.layers.58.mlp.experts.59.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.6.down_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.6.gate_proj": { "bits": 3 }, "model.language_model.layers.58.mlp.experts.6.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.60.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.60.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.60.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.61.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.61.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.61.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.62.down_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.62.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.62.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.63.down_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.63.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.63.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.64.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.64.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.64.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.65.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.65.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.65.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.66.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.66.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.66.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.67.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.67.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.67.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.68.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.68.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.68.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.69.down_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.69.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.69.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.7.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.7.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.7.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.70.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.70.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.70.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.71.down_proj": { "bits": 2 }, "model.language_model.layers.58.mlp.experts.71.gate_proj": { "bits": 2 }, "model.language_model.layers.58.mlp.experts.71.up_proj": { "bits": 2 }, "model.language_model.layers.58.mlp.experts.72.down_proj": { "bits": 3 }, "model.language_model.layers.58.mlp.experts.72.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.72.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.73.down_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.73.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.73.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.74.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.74.gate_proj": { "bits": 3 }, "model.language_model.layers.58.mlp.experts.74.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.75.down_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.75.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.75.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.76.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.76.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.76.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.77.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.77.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.77.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.78.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.78.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.78.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.79.down_proj": { "bits": 3 }, "model.language_model.layers.58.mlp.experts.79.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.79.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.8.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.8.gate_proj": { "bits": 3 }, "model.language_model.layers.58.mlp.experts.8.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.80.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.80.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.80.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.81.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.81.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.81.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.82.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.82.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.82.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.83.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.83.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.83.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.84.down_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.84.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.84.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.85.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.85.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.85.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.86.down_proj": { "bits": 2 }, "model.language_model.layers.58.mlp.experts.86.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.86.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.87.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.87.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.87.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.88.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.88.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.88.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.89.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.89.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.89.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.9.down_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.9.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.9.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.90.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.90.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.90.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.91.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.91.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.91.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.92.down_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.92.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.92.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.93.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.93.gate_proj": { "bits": 3 }, "model.language_model.layers.58.mlp.experts.93.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.94.down_proj": { "bits": 2 }, "model.language_model.layers.58.mlp.experts.94.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.94.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.95.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.95.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.95.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.96.down_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.96.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.96.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.97.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.97.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.97.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.98.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.experts.98.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.98.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.99.gate_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.experts.99.up_proj": { "bits": 4 }, "model.language_model.layers.58.mlp.shared_experts.down_proj": { "bits": 8 }, "model.language_model.layers.58.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.58.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.58.self_attn.o_proj": { "bits": 8 }, "model.language_model.layers.58.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.58.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.0.down_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.0.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.0.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.1.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.1.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.1.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.10.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.10.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.10.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.100.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.100.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.100.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.101.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.101.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.101.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.102.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.102.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.102.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.103.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.103.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.103.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.104.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.104.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.104.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.105.down_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.105.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.105.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.106.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.106.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.106.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.107.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.107.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.107.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.108.down_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.108.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.108.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.109.down_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.109.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.109.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.11.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.11.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.11.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.110.down_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.110.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.110.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.111.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.111.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.111.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.112.down_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.112.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.112.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.113.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.113.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.113.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.114.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.114.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.114.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.115.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.115.gate_proj": { "bits": 3 }, "model.language_model.layers.59.mlp.experts.115.up_proj": { "bits": 3 }, "model.language_model.layers.59.mlp.experts.116.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.116.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.116.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.117.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.117.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.117.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.118.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.118.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.118.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.119.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.119.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.119.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.12.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.12.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.12.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.120.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.120.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.120.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.121.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.121.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.121.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.122.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.122.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.122.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.123.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.123.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.123.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.124.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.124.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.124.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.125.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.125.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.125.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.126.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.126.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.126.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.127.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.127.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.127.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.13.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.13.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.13.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.14.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.14.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.14.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.15.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.15.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.15.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.16.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.16.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.16.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.17.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.17.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.17.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.18.down_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.18.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.18.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.19.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.19.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.19.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.2.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.2.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.2.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.20.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.20.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.20.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.21.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.21.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.21.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.22.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.22.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.22.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.23.down_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.23.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.23.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.24.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.24.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.24.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.25.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.25.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.25.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.26.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.26.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.26.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.27.down_proj": { "bits": 3 }, "model.language_model.layers.59.mlp.experts.27.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.27.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.28.down_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.28.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.28.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.29.down_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.29.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.29.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.3.down_proj": { "bits": 3 }, "model.language_model.layers.59.mlp.experts.3.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.3.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.30.down_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.30.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.30.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.31.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.31.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.31.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.32.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.32.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.32.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.33.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.33.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.33.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.34.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.34.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.34.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.35.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.35.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.35.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.36.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.36.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.36.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.37.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.37.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.37.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.38.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.38.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.38.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.39.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.39.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.39.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.4.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.4.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.4.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.40.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.40.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.40.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.41.down_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.41.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.41.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.42.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.42.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.42.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.43.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.43.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.43.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.44.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.44.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.44.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.45.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.45.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.45.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.46.down_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.46.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.46.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.47.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.47.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.47.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.48.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.48.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.48.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.49.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.49.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.49.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.5.down_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.5.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.5.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.50.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.50.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.50.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.51.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.51.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.51.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.52.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.52.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.52.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.53.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.53.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.53.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.54.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.54.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.54.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.55.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.55.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.55.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.56.down_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.56.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.56.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.57.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.57.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.57.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.58.down_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.58.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.58.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.59.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.59.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.59.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.6.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.6.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.6.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.60.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.60.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.60.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.61.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.61.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.61.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.62.down_proj": { "bits": 3 }, "model.language_model.layers.59.mlp.experts.62.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.62.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.63.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.63.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.63.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.64.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.64.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.64.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.65.down_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.65.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.65.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.66.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.66.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.66.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.67.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.67.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.67.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.68.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.68.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.68.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.69.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.69.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.69.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.7.down_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.7.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.7.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.70.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.70.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.70.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.71.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.71.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.71.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.72.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.72.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.72.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.73.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.73.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.73.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.74.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.74.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.74.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.75.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.75.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.75.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.76.down_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.76.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.76.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.77.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.77.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.77.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.78.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.78.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.78.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.79.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.79.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.79.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.8.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.8.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.8.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.80.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.80.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.80.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.81.down_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.81.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.81.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.82.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.82.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.82.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.83.down_proj": { "bits": 2 }, "model.language_model.layers.59.mlp.experts.83.gate_proj": { "bits": 2 }, "model.language_model.layers.59.mlp.experts.83.up_proj": { "bits": 2 }, "model.language_model.layers.59.mlp.experts.84.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.84.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.84.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.85.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.85.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.85.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.86.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.86.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.86.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.87.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.87.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.87.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.88.down_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.88.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.88.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.89.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.89.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.89.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.9.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.9.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.9.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.90.down_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.90.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.90.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.91.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.91.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.91.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.92.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.92.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.92.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.93.down_proj": { "bits": 3 }, "model.language_model.layers.59.mlp.experts.93.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.93.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.94.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.94.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.94.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.95.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.95.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.95.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.96.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.96.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.96.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.97.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.97.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.97.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.98.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.experts.98.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.98.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.99.down_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.99.gate_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.experts.99.up_proj": { "bits": 4 }, "model.language_model.layers.59.mlp.shared_experts.down_proj": { "bits": 8 }, "model.language_model.layers.59.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.59.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.59.self_attn.o_proj": { "bits": 8 }, "model.language_model.layers.59.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.59.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.6.mlp.experts.0.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.0.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.0.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.1.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.1.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.1.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.10.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.10.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.10.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.100.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.100.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.100.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.101.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.101.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.101.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.102.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.102.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.102.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.103.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.103.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.103.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.104.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.104.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.104.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.105.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.105.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.105.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.106.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.106.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.106.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.107.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.107.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.107.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.108.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.108.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.108.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.109.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.109.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.109.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.11.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.11.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.11.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.110.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.110.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.110.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.111.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.111.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.111.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.112.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.112.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.112.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.113.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.113.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.113.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.114.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.114.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.114.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.115.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.115.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.115.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.116.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.116.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.116.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.117.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.117.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.117.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.118.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.118.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.118.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.119.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.119.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.119.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.12.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.12.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.12.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.120.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.120.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.120.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.121.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.121.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.121.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.122.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.122.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.122.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.123.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.123.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.123.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.124.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.124.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.124.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.125.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.125.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.125.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.126.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.126.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.126.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.127.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.127.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.127.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.13.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.13.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.13.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.14.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.14.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.14.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.15.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.15.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.15.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.16.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.16.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.16.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.17.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.17.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.17.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.18.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.18.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.18.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.19.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.19.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.19.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.2.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.2.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.2.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.20.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.20.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.20.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.21.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.21.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.21.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.22.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.22.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.22.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.23.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.23.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.23.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.24.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.24.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.24.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.25.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.25.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.25.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.26.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.26.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.26.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.27.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.27.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.27.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.28.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.28.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.28.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.29.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.29.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.29.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.3.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.3.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.3.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.30.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.30.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.30.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.31.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.31.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.31.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.32.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.32.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.32.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.33.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.33.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.33.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.34.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.34.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.34.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.35.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.35.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.35.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.36.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.36.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.36.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.37.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.37.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.37.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.38.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.38.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.38.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.39.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.39.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.39.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.4.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.4.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.4.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.40.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.40.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.40.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.41.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.41.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.41.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.42.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.42.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.42.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.43.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.43.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.43.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.44.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.44.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.44.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.45.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.45.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.45.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.46.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.46.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.46.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.47.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.47.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.47.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.48.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.48.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.48.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.49.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.49.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.49.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.5.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.5.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.5.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.50.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.50.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.50.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.51.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.51.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.51.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.52.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.52.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.52.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.53.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.53.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.53.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.54.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.54.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.54.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.55.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.55.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.55.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.56.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.56.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.56.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.57.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.57.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.57.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.58.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.58.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.58.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.59.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.59.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.59.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.6.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.6.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.6.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.60.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.60.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.60.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.61.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.61.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.61.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.62.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.62.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.62.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.63.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.63.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.63.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.64.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.64.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.64.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.65.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.65.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.65.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.66.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.66.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.66.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.67.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.67.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.67.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.68.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.68.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.68.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.69.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.69.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.69.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.7.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.7.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.7.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.70.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.70.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.70.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.71.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.71.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.71.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.72.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.72.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.72.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.73.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.73.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.73.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.74.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.74.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.74.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.75.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.75.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.75.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.76.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.76.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.76.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.77.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.77.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.77.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.78.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.78.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.78.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.79.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.79.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.79.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.8.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.8.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.8.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.80.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.80.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.80.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.81.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.81.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.81.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.82.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.82.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.82.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.83.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.83.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.83.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.84.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.84.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.84.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.85.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.85.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.85.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.86.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.86.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.86.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.87.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.87.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.87.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.88.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.88.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.88.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.89.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.89.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.89.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.9.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.9.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.9.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.90.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.90.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.90.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.91.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.91.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.91.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.92.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.92.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.92.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.93.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.93.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.93.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.94.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.94.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.94.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.95.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.95.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.95.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.96.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.96.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.96.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.97.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.97.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.97.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.98.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.98.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.98.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.99.down_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.99.gate_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.experts.99.up_proj": { "bits": 2 }, "model.language_model.layers.6.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.6.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.6.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.6.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.6.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.6.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.7.mlp.experts.0.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.0.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.0.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.1.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.1.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.1.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.10.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.10.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.10.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.100.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.100.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.100.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.101.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.101.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.101.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.102.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.102.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.102.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.103.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.103.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.103.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.104.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.104.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.104.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.105.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.105.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.105.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.106.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.106.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.106.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.107.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.107.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.107.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.108.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.108.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.108.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.109.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.109.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.109.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.11.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.11.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.11.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.110.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.110.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.110.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.111.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.111.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.111.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.112.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.112.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.112.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.113.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.113.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.113.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.114.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.114.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.114.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.115.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.115.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.115.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.116.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.116.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.116.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.117.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.117.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.117.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.118.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.118.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.118.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.119.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.119.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.119.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.12.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.12.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.12.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.120.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.120.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.120.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.121.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.121.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.121.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.122.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.122.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.122.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.123.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.123.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.123.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.124.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.124.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.124.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.125.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.125.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.125.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.126.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.126.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.126.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.127.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.127.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.127.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.13.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.13.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.13.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.14.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.14.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.14.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.15.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.15.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.15.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.16.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.16.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.16.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.17.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.17.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.17.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.18.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.18.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.18.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.19.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.19.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.19.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.2.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.2.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.2.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.20.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.20.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.20.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.21.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.21.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.21.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.22.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.22.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.22.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.23.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.23.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.23.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.24.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.24.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.24.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.25.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.25.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.25.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.26.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.26.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.26.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.27.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.27.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.27.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.28.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.28.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.28.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.29.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.29.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.29.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.3.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.3.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.3.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.30.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.30.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.30.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.31.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.31.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.31.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.32.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.32.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.32.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.33.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.33.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.33.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.34.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.34.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.34.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.35.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.35.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.35.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.36.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.36.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.36.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.37.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.37.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.37.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.38.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.38.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.38.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.39.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.39.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.39.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.4.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.4.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.4.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.40.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.40.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.40.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.41.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.41.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.41.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.42.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.42.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.42.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.43.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.43.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.43.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.44.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.44.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.44.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.45.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.45.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.45.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.46.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.46.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.46.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.47.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.47.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.47.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.48.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.48.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.48.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.49.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.49.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.49.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.5.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.5.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.5.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.50.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.50.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.50.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.51.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.51.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.51.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.52.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.52.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.52.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.53.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.53.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.53.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.54.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.54.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.54.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.55.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.55.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.55.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.56.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.56.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.56.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.57.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.57.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.57.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.58.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.58.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.58.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.59.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.59.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.59.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.6.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.6.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.6.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.60.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.60.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.60.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.61.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.61.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.61.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.62.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.62.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.62.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.63.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.63.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.63.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.64.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.64.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.64.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.65.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.65.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.65.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.66.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.66.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.66.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.67.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.67.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.67.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.68.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.68.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.68.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.69.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.69.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.69.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.7.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.7.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.7.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.70.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.70.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.70.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.71.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.71.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.71.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.72.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.72.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.72.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.73.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.73.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.73.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.74.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.74.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.74.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.75.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.75.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.75.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.76.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.76.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.76.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.77.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.77.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.77.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.78.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.78.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.78.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.79.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.79.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.79.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.8.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.8.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.8.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.80.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.80.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.80.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.81.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.81.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.81.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.82.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.82.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.82.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.83.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.83.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.83.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.84.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.84.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.84.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.85.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.85.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.85.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.86.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.86.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.86.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.87.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.87.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.87.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.88.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.88.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.88.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.89.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.89.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.89.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.9.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.9.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.9.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.90.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.90.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.90.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.91.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.91.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.91.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.92.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.92.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.92.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.93.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.93.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.93.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.94.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.94.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.94.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.95.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.95.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.95.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.96.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.96.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.96.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.97.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.97.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.97.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.98.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.98.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.98.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.99.down_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.99.gate_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.experts.99.up_proj": { "bits": 2 }, "model.language_model.layers.7.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.7.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.7.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.7.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.7.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.7.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.8.mlp.experts.0.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.0.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.0.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.1.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.1.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.1.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.10.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.10.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.10.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.100.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.100.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.100.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.101.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.101.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.101.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.102.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.102.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.102.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.103.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.103.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.103.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.104.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.104.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.104.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.105.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.105.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.105.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.106.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.106.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.106.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.107.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.107.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.107.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.108.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.108.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.108.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.109.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.109.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.109.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.11.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.11.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.11.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.110.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.110.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.110.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.111.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.111.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.111.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.112.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.112.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.112.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.113.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.113.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.113.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.114.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.114.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.114.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.115.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.115.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.115.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.116.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.116.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.116.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.117.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.117.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.117.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.118.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.118.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.118.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.119.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.119.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.119.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.12.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.12.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.12.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.120.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.120.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.120.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.121.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.121.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.121.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.122.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.122.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.122.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.123.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.123.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.123.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.124.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.124.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.124.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.125.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.125.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.125.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.126.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.126.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.126.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.127.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.127.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.127.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.13.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.13.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.13.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.14.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.14.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.14.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.15.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.15.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.15.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.16.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.16.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.16.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.17.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.17.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.17.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.18.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.18.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.18.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.19.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.19.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.19.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.2.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.2.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.2.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.20.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.20.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.20.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.21.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.21.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.21.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.22.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.22.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.22.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.23.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.23.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.23.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.24.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.24.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.24.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.25.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.25.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.25.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.26.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.26.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.26.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.27.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.27.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.27.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.28.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.28.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.28.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.29.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.29.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.29.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.3.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.3.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.3.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.30.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.30.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.30.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.31.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.31.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.31.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.32.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.32.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.32.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.33.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.33.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.33.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.34.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.34.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.34.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.35.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.35.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.35.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.36.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.36.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.36.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.37.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.37.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.37.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.38.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.38.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.38.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.39.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.39.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.39.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.4.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.4.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.4.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.40.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.40.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.40.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.41.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.41.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.41.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.42.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.42.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.42.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.43.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.43.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.43.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.44.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.44.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.44.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.45.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.45.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.45.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.46.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.46.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.46.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.47.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.47.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.47.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.48.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.48.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.48.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.49.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.49.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.49.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.5.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.5.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.5.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.50.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.50.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.50.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.51.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.51.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.51.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.52.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.52.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.52.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.53.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.53.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.53.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.54.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.54.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.54.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.55.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.55.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.55.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.56.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.56.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.56.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.57.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.57.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.57.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.58.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.58.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.58.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.59.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.59.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.59.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.6.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.6.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.6.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.60.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.60.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.60.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.61.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.61.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.61.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.62.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.62.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.62.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.63.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.63.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.63.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.64.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.64.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.64.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.65.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.65.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.65.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.66.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.66.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.66.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.67.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.67.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.67.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.68.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.68.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.68.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.69.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.69.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.69.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.7.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.7.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.7.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.70.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.70.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.70.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.71.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.71.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.71.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.72.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.72.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.72.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.73.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.73.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.73.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.74.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.74.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.74.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.75.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.75.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.75.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.76.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.76.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.76.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.77.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.77.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.77.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.78.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.78.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.78.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.79.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.79.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.79.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.8.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.8.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.8.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.80.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.80.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.80.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.81.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.81.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.81.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.82.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.82.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.82.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.83.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.83.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.83.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.84.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.84.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.84.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.85.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.85.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.85.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.86.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.86.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.86.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.87.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.87.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.87.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.88.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.88.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.88.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.89.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.89.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.89.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.9.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.9.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.9.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.90.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.90.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.90.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.91.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.91.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.91.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.92.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.92.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.92.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.93.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.93.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.93.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.94.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.94.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.94.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.95.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.95.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.95.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.96.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.96.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.96.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.97.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.97.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.97.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.98.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.98.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.98.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.99.down_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.99.gate_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.experts.99.up_proj": { "bits": 2 }, "model.language_model.layers.8.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.8.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.8.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.8.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.8.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.8.self_attn.v_proj": { "bits": 4 }, "model.language_model.layers.9.mlp.experts.0.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.0.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.0.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.1.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.1.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.1.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.10.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.10.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.10.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.100.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.100.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.100.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.101.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.101.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.101.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.102.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.102.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.102.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.103.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.103.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.103.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.104.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.104.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.104.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.105.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.105.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.105.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.106.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.106.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.106.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.107.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.107.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.107.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.108.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.108.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.108.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.109.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.109.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.109.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.11.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.11.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.11.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.110.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.110.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.110.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.111.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.111.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.111.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.112.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.112.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.112.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.113.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.113.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.113.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.114.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.114.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.114.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.115.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.115.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.115.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.116.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.116.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.116.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.117.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.117.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.117.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.118.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.118.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.118.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.119.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.119.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.119.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.12.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.12.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.12.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.120.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.120.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.120.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.121.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.121.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.121.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.122.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.122.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.122.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.123.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.123.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.123.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.124.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.124.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.124.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.125.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.125.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.125.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.126.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.126.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.126.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.127.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.127.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.127.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.13.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.13.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.13.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.14.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.14.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.14.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.15.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.15.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.15.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.16.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.16.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.16.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.17.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.17.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.17.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.18.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.18.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.18.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.19.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.19.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.19.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.2.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.2.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.2.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.20.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.20.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.20.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.21.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.21.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.21.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.22.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.22.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.22.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.23.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.23.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.23.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.24.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.24.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.24.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.25.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.25.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.25.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.26.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.26.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.26.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.27.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.27.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.27.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.28.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.28.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.28.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.29.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.29.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.29.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.3.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.3.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.3.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.30.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.30.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.30.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.31.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.31.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.31.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.32.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.32.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.32.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.33.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.33.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.33.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.34.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.34.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.34.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.35.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.35.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.35.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.36.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.36.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.36.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.37.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.37.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.37.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.38.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.38.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.38.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.39.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.39.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.39.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.4.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.4.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.4.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.40.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.40.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.40.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.41.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.41.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.41.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.42.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.42.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.42.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.43.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.43.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.43.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.44.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.44.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.44.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.45.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.45.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.45.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.46.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.46.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.46.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.47.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.47.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.47.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.48.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.48.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.48.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.49.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.49.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.49.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.5.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.5.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.5.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.50.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.50.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.50.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.51.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.51.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.51.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.52.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.52.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.52.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.53.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.53.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.53.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.54.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.54.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.54.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.55.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.55.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.55.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.56.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.56.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.56.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.57.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.57.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.57.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.58.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.58.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.58.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.59.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.59.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.59.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.6.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.6.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.6.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.60.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.60.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.60.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.61.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.61.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.61.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.62.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.62.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.62.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.63.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.63.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.63.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.64.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.64.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.64.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.65.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.65.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.65.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.66.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.66.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.66.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.67.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.67.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.67.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.68.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.68.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.68.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.69.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.69.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.69.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.7.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.7.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.7.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.70.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.70.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.70.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.71.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.71.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.71.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.72.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.72.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.72.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.73.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.73.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.73.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.74.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.74.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.74.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.75.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.75.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.75.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.76.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.76.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.76.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.77.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.77.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.77.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.78.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.78.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.78.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.79.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.79.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.79.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.8.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.8.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.8.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.80.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.80.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.80.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.81.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.81.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.81.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.82.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.82.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.82.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.83.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.83.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.83.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.84.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.84.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.84.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.85.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.85.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.85.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.86.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.86.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.86.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.87.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.87.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.87.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.88.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.88.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.88.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.89.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.89.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.89.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.9.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.9.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.9.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.90.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.90.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.90.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.91.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.91.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.91.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.92.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.92.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.92.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.93.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.93.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.93.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.94.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.94.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.94.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.95.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.95.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.95.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.96.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.96.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.96.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.97.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.97.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.97.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.98.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.98.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.98.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.99.down_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.99.gate_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.experts.99.up_proj": { "bits": 2 }, "model.language_model.layers.9.mlp.shared_experts.down_proj": { "bits": 4 }, "model.language_model.layers.9.mlp.shared_experts.gate_up_proj": { "bits": 4 }, "model.language_model.layers.9.self_attn.k_proj": { "bits": 4 }, "model.language_model.layers.9.self_attn.o_proj": { "bits": 4 }, "model.language_model.layers.9.self_attn.q_proj": { "bits": 4 }, "model.language_model.layers.9.self_attn.v_proj": { "bits": 4 } }, "group_size": 128, "iters": 1000, "low_gpu_mem_usage": true, "packing_format": "auto_round:auto_gptq", "quant_method": "auto-round", "seqlen": 512, "sym": true }, "text_config": { "architectures": [ "MiniMaxM3SparseForCausalLM" ], "attention_dropout": 0.0, "attention_output_gate": false, "bos_token_id": 200034, "dense_intermediate_size": 12288, "dtype": "bfloat16", "eos_token_id": 200020, "head_dim": 128, "hidden_act": "silu", "hidden_size": 6144, "index_block_size": 128, "index_head_dim": 128, "index_local_blocks": 1, "index_n_heads": 4, "index_topk_blocks": 16, "initializer_range": 0.02, "intermediate_size": 3072, "layer_types": [ "full_attention", "full_attention", "full_attention", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse", "minimax_m3_sparse" ], "max_position_embeddings": 1048576, "mlp_layer_types": [ "dense", "dense", "dense", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse", "sparse" ], "model_type": "minimax_m3_vl_text", "n_shared_experts": 1, "num_attention_heads": 64, "num_experts_per_tok": 4, "num_hidden_layers": 60, "num_key_value_heads": 4, "num_local_experts": 128, "num_mtp_modules": 0, "num_nextn_predict_layers": 0, "output_router_logits": false, "pad_token_id": null, "partial_rotary_factor": 0.5, "qk_norm_type": "per_head", "rms_norm_eps": 1e-06, "rope_parameters": { "partial_rotary_factor": 0.5, "rope_theta": 5000000, "rope_type": "default" }, "rotary_dim": 64, "routed_scaling_factor": 2.0, "router_aux_loss_coef": 0.001, "router_jitter_noise": 0.0, "scoring_func": "sigmoid", "shared_intermediate_size": 3072, "swiglu_alpha": 1.702, "swiglu_limit": 7.0, "tie_word_embeddings": false, "use_cache": true, "use_gemma_norm": true, "use_qk_norm": true, "use_routing_bias": true, "vocab_size": 200064 }, "tie_word_embeddings": false, "transformers_version": "5.12.0", "video_token_index": 200026, "vision_config": { "attention_dropout": 0.0, "dtype": "bfloat16", "hidden_act": "gelu", "hidden_size": 1280, "image_size": 2016, "img_token_compression_config": { "image_token_compression_method": "patch_merge", "spatial_merge_size": 2, "temporal_patch_size": 2 }, "initializer_factor": 1.0, "initializer_range": 0.02, "intermediate_size": 5120, "layer_norm_eps": 1e-05, "model_type": "minimax_m3_vl_vision", "num_attention_heads": 16, "num_channels": 3, "num_hidden_layers": 32, "patch_size": 14, "position_embedding_type": "rope", "projection_dim": 6144, "rope_mode": "3d", "rope_parameters": { "rope_theta": 10000.0, "rope_type": "default" }, "spatial_merge_size": 2, "temporal_patch_size": 2, "vision_segment_max_frames": 4, "vocab_size": 32000 }, "vision_feature_layer": -1, "vision_feature_select_strategy": "full" }