{ "add_skip_keys": true, "dequantize_fp32": false, "dynamic_loss_threshold": null, "group_size": 0, "hadamard_group_size": 256, "is_integer": true, "is_training": false, "minimum_allowed_numel": 16384, "modules_dtype_dict": {}, "modules_quant_config": {}, "modules_to_not_convert": [ "time_embedder", ".proj_out", "time_modulation", ".context_embedder", ".t_embedder", ".vid_out", "norm_post_attn", "time_text_embed", "norm_out_modulation", ".txt_in", "patch_emb", "patch_embedding", ".condition_embedder", "wte", ".time_embed", "scale_shift_table", ".txt_out", "norm1", ".vid_in", ".final_layer", ".img_in", "norm", "patch_embed", ".x_embedder", "norm2", ".emb_in", "norm_k", "multi_modal_projector", ".img_out", ".emb_out", "norm_out", "router", "lm_head", ".norm_out", "norm_q", ".y_embedder", "norm_post_ffn", "blocks.0.ffn.experts.w1_sdnq.weight", "blocks.0.ffn.experts.w2_sdnq.weight", "blocks.0.ffn.experts.w3_sdnq.weight", "blocks.1.ffn.experts.w1_sdnq.weight", "blocks.1.ffn.experts.w2_sdnq.weight", "blocks.1.ffn.experts.w3_sdnq.weight", "blocks.2.ffn.experts.w1_sdnq.weight", "blocks.2.ffn.experts.w2_sdnq.weight", "blocks.2.ffn.experts.w3_sdnq.weight", "blocks.3.ffn.experts.w1_sdnq.weight", "blocks.3.ffn.experts.w2_sdnq.weight", "blocks.3.ffn.experts.w3_sdnq.weight", "blocks.4.ffn.experts.w1_sdnq.weight", "blocks.4.ffn.experts.w2_sdnq.weight", "blocks.4.ffn.experts.w3_sdnq.weight", "blocks.5.ffn.experts.w1_sdnq.weight", "blocks.5.ffn.experts.w2_sdnq.weight", "blocks.5.ffn.experts.w3_sdnq.weight", "blocks.6.ffn.experts.w1_sdnq.weight", "blocks.6.ffn.experts.w2_sdnq.weight", "blocks.6.ffn.experts.w3_sdnq.weight", "blocks.7.ffn.experts.w1_sdnq.weight", "blocks.7.ffn.experts.w2_sdnq.weight", "blocks.7.ffn.experts.w3_sdnq.weight", "blocks.8.ffn.experts.w1_sdnq.weight", "blocks.8.ffn.experts.w2_sdnq.weight", "blocks.8.ffn.experts.w3_sdnq.weight", "blocks.9.ffn.experts.w1_sdnq.weight", "blocks.9.ffn.experts.w2_sdnq.weight", "blocks.9.ffn.experts.w3_sdnq.weight", "blocks.10.ffn.experts.w1_sdnq.weight", "blocks.10.ffn.experts.w2_sdnq.weight", "blocks.10.ffn.experts.w3_sdnq.weight", "blocks.11.ffn.experts.w1_sdnq.weight", "blocks.11.ffn.experts.w2_sdnq.weight", "blocks.11.ffn.experts.w3_sdnq.weight", "blocks.12.ffn.experts.w1_sdnq.weight", "blocks.12.ffn.experts.w2_sdnq.weight", "blocks.12.ffn.experts.w3_sdnq.weight", "blocks.13.ffn.experts.w1_sdnq.weight", "blocks.13.ffn.experts.w2_sdnq.weight", "blocks.13.ffn.experts.w3_sdnq.weight", "blocks.14.ffn.experts.w1_sdnq.weight", "blocks.14.ffn.experts.w2_sdnq.weight", "blocks.14.ffn.experts.w3_sdnq.weight", "blocks.15.ffn.experts.w1_sdnq.weight", "blocks.15.ffn.experts.w2_sdnq.weight", "blocks.15.ffn.experts.w3_sdnq.weight", "blocks.16.ffn.experts.w1_sdnq.weight", "blocks.16.ffn.experts.w2_sdnq.weight", "blocks.16.ffn.experts.w3_sdnq.weight", "blocks.17.ffn.experts.w1_sdnq.weight", "blocks.17.ffn.experts.w2_sdnq.weight", "blocks.17.ffn.experts.w3_sdnq.weight", "blocks.18.ffn.experts.w1_sdnq.weight", "blocks.18.ffn.experts.w2_sdnq.weight", "blocks.18.ffn.experts.w3_sdnq.weight", "blocks.19.ffn.experts.w1_sdnq.weight", "blocks.19.ffn.experts.w2_sdnq.weight", "blocks.19.ffn.experts.w3_sdnq.weight", "blocks.20.ffn.experts.w1_sdnq.weight", "blocks.20.ffn.experts.w2_sdnq.weight", "blocks.20.ffn.experts.w3_sdnq.weight", "blocks.21.ffn.experts.w1_sdnq.weight", "blocks.21.ffn.experts.w2_sdnq.weight", "blocks.21.ffn.experts.w3_sdnq.weight", "blocks.22.ffn.experts.w1_sdnq.weight", "blocks.22.ffn.experts.w2_sdnq.weight", "blocks.22.ffn.experts.w3_sdnq.weight", "blocks.23.ffn.experts.w1_sdnq.weight", "blocks.23.ffn.experts.w2_sdnq.weight", "blocks.23.ffn.experts.w3_sdnq.weight", "blocks.24.ffn.experts.w1_sdnq.weight", "blocks.24.ffn.experts.w2_sdnq.weight", "blocks.24.ffn.experts.w3_sdnq.weight", "blocks.25.ffn.experts.w1_sdnq.weight", "blocks.25.ffn.experts.w2_sdnq.weight", "blocks.25.ffn.experts.w3_sdnq.weight", "blocks.26.ffn.experts.w1_sdnq.weight", "blocks.26.ffn.experts.w2_sdnq.weight", "blocks.26.ffn.experts.w3_sdnq.weight", "blocks.27.ffn.experts.w1_sdnq.weight", "blocks.27.ffn.experts.w2_sdnq.weight", "blocks.27.ffn.experts.w3_sdnq.weight", "blocks.28.ffn.experts.w1_sdnq.weight", "blocks.28.ffn.experts.w2_sdnq.weight", "blocks.28.ffn.experts.w3_sdnq.weight", "blocks.29.ffn.experts.w1_sdnq.weight", "blocks.29.ffn.experts.w2_sdnq.weight", "blocks.29.ffn.experts.w3_sdnq.weight", "blocks.30.ffn.experts.w1_sdnq.weight", "blocks.30.ffn.experts.w2_sdnq.weight", "blocks.30.ffn.experts.w3_sdnq.weight", "blocks.31.ffn.experts.w1_sdnq.weight", "blocks.31.ffn.experts.w2_sdnq.weight", "blocks.31.ffn.experts.w3_sdnq.weight", "blocks.32.ffn.experts.w1_sdnq.weight", "blocks.32.ffn.experts.w2_sdnq.weight", "blocks.32.ffn.experts.w3_sdnq.weight", "blocks.33.ffn.experts.w1_sdnq.weight", "blocks.33.ffn.experts.w2_sdnq.weight", "blocks.33.ffn.experts.w3_sdnq.weight", "blocks.34.ffn.experts.w1_sdnq.weight", "blocks.34.ffn.experts.w2_sdnq.weight", "blocks.34.ffn.experts.w3_sdnq.weight", "blocks.35.ffn.experts.w1_sdnq.weight", "blocks.35.ffn.experts.w2_sdnq.weight", "blocks.35.ffn.experts.w3_sdnq.weight", "blocks.36.ffn.experts.w1_sdnq.weight", "blocks.36.ffn.experts.w2_sdnq.weight", "blocks.36.ffn.experts.w3_sdnq.weight", "blocks.37.ffn.experts.w1_sdnq.weight", "blocks.37.ffn.experts.w2_sdnq.weight", "blocks.37.ffn.experts.w3_sdnq.weight", "blocks.38.ffn.experts.w1_sdnq.weight", "blocks.38.ffn.experts.w2_sdnq.weight", "blocks.38.ffn.experts.w3_sdnq.weight", "blocks.39.ffn.experts.w1_sdnq.weight", "blocks.39.ffn.experts.w2_sdnq.weight", "blocks.39.ffn.experts.w3_sdnq.weight", "blocks.40.ffn.experts.w1_sdnq.weight", "blocks.40.ffn.experts.w2_sdnq.weight", "blocks.40.ffn.experts.w3_sdnq.weight", "blocks.41.ffn.experts.w1_sdnq.weight", "blocks.41.ffn.experts.w2_sdnq.weight", "blocks.41.ffn.experts.w3_sdnq.weight", "blocks.42.ffn.experts.w1_sdnq.weight", "blocks.42.ffn.experts.w2_sdnq.weight", "blocks.42.ffn.experts.w3_sdnq.weight", "blocks.43.ffn.experts.w1_sdnq.weight", "blocks.43.ffn.experts.w2_sdnq.weight", "blocks.43.ffn.experts.w3_sdnq.weight", "blocks.44.ffn.experts.w1_sdnq.weight", "blocks.44.ffn.experts.w2_sdnq.weight", "blocks.44.ffn.experts.w3_sdnq.weight", "blocks.45.ffn.experts.w1_sdnq.weight", "blocks.45.ffn.experts.w2_sdnq.weight", "blocks.45.ffn.experts.w3_sdnq.weight", "blocks.46.ffn.experts.w1_sdnq.weight", "blocks.46.ffn.experts.w2_sdnq.weight", "blocks.46.ffn.experts.w3_sdnq.weight", "blocks.47.ffn.experts.w1_sdnq.weight", "blocks.47.ffn.experts.w2_sdnq.weight", "blocks.47.ffn.experts.w3_sdnq.weight" ], "modules_to_not_use_matmul": [], "non_blocking": false, "quant_conv": false, "quant_embedding": false, "quant_method": "sdnq", "quantization_device": "cuda", "quantized_matmul_dtype": null, "return_device": "cpu", "sdnq_version": "0.2.2", "svd_rank": 32, "svd_steps": 8, "use_dynamic_quantization": false, "use_grad_ckpt": true, "use_hadamard": false, "use_quantized_matmul": false, "use_quantized_matmul_conv": false, "use_static_quantization": true, "use_stochastic_rounding": false, "use_svd": false, "weights_dtype": "uint4" }