| { |
| "_name_or_path": "mgoin/pixtral-12b", |
| "architectures": [ |
| "LlavaForConditionalGeneration" |
| ], |
| "ignore_index": -100, |
| "image_seq_length": 1, |
| "image_token_index": 10, |
| "model_type": "llava", |
| "projector_hidden_act": "gelu", |
| "quantization_config": { |
| "config_groups": { |
| "group_0": { |
| "input_activations": { |
| "actorder": null, |
| "block_structure": null, |
| "dynamic": true, |
| "group_size": null, |
| "num_bits": 8, |
| "observer": null, |
| "observer_kwargs": {}, |
| "strategy": "token", |
| "symmetric": true, |
| "type": "int" |
| }, |
| "output_activations": null, |
| "targets": [ |
| "Linear" |
| ], |
| "weights": { |
| "actorder": null, |
| "block_structure": null, |
| "dynamic": false, |
| "group_size": null, |
| "num_bits": 8, |
| "observer": "minmax", |
| "observer_kwargs": {}, |
| "strategy": "channel", |
| "symmetric": true, |
| "type": "int" |
| } |
| } |
| }, |
| "format": "int-quantized", |
| "global_compression_ratio": 1.254986373686197, |
| "ignore": [ |
| "vision_tower.transformer.layers.0.feed_forward.gate_proj", |
| "vision_tower.transformer.layers.0.feed_forward.up_proj", |
| "vision_tower.transformer.layers.0.feed_forward.down_proj", |
| "vision_tower.transformer.layers.0.attention.k_proj", |
| "vision_tower.transformer.layers.0.attention.v_proj", |
| "vision_tower.transformer.layers.0.attention.q_proj", |
| "vision_tower.transformer.layers.0.attention.o_proj", |
| "vision_tower.transformer.layers.1.feed_forward.gate_proj", |
| "vision_tower.transformer.layers.1.feed_forward.up_proj", |
| "vision_tower.transformer.layers.1.feed_forward.down_proj", |
| "vision_tower.transformer.layers.1.attention.k_proj", |
| "vision_tower.transformer.layers.1.attention.v_proj", |
| "vision_tower.transformer.layers.1.attention.q_proj", |
| "vision_tower.transformer.layers.1.attention.o_proj", |
| "vision_tower.transformer.layers.2.feed_forward.gate_proj", |
| "vision_tower.transformer.layers.2.feed_forward.up_proj", |
| "vision_tower.transformer.layers.2.feed_forward.down_proj", |
| "vision_tower.transformer.layers.2.attention.k_proj", |
| "vision_tower.transformer.layers.2.attention.v_proj", |
| "vision_tower.transformer.layers.2.attention.q_proj", |
| "vision_tower.transformer.layers.2.attention.o_proj", |
| "vision_tower.transformer.layers.3.feed_forward.gate_proj", |
| "vision_tower.transformer.layers.3.feed_forward.up_proj", |
| "vision_tower.transformer.layers.3.feed_forward.down_proj", |
| "vision_tower.transformer.layers.3.attention.k_proj", |
| "vision_tower.transformer.layers.3.attention.v_proj", |
| "vision_tower.transformer.layers.3.attention.q_proj", |
| "vision_tower.transformer.layers.3.attention.o_proj", |
| "vision_tower.transformer.layers.4.feed_forward.gate_proj", |
| "vision_tower.transformer.layers.4.feed_forward.up_proj", |
| "vision_tower.transformer.layers.4.feed_forward.down_proj", |
| "vision_tower.transformer.layers.4.attention.k_proj", |
| "vision_tower.transformer.layers.4.attention.v_proj", |
| "vision_tower.transformer.layers.4.attention.q_proj", |
| "vision_tower.transformer.layers.4.attention.o_proj", |
| "vision_tower.transformer.layers.5.feed_forward.gate_proj", |
| "vision_tower.transformer.layers.5.feed_forward.up_proj", |
| "vision_tower.transformer.layers.5.feed_forward.down_proj", |
| "vision_tower.transformer.layers.5.attention.k_proj", |
| "vision_tower.transformer.layers.5.attention.v_proj", |
| "vision_tower.transformer.layers.5.attention.q_proj", |
| "vision_tower.transformer.layers.5.attention.o_proj", |
| "vision_tower.transformer.layers.6.feed_forward.gate_proj", |
| "vision_tower.transformer.layers.6.feed_forward.up_proj", |
| "vision_tower.transformer.layers.6.feed_forward.down_proj", |
| "vision_tower.transformer.layers.6.attention.k_proj", |
| "vision_tower.transformer.layers.6.attention.v_proj", |
| "vision_tower.transformer.layers.6.attention.q_proj", |
| "vision_tower.transformer.layers.6.attention.o_proj", |
| "vision_tower.transformer.layers.7.feed_forward.gate_proj", |
| "vision_tower.transformer.layers.7.feed_forward.up_proj", |
| "vision_tower.transformer.layers.7.feed_forward.down_proj", |
| "vision_tower.transformer.layers.7.attention.k_proj", |
| "vision_tower.transformer.layers.7.attention.v_proj", |
| "vision_tower.transformer.layers.7.attention.q_proj", |
| "vision_tower.transformer.layers.7.attention.o_proj", |
| "vision_tower.transformer.layers.8.feed_forward.gate_proj", |
| "vision_tower.transformer.layers.8.feed_forward.up_proj", |
| "vision_tower.transformer.layers.8.feed_forward.down_proj", |
| "vision_tower.transformer.layers.8.attention.k_proj", |
| "vision_tower.transformer.layers.8.attention.v_proj", |
| "vision_tower.transformer.layers.8.attention.q_proj", |
| "vision_tower.transformer.layers.8.attention.o_proj", |
| "vision_tower.transformer.layers.9.feed_forward.gate_proj", |
| "vision_tower.transformer.layers.9.feed_forward.up_proj", |
| "vision_tower.transformer.layers.9.feed_forward.down_proj", |
| "vision_tower.transformer.layers.9.attention.k_proj", |
| "vision_tower.transformer.layers.9.attention.v_proj", |
| "vision_tower.transformer.layers.9.attention.q_proj", |
| "vision_tower.transformer.layers.9.attention.o_proj", |
| "vision_tower.transformer.layers.10.feed_forward.gate_proj", |
| "vision_tower.transformer.layers.10.feed_forward.up_proj", |
| "vision_tower.transformer.layers.10.feed_forward.down_proj", |
| "vision_tower.transformer.layers.10.attention.k_proj", |
| "vision_tower.transformer.layers.10.attention.v_proj", |
| "vision_tower.transformer.layers.10.attention.q_proj", |
| "vision_tower.transformer.layers.10.attention.o_proj", |
| "vision_tower.transformer.layers.11.feed_forward.gate_proj", |
| "vision_tower.transformer.layers.11.feed_forward.up_proj", |
| "vision_tower.transformer.layers.11.feed_forward.down_proj", |
| "vision_tower.transformer.layers.11.attention.k_proj", |
| "vision_tower.transformer.layers.11.attention.v_proj", |
| "vision_tower.transformer.layers.11.attention.q_proj", |
| "vision_tower.transformer.layers.11.attention.o_proj", |
| "vision_tower.transformer.layers.12.feed_forward.gate_proj", |
| "vision_tower.transformer.layers.12.feed_forward.up_proj", |
| "vision_tower.transformer.layers.12.feed_forward.down_proj", |
| "vision_tower.transformer.layers.12.attention.k_proj", |
| "vision_tower.transformer.layers.12.attention.v_proj", |
| "vision_tower.transformer.layers.12.attention.q_proj", |
| "vision_tower.transformer.layers.12.attention.o_proj", |
| "vision_tower.transformer.layers.13.feed_forward.gate_proj", |
| "vision_tower.transformer.layers.13.feed_forward.up_proj", |
| "vision_tower.transformer.layers.13.feed_forward.down_proj", |
| "vision_tower.transformer.layers.13.attention.k_proj", |
| "vision_tower.transformer.layers.13.attention.v_proj", |
| "vision_tower.transformer.layers.13.attention.q_proj", |
| "vision_tower.transformer.layers.13.attention.o_proj", |
| "vision_tower.transformer.layers.14.feed_forward.gate_proj", |
| "vision_tower.transformer.layers.14.feed_forward.up_proj", |
| "vision_tower.transformer.layers.14.feed_forward.down_proj", |
| "vision_tower.transformer.layers.14.attention.k_proj", |
| "vision_tower.transformer.layers.14.attention.v_proj", |
| "vision_tower.transformer.layers.14.attention.q_proj", |
| "vision_tower.transformer.layers.14.attention.o_proj", |
| "vision_tower.transformer.layers.15.feed_forward.gate_proj", |
| "vision_tower.transformer.layers.15.feed_forward.up_proj", |
| "vision_tower.transformer.layers.15.feed_forward.down_proj", |
| "vision_tower.transformer.layers.15.attention.k_proj", |
| "vision_tower.transformer.layers.15.attention.v_proj", |
| "vision_tower.transformer.layers.15.attention.q_proj", |
| "vision_tower.transformer.layers.15.attention.o_proj", |
| "vision_tower.transformer.layers.16.feed_forward.gate_proj", |
| "vision_tower.transformer.layers.16.feed_forward.up_proj", |
| "vision_tower.transformer.layers.16.feed_forward.down_proj", |
| "vision_tower.transformer.layers.16.attention.k_proj", |
| "vision_tower.transformer.layers.16.attention.v_proj", |
| "vision_tower.transformer.layers.16.attention.q_proj", |
| "vision_tower.transformer.layers.16.attention.o_proj", |
| "vision_tower.transformer.layers.17.feed_forward.gate_proj", |
| "vision_tower.transformer.layers.17.feed_forward.up_proj", |
| "vision_tower.transformer.layers.17.feed_forward.down_proj", |
| "vision_tower.transformer.layers.17.attention.k_proj", |
| "vision_tower.transformer.layers.17.attention.v_proj", |
| "vision_tower.transformer.layers.17.attention.q_proj", |
| "vision_tower.transformer.layers.17.attention.o_proj", |
| "vision_tower.transformer.layers.18.feed_forward.gate_proj", |
| "vision_tower.transformer.layers.18.feed_forward.up_proj", |
| "vision_tower.transformer.layers.18.feed_forward.down_proj", |
| "vision_tower.transformer.layers.18.attention.k_proj", |
| "vision_tower.transformer.layers.18.attention.v_proj", |
| "vision_tower.transformer.layers.18.attention.q_proj", |
| "vision_tower.transformer.layers.18.attention.o_proj", |
| "vision_tower.transformer.layers.19.feed_forward.gate_proj", |
| "vision_tower.transformer.layers.19.feed_forward.up_proj", |
| "vision_tower.transformer.layers.19.feed_forward.down_proj", |
| "vision_tower.transformer.layers.19.attention.k_proj", |
| "vision_tower.transformer.layers.19.attention.v_proj", |
| "vision_tower.transformer.layers.19.attention.q_proj", |
| "vision_tower.transformer.layers.19.attention.o_proj", |
| "vision_tower.transformer.layers.20.feed_forward.gate_proj", |
| "vision_tower.transformer.layers.20.feed_forward.up_proj", |
| "vision_tower.transformer.layers.20.feed_forward.down_proj", |
| "vision_tower.transformer.layers.20.attention.k_proj", |
| "vision_tower.transformer.layers.20.attention.v_proj", |
| "vision_tower.transformer.layers.20.attention.q_proj", |
| "vision_tower.transformer.layers.20.attention.o_proj", |
| "vision_tower.transformer.layers.21.feed_forward.gate_proj", |
| "vision_tower.transformer.layers.21.feed_forward.up_proj", |
| "vision_tower.transformer.layers.21.feed_forward.down_proj", |
| "vision_tower.transformer.layers.21.attention.k_proj", |
| "vision_tower.transformer.layers.21.attention.v_proj", |
| "vision_tower.transformer.layers.21.attention.q_proj", |
| "vision_tower.transformer.layers.21.attention.o_proj", |
| "vision_tower.transformer.layers.22.feed_forward.gate_proj", |
| "vision_tower.transformer.layers.22.feed_forward.up_proj", |
| "vision_tower.transformer.layers.22.feed_forward.down_proj", |
| "vision_tower.transformer.layers.22.attention.k_proj", |
| "vision_tower.transformer.layers.22.attention.v_proj", |
| "vision_tower.transformer.layers.22.attention.q_proj", |
| "vision_tower.transformer.layers.22.attention.o_proj", |
| "vision_tower.transformer.layers.23.feed_forward.gate_proj", |
| "vision_tower.transformer.layers.23.feed_forward.up_proj", |
| "vision_tower.transformer.layers.23.feed_forward.down_proj", |
| "vision_tower.transformer.layers.23.attention.k_proj", |
| "vision_tower.transformer.layers.23.attention.v_proj", |
| "vision_tower.transformer.layers.23.attention.q_proj", |
| "vision_tower.transformer.layers.23.attention.o_proj", |
| "multi_modal_projector.linear_1", |
| "multi_modal_projector.linear_2", |
| "language_model.lm_head" |
| ], |
| "kv_cache_scheme": null, |
| "quant_method": "compressed-tensors", |
| "quantization_status": "compressed" |
| }, |
| "text_config": { |
| "hidden_size": 5120, |
| "intermediate_size": 14336, |
| "is_composition": true, |
| "max_position_embeddings": 1024000, |
| "model_type": "mistral", |
| "num_hidden_layers": 40, |
| "num_key_value_heads": 8, |
| "rms_norm_eps": 1e-05, |
| "rope_theta": 1000000000.0, |
| "sliding_window": null, |
| "vocab_size": 131072 |
| }, |
| "torch_dtype": "bfloat16", |
| "transformers_version": "4.47.0", |
| "vision_config": { |
| "head_dim": 64, |
| "hidden_act": "silu", |
| "image_size": 1024, |
| "is_composition": true, |
| "model_type": "pixtral", |
| "patch_size": 16, |
| "rope_theta": 10000.0, |
| "tie_word_embeddings": false |
| }, |
| "vision_feature_layer": -1, |
| "vision_feature_select_strategy": "full" |
| } |
