{ "architectures": [ "InklingForConditionalGeneration" ], "model_type": "inkling_mm_model", "eos_token_id": 200006, "text_config": { "model_max_length": 1048576, "torch_dtype": "bfloat16", "hidden_size": 6144, "num_hidden_layers": 66, "vocab_size": 201024, "num_attention_heads": 64, "num_key_value_heads": 8, "head_dim": 128, "d_rel": 16, "rel_extent": 1024, "q_bias": false, "o_bias": false, "log_scaling_n_floor": 128000, "log_scaling_alpha": 0.1, "rms_norm_eps": 1e-06, "use_embed_norm": true, "local_layer_ids": [ 0, 1, 2, 3, 4, 6, 7, 8, 9, 10, 12, 13, 14, 15, 16, 18, 19, 20, 21, 22, 24, 25, 26, 27, 28, 30, 31, 32, 33, 34, 36, 37, 38, 39, 40, 42, 43, 44, 45, 46, 48, 49, 50, 51, 52, 54, 55, 56, 57, 58, 60, 61, 62, 63, 64 ], "dense_mlp_idx": 2, "use_sconv": true, "sconv_kernel_size": 4, "unpadded_vocab_size": 200058, "logits_mup_width_multiplier": 24.0, "final_logit_softcapping": null, "swa_head_dim": 128, "swa_num_attention_heads": 64, "swa_num_key_value_heads": 16, "sliding_window_size": 512, "n_routed_experts": 256, "num_experts_per_tok": 6, "n_shared_experts": 2, "shared_expert_sink": true, "dense_intermediate_size": 24576, "intermediate_size": 3072, "route_scale": 8.0, "use_gate_bias": true, "gate_activation": "sigmoid", "norm_after_topk": true, "use_global_scale": true }, "audio_config": { "decoder_dmodel": 6144, "n_mel_bins": 80, "mel_vocab_size": 16, "bias": false, "dmel_min_value": -7.0, "dmel_max_value": 2.0, "use_audio_norm": true, "audio_mode": "dmel" }, "vision_config": { "vision_encoder_type": "hmlp", "decoder_dmodel": 6144, "patch_size": 40, "temporal_patch_size": 2, "n_channels": 3, "n_layers": 4, "use_vision_norm": true }, "mtp_config": { "num_nextn_predict_layers": 8, "chain_hidden_post_norm": false, "local_layer_ids": [ 0, 2, 4, 5, 6, 7 ] }, "quantization": { "group_size": 64, "bits": 4 }, "mlx_conversion": { "source": "thinkingmachines/Inkling", "scheme": "bf16->mlx_affine", "bits": 4, "group_size": 64, "scope": "text_backbone_only", "quantized_modules": [ "model.llm.layers.10.mlp.experts.w13_weight", "model.llm.layers.10.mlp.experts.w2_weight", "model.llm.layers.11.mlp.experts.w13_weight", "model.llm.layers.11.mlp.experts.w2_weight", "model.llm.layers.12.mlp.experts.w13_weight", "model.llm.layers.12.mlp.experts.w2_weight", "model.llm.layers.13.mlp.experts.w13_weight", "model.llm.layers.13.mlp.experts.w2_weight", "model.llm.layers.14.mlp.experts.w13_weight", "model.llm.layers.14.mlp.experts.w2_weight", "model.llm.layers.15.mlp.experts.w13_weight", "model.llm.layers.15.mlp.experts.w2_weight", "model.llm.layers.16.mlp.experts.w13_weight", "model.llm.layers.16.mlp.experts.w2_weight", "model.llm.layers.17.mlp.experts.w13_weight", "model.llm.layers.17.mlp.experts.w2_weight", "model.llm.layers.18.mlp.experts.w13_weight", "model.llm.layers.18.mlp.experts.w2_weight", "model.llm.layers.19.mlp.experts.w13_weight", "model.llm.layers.19.mlp.experts.w2_weight", "model.llm.layers.2.mlp.experts.w13_weight", "model.llm.layers.2.mlp.experts.w2_weight", "model.llm.layers.20.mlp.experts.w13_weight", "model.llm.layers.20.mlp.experts.w2_weight", "model.llm.layers.21.mlp.experts.w13_weight", "model.llm.layers.21.mlp.experts.w2_weight", "model.llm.layers.22.mlp.experts.w13_weight", "model.llm.layers.22.mlp.experts.w2_weight", "model.llm.layers.23.mlp.experts.w13_weight", "model.llm.layers.23.mlp.experts.w2_weight", "model.llm.layers.24.mlp.experts.w13_weight", "model.llm.layers.24.mlp.experts.w2_weight", "model.llm.layers.25.mlp.experts.w13_weight", "model.llm.layers.25.mlp.experts.w2_weight", "model.llm.layers.26.mlp.experts.w13_weight", "model.llm.layers.26.mlp.experts.w2_weight", "model.llm.layers.27.mlp.experts.w13_weight", "model.llm.layers.27.mlp.experts.w2_weight", "model.llm.layers.28.mlp.experts.w13_weight", "model.llm.layers.28.mlp.experts.w2_weight", "model.llm.layers.29.mlp.experts.w13_weight", "model.llm.layers.29.mlp.experts.w2_weight", "model.llm.layers.3.mlp.experts.w13_weight", "model.llm.layers.3.mlp.experts.w2_weight", "model.llm.layers.30.mlp.experts.w13_weight", "model.llm.layers.30.mlp.experts.w2_weight", "model.llm.layers.31.mlp.experts.w13_weight", "model.llm.layers.31.mlp.experts.w2_weight", "model.llm.layers.32.mlp.experts.w13_weight", "model.llm.layers.32.mlp.experts.w2_weight", "model.llm.layers.33.mlp.experts.w13_weight", "model.llm.layers.33.mlp.experts.w2_weight", "model.llm.layers.34.mlp.experts.w13_weight", "model.llm.layers.34.mlp.experts.w2_weight", "model.llm.layers.35.mlp.experts.w13_weight", "model.llm.layers.35.mlp.experts.w2_weight", "model.llm.layers.36.mlp.experts.w13_weight", "model.llm.layers.36.mlp.experts.w2_weight", "model.llm.layers.37.mlp.experts.w13_weight", "model.llm.layers.37.mlp.experts.w2_weight", "model.llm.layers.38.mlp.experts.w13_weight", "model.llm.layers.38.mlp.experts.w2_weight", "model.llm.layers.39.mlp.experts.w13_weight", "model.llm.layers.39.mlp.experts.w2_weight", "model.llm.layers.4.mlp.experts.w13_weight", "model.llm.layers.4.mlp.experts.w2_weight", "model.llm.layers.40.mlp.experts.w13_weight", "model.llm.layers.40.mlp.experts.w2_weight", "model.llm.layers.41.mlp.experts.w13_weight", "model.llm.layers.41.mlp.experts.w2_weight", "model.llm.layers.42.mlp.experts.w13_weight", "model.llm.layers.42.mlp.experts.w2_weight", "model.llm.layers.43.mlp.experts.w13_weight", "model.llm.layers.43.mlp.experts.w2_weight", "model.llm.layers.44.mlp.experts.w13_weight", "model.llm.layers.44.mlp.experts.w2_weight", "model.llm.layers.45.mlp.experts.w13_weight", "model.llm.layers.45.mlp.experts.w2_weight", "model.llm.layers.46.mlp.experts.w13_weight", "model.llm.layers.46.mlp.experts.w2_weight", "model.llm.layers.47.mlp.experts.w13_weight", "model.llm.layers.47.mlp.experts.w2_weight", "model.llm.layers.48.mlp.experts.w13_weight", "model.llm.layers.48.mlp.experts.w2_weight", "model.llm.layers.49.mlp.experts.w13_weight", "model.llm.layers.49.mlp.experts.w2_weight", "model.llm.layers.5.mlp.experts.w13_weight", "model.llm.layers.5.mlp.experts.w2_weight", "model.llm.layers.50.mlp.experts.w13_weight", "model.llm.layers.50.mlp.experts.w2_weight", "model.llm.layers.51.mlp.experts.w13_weight", "model.llm.layers.51.mlp.experts.w2_weight", "model.llm.layers.52.mlp.experts.w13_weight", "model.llm.layers.52.mlp.experts.w2_weight", "model.llm.layers.53.mlp.experts.w13_weight", "model.llm.layers.53.mlp.experts.w2_weight", "model.llm.layers.54.mlp.experts.w13_weight", "model.llm.layers.54.mlp.experts.w2_weight", "model.llm.layers.55.mlp.experts.w13_weight", "model.llm.layers.55.mlp.experts.w2_weight", "model.llm.layers.56.mlp.experts.w13_weight", "model.llm.layers.56.mlp.experts.w2_weight", "model.llm.layers.57.mlp.experts.w13_weight", "model.llm.layers.57.mlp.experts.w2_weight", "model.llm.layers.58.mlp.experts.w13_weight", "model.llm.layers.58.mlp.experts.w2_weight", "model.llm.layers.59.mlp.experts.w13_weight", "model.llm.layers.59.mlp.experts.w2_weight", "model.llm.layers.6.mlp.experts.w13_weight", "model.llm.layers.6.mlp.experts.w2_weight", "model.llm.layers.60.mlp.experts.w13_weight", "model.llm.layers.60.mlp.experts.w2_weight", "model.llm.layers.61.mlp.experts.w13_weight", "model.llm.layers.61.mlp.experts.w2_weight", "model.llm.layers.62.mlp.experts.w13_weight", "model.llm.layers.62.mlp.experts.w2_weight", "model.llm.layers.63.mlp.experts.w13_weight", "model.llm.layers.63.mlp.experts.w2_weight", "model.llm.layers.64.mlp.experts.w13_weight", "model.llm.layers.64.mlp.experts.w2_weight", "model.llm.layers.65.mlp.experts.w13_weight", "model.llm.layers.65.mlp.experts.w2_weight", "model.llm.layers.7.mlp.experts.w13_weight", "model.llm.layers.7.mlp.experts.w2_weight", "model.llm.layers.8.mlp.experts.w13_weight", "model.llm.layers.8.mlp.experts.w2_weight", "model.llm.layers.9.mlp.experts.w13_weight", "model.llm.layers.9.mlp.experts.w2_weight" ], "bf16_fallback_modules": [] } }