| { |
| "architectures": [ |
| "Phi4MMForCausalLM" |
| ], |
| "attention_bias": false, |
| "attention_dropout": 0.0, |
| "audio_processor": { |
| "config": { |
| "activation": "swish", |
| "activation_checkpointing": { |
| "interval": 1, |
| "module": "transformer", |
| "offload": false |
| }, |
| "attention_dim": 1024, |
| "attention_heads": 16, |
| "batch_norm": false, |
| "bias_in_glu": true, |
| "causal": true, |
| "chunk_size": -1, |
| "cnn_layer_norm": true, |
| "conv_activation": "swish", |
| "conv_glu_type": "swish", |
| "depthwise_multiplier": 1, |
| "depthwise_seperable_out_channel": 1024, |
| "dropout_rate": 0.0, |
| "encoder_embedding_config": { |
| "input_size": 80 |
| }, |
| "ext_pw_kernel_size": 1, |
| "ext_pw_out_channel": 1024, |
| "input_layer": "nemo_conv", |
| "input_size": 80, |
| "kernel_size": 3, |
| "left_chunk": 18, |
| "linear_units": 1536, |
| "nemo_conv_settings": { |
| "conv_channels": 1024 |
| }, |
| "num_blocks": 24, |
| "relative_attention_bias_args": { |
| "t5_bias_max_distance": 500, |
| "type": "t5" |
| }, |
| "time_reduction": 8 |
| }, |
| "name": "cascades" |
| }, |
| "auto_map": { |
| "AutoConfig": "configuration_phi4mm.Phi4MMConfig", |
| "AutoModelForCausalLM": "modeling_phi4mm.Phi4MMForCausalLM", |
| "AutoTokenizer": "Xenova/gpt-4o" |
| }, |
| "bos_token_id": 199999, |
| "dtype": "bfloat16", |
| "embd_layer": { |
| "audio_embd_layer": { |
| "compression_rate": 8, |
| "downsample_rate": 1, |
| "embedding_cls": "audio", |
| "enable_gradient_checkpointing": true, |
| "projection_cls": "mlp", |
| "use_conv_downsample": false, |
| "use_qformer": false |
| }, |
| "embedding_cls": "image_audio", |
| "image_embd_layer": { |
| "crop_size": 448, |
| "embedding_cls": "tune_image", |
| "enable_gradient_checkpointing": true, |
| "hd_transform_order": "sub_glb", |
| "image_token_compression_cls": "avg_pool_2d", |
| "projection_cls": "mlp", |
| "use_hd_transform": true, |
| "with_learnable_separator": true |
| } |
| }, |
| "embd_pdrop": 0.0, |
| "eos_token_id": 199999, |
| "full_attn_mod": 1, |
| "hidden_act": "silu", |
| "hidden_size": 3072, |
| "img_processor": null, |
| "initializer_range": 0.02, |
| "intermediate_size": 8192, |
| "interpolate_factor": 1, |
| "lm_head_bias": false, |
| "max_position_embeddings": 131072, |
| "mlp_bias": false, |
| "model_type": "phi4mm", |
| "num_attention_heads": 24, |
| "num_hidden_layers": 32, |
| "num_key_value_heads": 8, |
| "original_max_position_embeddings": 4096, |
| "pad_token_id": 199999, |
| "partial_rotary_factor": 0.75, |
| "quantization_config": { |
| "config_groups": { |
| "group_0": { |
| "format": "pack-quantized", |
| "input_activations": null, |
| "output_activations": null, |
| "targets": [ |
| "Linear" |
| ], |
| "weights": { |
| "actorder": "static", |
| "block_structure": null, |
| "dynamic": false, |
| "group_size": 128, |
| "num_bits": 4, |
| "observer": "memoryless_minmax", |
| "observer_kwargs": {}, |
| "scale_dtype": null, |
| "strategy": "group", |
| "symmetric": true, |
| "type": "int", |
| "zp_dtype": null |
| } |
| } |
| }, |
| "format": "pack-quantized", |
| "global_compression_ratio": null, |
| "ignore": [ |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.0.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.0.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.0.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.0.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.0.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.0.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.1.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.1.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.1.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.1.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.1.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.1.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.2.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.2.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.2.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.2.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.2.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.2.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.3.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.3.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.3.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.3.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.3.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.3.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.4.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.4.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.4.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.4.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.4.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.4.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.5.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.5.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.5.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.5.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.5.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.5.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.6.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.6.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.6.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.6.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.6.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.6.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.7.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.7.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.7.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.7.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.7.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.7.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.8.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.8.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.8.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.8.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.8.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.8.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.9.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.9.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.9.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.9.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.9.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.9.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.10.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.10.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.10.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.10.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.10.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.10.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.11.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.11.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.11.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.11.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.11.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.11.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.12.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.12.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.12.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.12.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.12.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.12.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.13.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.13.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.13.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.13.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.13.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.13.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.14.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.14.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.14.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.14.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.14.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.14.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.15.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.15.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.15.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.15.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.15.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.15.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.16.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.16.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.16.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.16.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.16.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.16.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.17.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.17.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.17.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.17.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.17.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.17.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.18.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.18.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.18.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.18.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.18.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.18.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.19.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.19.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.19.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.19.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.19.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.19.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.20.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.20.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.20.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.20.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.20.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.20.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.21.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.21.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.21.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.21.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.21.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.21.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.22.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.22.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.22.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.22.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.22.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.22.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.23.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.23.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.23.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.23.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.23.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.23.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.24.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.24.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.24.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.24.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.24.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.24.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.25.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.25.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.25.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.25.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.25.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.25.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.26.self_attn.k_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.26.self_attn.v_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.26.self_attn.q_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.26.self_attn.out_proj", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.26.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.encoder.layers.26.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_processor.head.mlp.fc1", |
| "model.embed_tokens_extend.image_embed.img_processor.head.mlp.fc2", |
| "model.embed_tokens_extend.image_embed.img_projection.0", |
| "model.embed_tokens_extend.image_embed.img_projection.2", |
| "model.embed_tokens_extend.audio_embed.encoder.embed.out", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.0._checkpoint_wrapped_module.feed_forward_in.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.0._checkpoint_wrapped_module.feed_forward_in.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.0._checkpoint_wrapped_module.self_attn.linear_q", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.0._checkpoint_wrapped_module.self_attn.linear_k", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.0._checkpoint_wrapped_module.self_attn.linear_v", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.0._checkpoint_wrapped_module.self_attn.linear_out", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.0._checkpoint_wrapped_module.feed_forward_out.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.0._checkpoint_wrapped_module.feed_forward_out.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.1._checkpoint_wrapped_module.feed_forward_in.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.1._checkpoint_wrapped_module.feed_forward_in.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.1._checkpoint_wrapped_module.self_attn.linear_q", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.1._checkpoint_wrapped_module.self_attn.linear_k", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.1._checkpoint_wrapped_module.self_attn.linear_v", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.1._checkpoint_wrapped_module.self_attn.linear_out", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.1._checkpoint_wrapped_module.feed_forward_out.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.1._checkpoint_wrapped_module.feed_forward_out.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.2._checkpoint_wrapped_module.feed_forward_in.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.2._checkpoint_wrapped_module.feed_forward_in.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.2._checkpoint_wrapped_module.self_attn.linear_q", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.2._checkpoint_wrapped_module.self_attn.linear_k", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.2._checkpoint_wrapped_module.self_attn.linear_v", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.2._checkpoint_wrapped_module.self_attn.linear_out", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.2._checkpoint_wrapped_module.feed_forward_out.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.2._checkpoint_wrapped_module.feed_forward_out.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.3._checkpoint_wrapped_module.feed_forward_in.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.3._checkpoint_wrapped_module.feed_forward_in.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.3._checkpoint_wrapped_module.self_attn.linear_q", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.3._checkpoint_wrapped_module.self_attn.linear_k", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.3._checkpoint_wrapped_module.self_attn.linear_v", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.3._checkpoint_wrapped_module.self_attn.linear_out", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.3._checkpoint_wrapped_module.feed_forward_out.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.3._checkpoint_wrapped_module.feed_forward_out.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.4._checkpoint_wrapped_module.feed_forward_in.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.4._checkpoint_wrapped_module.feed_forward_in.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.4._checkpoint_wrapped_module.self_attn.linear_q", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.4._checkpoint_wrapped_module.self_attn.linear_k", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.4._checkpoint_wrapped_module.self_attn.linear_v", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.4._checkpoint_wrapped_module.self_attn.linear_out", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.4._checkpoint_wrapped_module.feed_forward_out.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.4._checkpoint_wrapped_module.feed_forward_out.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.5._checkpoint_wrapped_module.feed_forward_in.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.5._checkpoint_wrapped_module.feed_forward_in.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.5._checkpoint_wrapped_module.self_attn.linear_q", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.5._checkpoint_wrapped_module.self_attn.linear_k", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.5._checkpoint_wrapped_module.self_attn.linear_v", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.5._checkpoint_wrapped_module.self_attn.linear_out", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.5._checkpoint_wrapped_module.feed_forward_out.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.5._checkpoint_wrapped_module.feed_forward_out.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.6._checkpoint_wrapped_module.feed_forward_in.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.6._checkpoint_wrapped_module.feed_forward_in.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.6._checkpoint_wrapped_module.self_attn.linear_q", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.6._checkpoint_wrapped_module.self_attn.linear_k", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.6._checkpoint_wrapped_module.self_attn.linear_v", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.6._checkpoint_wrapped_module.self_attn.linear_out", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.6._checkpoint_wrapped_module.feed_forward_out.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.6._checkpoint_wrapped_module.feed_forward_out.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.7._checkpoint_wrapped_module.feed_forward_in.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.7._checkpoint_wrapped_module.feed_forward_in.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.7._checkpoint_wrapped_module.self_attn.linear_q", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.7._checkpoint_wrapped_module.self_attn.linear_k", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.7._checkpoint_wrapped_module.self_attn.linear_v", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.7._checkpoint_wrapped_module.self_attn.linear_out", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.7._checkpoint_wrapped_module.feed_forward_out.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.7._checkpoint_wrapped_module.feed_forward_out.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.8._checkpoint_wrapped_module.feed_forward_in.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.8._checkpoint_wrapped_module.feed_forward_in.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.8._checkpoint_wrapped_module.self_attn.linear_q", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.8._checkpoint_wrapped_module.self_attn.linear_k", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.8._checkpoint_wrapped_module.self_attn.linear_v", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.8._checkpoint_wrapped_module.self_attn.linear_out", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.8._checkpoint_wrapped_module.feed_forward_out.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.8._checkpoint_wrapped_module.feed_forward_out.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.9._checkpoint_wrapped_module.feed_forward_in.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.9._checkpoint_wrapped_module.feed_forward_in.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.9._checkpoint_wrapped_module.self_attn.linear_q", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.9._checkpoint_wrapped_module.self_attn.linear_k", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.9._checkpoint_wrapped_module.self_attn.linear_v", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.9._checkpoint_wrapped_module.self_attn.linear_out", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.9._checkpoint_wrapped_module.feed_forward_out.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.9._checkpoint_wrapped_module.feed_forward_out.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.10._checkpoint_wrapped_module.feed_forward_in.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.10._checkpoint_wrapped_module.feed_forward_in.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.10._checkpoint_wrapped_module.self_attn.linear_q", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.10._checkpoint_wrapped_module.self_attn.linear_k", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.10._checkpoint_wrapped_module.self_attn.linear_v", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.10._checkpoint_wrapped_module.self_attn.linear_out", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.10._checkpoint_wrapped_module.feed_forward_out.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.10._checkpoint_wrapped_module.feed_forward_out.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.11._checkpoint_wrapped_module.feed_forward_in.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.11._checkpoint_wrapped_module.feed_forward_in.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.11._checkpoint_wrapped_module.self_attn.linear_q", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.11._checkpoint_wrapped_module.self_attn.linear_k", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.11._checkpoint_wrapped_module.self_attn.linear_v", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.11._checkpoint_wrapped_module.self_attn.linear_out", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.11._checkpoint_wrapped_module.feed_forward_out.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.11._checkpoint_wrapped_module.feed_forward_out.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.12._checkpoint_wrapped_module.feed_forward_in.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.12._checkpoint_wrapped_module.feed_forward_in.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.12._checkpoint_wrapped_module.self_attn.linear_q", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.12._checkpoint_wrapped_module.self_attn.linear_k", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.12._checkpoint_wrapped_module.self_attn.linear_v", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.12._checkpoint_wrapped_module.self_attn.linear_out", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.12._checkpoint_wrapped_module.feed_forward_out.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.12._checkpoint_wrapped_module.feed_forward_out.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.13._checkpoint_wrapped_module.feed_forward_in.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.13._checkpoint_wrapped_module.feed_forward_in.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.13._checkpoint_wrapped_module.self_attn.linear_q", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.13._checkpoint_wrapped_module.self_attn.linear_k", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.13._checkpoint_wrapped_module.self_attn.linear_v", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.13._checkpoint_wrapped_module.self_attn.linear_out", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.13._checkpoint_wrapped_module.feed_forward_out.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.13._checkpoint_wrapped_module.feed_forward_out.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.14._checkpoint_wrapped_module.feed_forward_in.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.14._checkpoint_wrapped_module.feed_forward_in.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.14._checkpoint_wrapped_module.self_attn.linear_q", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.14._checkpoint_wrapped_module.self_attn.linear_k", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.14._checkpoint_wrapped_module.self_attn.linear_v", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.14._checkpoint_wrapped_module.self_attn.linear_out", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.14._checkpoint_wrapped_module.feed_forward_out.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.14._checkpoint_wrapped_module.feed_forward_out.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.15._checkpoint_wrapped_module.feed_forward_in.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.15._checkpoint_wrapped_module.feed_forward_in.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.15._checkpoint_wrapped_module.self_attn.linear_q", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.15._checkpoint_wrapped_module.self_attn.linear_k", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.15._checkpoint_wrapped_module.self_attn.linear_v", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.15._checkpoint_wrapped_module.self_attn.linear_out", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.15._checkpoint_wrapped_module.feed_forward_out.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.15._checkpoint_wrapped_module.feed_forward_out.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.16._checkpoint_wrapped_module.feed_forward_in.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.16._checkpoint_wrapped_module.feed_forward_in.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.16._checkpoint_wrapped_module.self_attn.linear_q", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.16._checkpoint_wrapped_module.self_attn.linear_k", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.16._checkpoint_wrapped_module.self_attn.linear_v", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.16._checkpoint_wrapped_module.self_attn.linear_out", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.16._checkpoint_wrapped_module.feed_forward_out.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.16._checkpoint_wrapped_module.feed_forward_out.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.17._checkpoint_wrapped_module.feed_forward_in.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.17._checkpoint_wrapped_module.feed_forward_in.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.17._checkpoint_wrapped_module.self_attn.linear_q", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.17._checkpoint_wrapped_module.self_attn.linear_k", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.17._checkpoint_wrapped_module.self_attn.linear_v", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.17._checkpoint_wrapped_module.self_attn.linear_out", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.17._checkpoint_wrapped_module.feed_forward_out.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.17._checkpoint_wrapped_module.feed_forward_out.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.18._checkpoint_wrapped_module.feed_forward_in.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.18._checkpoint_wrapped_module.feed_forward_in.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.18._checkpoint_wrapped_module.self_attn.linear_q", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.18._checkpoint_wrapped_module.self_attn.linear_k", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.18._checkpoint_wrapped_module.self_attn.linear_v", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.18._checkpoint_wrapped_module.self_attn.linear_out", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.18._checkpoint_wrapped_module.feed_forward_out.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.18._checkpoint_wrapped_module.feed_forward_out.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.19._checkpoint_wrapped_module.feed_forward_in.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.19._checkpoint_wrapped_module.feed_forward_in.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.19._checkpoint_wrapped_module.self_attn.linear_q", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.19._checkpoint_wrapped_module.self_attn.linear_k", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.19._checkpoint_wrapped_module.self_attn.linear_v", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.19._checkpoint_wrapped_module.self_attn.linear_out", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.19._checkpoint_wrapped_module.feed_forward_out.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.19._checkpoint_wrapped_module.feed_forward_out.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.20._checkpoint_wrapped_module.feed_forward_in.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.20._checkpoint_wrapped_module.feed_forward_in.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.20._checkpoint_wrapped_module.self_attn.linear_q", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.20._checkpoint_wrapped_module.self_attn.linear_k", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.20._checkpoint_wrapped_module.self_attn.linear_v", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.20._checkpoint_wrapped_module.self_attn.linear_out", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.20._checkpoint_wrapped_module.feed_forward_out.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.20._checkpoint_wrapped_module.feed_forward_out.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.21._checkpoint_wrapped_module.feed_forward_in.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.21._checkpoint_wrapped_module.feed_forward_in.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.21._checkpoint_wrapped_module.self_attn.linear_q", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.21._checkpoint_wrapped_module.self_attn.linear_k", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.21._checkpoint_wrapped_module.self_attn.linear_v", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.21._checkpoint_wrapped_module.self_attn.linear_out", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.21._checkpoint_wrapped_module.feed_forward_out.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.21._checkpoint_wrapped_module.feed_forward_out.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.22._checkpoint_wrapped_module.feed_forward_in.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.22._checkpoint_wrapped_module.feed_forward_in.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.22._checkpoint_wrapped_module.self_attn.linear_q", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.22._checkpoint_wrapped_module.self_attn.linear_k", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.22._checkpoint_wrapped_module.self_attn.linear_v", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.22._checkpoint_wrapped_module.self_attn.linear_out", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.22._checkpoint_wrapped_module.feed_forward_out.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.22._checkpoint_wrapped_module.feed_forward_out.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.23._checkpoint_wrapped_module.feed_forward_in.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.23._checkpoint_wrapped_module.feed_forward_in.net.2", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.23._checkpoint_wrapped_module.self_attn.linear_q", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.23._checkpoint_wrapped_module.self_attn.linear_k", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.23._checkpoint_wrapped_module.self_attn.linear_v", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.23._checkpoint_wrapped_module.self_attn.linear_out", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.23._checkpoint_wrapped_module.feed_forward_out.net.0.linear", |
| "model.embed_tokens_extend.audio_embed.encoder.encoders.23._checkpoint_wrapped_module.feed_forward_out.net.2", |
| "model.embed_tokens_extend.audio_embed.audio_projection.speech.0", |
| "model.embed_tokens_extend.audio_embed.audio_projection.speech.2", |
| "model.embed_tokens_extend.audio_embed.audio_projection.vision.0", |
| "model.embed_tokens_extend.audio_embed.audio_projection.vision.2", |
| "model.layers.0.self_attn.o_proj.lora_A.vision", |
| "model.layers.0.self_attn.o_proj.lora_A.speech", |
| "model.layers.0.self_attn.o_proj.lora_B.vision", |
| "model.layers.0.self_attn.o_proj.lora_B.speech", |
| "model.layers.0.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.0.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.0.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.0.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.0.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.0.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.0.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.0.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.0.mlp.down_proj.lora_A.vision", |
| "model.layers.0.mlp.down_proj.lora_A.speech", |
| "model.layers.0.mlp.down_proj.lora_B.vision", |
| "model.layers.0.mlp.down_proj.lora_B.speech", |
| "model.layers.1.self_attn.o_proj.lora_A.vision", |
| "model.layers.1.self_attn.o_proj.lora_A.speech", |
| "model.layers.1.self_attn.o_proj.lora_B.vision", |
| "model.layers.1.self_attn.o_proj.lora_B.speech", |
| "model.layers.1.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.1.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.1.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.1.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.1.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.1.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.1.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.1.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.1.mlp.down_proj.lora_A.vision", |
| "model.layers.1.mlp.down_proj.lora_A.speech", |
| "model.layers.1.mlp.down_proj.lora_B.vision", |
| "model.layers.1.mlp.down_proj.lora_B.speech", |
| "model.layers.2.self_attn.o_proj.lora_A.vision", |
| "model.layers.2.self_attn.o_proj.lora_A.speech", |
| "model.layers.2.self_attn.o_proj.lora_B.vision", |
| "model.layers.2.self_attn.o_proj.lora_B.speech", |
| "model.layers.2.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.2.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.2.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.2.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.2.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.2.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.2.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.2.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.2.mlp.down_proj.lora_A.vision", |
| "model.layers.2.mlp.down_proj.lora_A.speech", |
| "model.layers.2.mlp.down_proj.lora_B.vision", |
| "model.layers.2.mlp.down_proj.lora_B.speech", |
| "model.layers.3.self_attn.o_proj.lora_A.vision", |
| "model.layers.3.self_attn.o_proj.lora_A.speech", |
| "model.layers.3.self_attn.o_proj.lora_B.vision", |
| "model.layers.3.self_attn.o_proj.lora_B.speech", |
| "model.layers.3.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.3.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.3.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.3.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.3.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.3.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.3.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.3.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.3.mlp.down_proj.lora_A.vision", |
| "model.layers.3.mlp.down_proj.lora_A.speech", |
| "model.layers.3.mlp.down_proj.lora_B.vision", |
| "model.layers.3.mlp.down_proj.lora_B.speech", |
| "model.layers.4.self_attn.o_proj.lora_A.vision", |
| "model.layers.4.self_attn.o_proj.lora_A.speech", |
| "model.layers.4.self_attn.o_proj.lora_B.vision", |
| "model.layers.4.self_attn.o_proj.lora_B.speech", |
| "model.layers.4.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.4.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.4.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.4.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.4.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.4.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.4.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.4.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.4.mlp.down_proj.lora_A.vision", |
| "model.layers.4.mlp.down_proj.lora_A.speech", |
| "model.layers.4.mlp.down_proj.lora_B.vision", |
| "model.layers.4.mlp.down_proj.lora_B.speech", |
| "model.layers.5.self_attn.o_proj.lora_A.vision", |
| "model.layers.5.self_attn.o_proj.lora_A.speech", |
| "model.layers.5.self_attn.o_proj.lora_B.vision", |
| "model.layers.5.self_attn.o_proj.lora_B.speech", |
| "model.layers.5.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.5.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.5.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.5.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.5.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.5.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.5.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.5.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.5.mlp.down_proj.lora_A.vision", |
| "model.layers.5.mlp.down_proj.lora_A.speech", |
| "model.layers.5.mlp.down_proj.lora_B.vision", |
| "model.layers.5.mlp.down_proj.lora_B.speech", |
| "model.layers.6.self_attn.o_proj.lora_A.vision", |
| "model.layers.6.self_attn.o_proj.lora_A.speech", |
| "model.layers.6.self_attn.o_proj.lora_B.vision", |
| "model.layers.6.self_attn.o_proj.lora_B.speech", |
| "model.layers.6.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.6.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.6.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.6.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.6.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.6.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.6.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.6.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.6.mlp.down_proj.lora_A.vision", |
| "model.layers.6.mlp.down_proj.lora_A.speech", |
| "model.layers.6.mlp.down_proj.lora_B.vision", |
| "model.layers.6.mlp.down_proj.lora_B.speech", |
| "model.layers.7.self_attn.o_proj.lora_A.vision", |
| "model.layers.7.self_attn.o_proj.lora_A.speech", |
| "model.layers.7.self_attn.o_proj.lora_B.vision", |
| "model.layers.7.self_attn.o_proj.lora_B.speech", |
| "model.layers.7.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.7.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.7.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.7.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.7.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.7.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.7.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.7.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.7.mlp.down_proj.lora_A.vision", |
| "model.layers.7.mlp.down_proj.lora_A.speech", |
| "model.layers.7.mlp.down_proj.lora_B.vision", |
| "model.layers.7.mlp.down_proj.lora_B.speech", |
| "model.layers.8.self_attn.o_proj.lora_A.vision", |
| "model.layers.8.self_attn.o_proj.lora_A.speech", |
| "model.layers.8.self_attn.o_proj.lora_B.vision", |
| "model.layers.8.self_attn.o_proj.lora_B.speech", |
| "model.layers.8.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.8.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.8.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.8.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.8.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.8.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.8.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.8.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.8.mlp.down_proj.lora_A.vision", |
| "model.layers.8.mlp.down_proj.lora_A.speech", |
| "model.layers.8.mlp.down_proj.lora_B.vision", |
| "model.layers.8.mlp.down_proj.lora_B.speech", |
| "model.layers.9.self_attn.o_proj.lora_A.vision", |
| "model.layers.9.self_attn.o_proj.lora_A.speech", |
| "model.layers.9.self_attn.o_proj.lora_B.vision", |
| "model.layers.9.self_attn.o_proj.lora_B.speech", |
| "model.layers.9.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.9.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.9.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.9.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.9.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.9.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.9.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.9.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.9.mlp.down_proj.lora_A.vision", |
| "model.layers.9.mlp.down_proj.lora_A.speech", |
| "model.layers.9.mlp.down_proj.lora_B.vision", |
| "model.layers.9.mlp.down_proj.lora_B.speech", |
| "model.layers.10.self_attn.o_proj.lora_A.vision", |
| "model.layers.10.self_attn.o_proj.lora_A.speech", |
| "model.layers.10.self_attn.o_proj.lora_B.vision", |
| "model.layers.10.self_attn.o_proj.lora_B.speech", |
| "model.layers.10.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.10.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.10.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.10.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.10.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.10.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.10.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.10.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.10.mlp.down_proj.lora_A.vision", |
| "model.layers.10.mlp.down_proj.lora_A.speech", |
| "model.layers.10.mlp.down_proj.lora_B.vision", |
| "model.layers.10.mlp.down_proj.lora_B.speech", |
| "model.layers.11.self_attn.o_proj.lora_A.vision", |
| "model.layers.11.self_attn.o_proj.lora_A.speech", |
| "model.layers.11.self_attn.o_proj.lora_B.vision", |
| "model.layers.11.self_attn.o_proj.lora_B.speech", |
| "model.layers.11.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.11.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.11.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.11.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.11.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.11.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.11.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.11.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.11.mlp.down_proj.lora_A.vision", |
| "model.layers.11.mlp.down_proj.lora_A.speech", |
| "model.layers.11.mlp.down_proj.lora_B.vision", |
| "model.layers.11.mlp.down_proj.lora_B.speech", |
| "model.layers.12.self_attn.o_proj.lora_A.vision", |
| "model.layers.12.self_attn.o_proj.lora_A.speech", |
| "model.layers.12.self_attn.o_proj.lora_B.vision", |
| "model.layers.12.self_attn.o_proj.lora_B.speech", |
| "model.layers.12.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.12.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.12.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.12.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.12.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.12.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.12.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.12.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.12.mlp.down_proj.lora_A.vision", |
| "model.layers.12.mlp.down_proj.lora_A.speech", |
| "model.layers.12.mlp.down_proj.lora_B.vision", |
| "model.layers.12.mlp.down_proj.lora_B.speech", |
| "model.layers.13.self_attn.o_proj.lora_A.vision", |
| "model.layers.13.self_attn.o_proj.lora_A.speech", |
| "model.layers.13.self_attn.o_proj.lora_B.vision", |
| "model.layers.13.self_attn.o_proj.lora_B.speech", |
| "model.layers.13.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.13.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.13.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.13.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.13.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.13.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.13.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.13.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.13.mlp.down_proj.lora_A.vision", |
| "model.layers.13.mlp.down_proj.lora_A.speech", |
| "model.layers.13.mlp.down_proj.lora_B.vision", |
| "model.layers.13.mlp.down_proj.lora_B.speech", |
| "model.layers.14.self_attn.o_proj.lora_A.vision", |
| "model.layers.14.self_attn.o_proj.lora_A.speech", |
| "model.layers.14.self_attn.o_proj.lora_B.vision", |
| "model.layers.14.self_attn.o_proj.lora_B.speech", |
| "model.layers.14.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.14.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.14.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.14.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.14.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.14.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.14.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.14.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.14.mlp.down_proj.lora_A.vision", |
| "model.layers.14.mlp.down_proj.lora_A.speech", |
| "model.layers.14.mlp.down_proj.lora_B.vision", |
| "model.layers.14.mlp.down_proj.lora_B.speech", |
| "model.layers.15.self_attn.o_proj.lora_A.vision", |
| "model.layers.15.self_attn.o_proj.lora_A.speech", |
| "model.layers.15.self_attn.o_proj.lora_B.vision", |
| "model.layers.15.self_attn.o_proj.lora_B.speech", |
| "model.layers.15.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.15.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.15.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.15.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.15.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.15.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.15.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.15.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.15.mlp.down_proj.lora_A.vision", |
| "model.layers.15.mlp.down_proj.lora_A.speech", |
| "model.layers.15.mlp.down_proj.lora_B.vision", |
| "model.layers.15.mlp.down_proj.lora_B.speech", |
| "model.layers.16.self_attn.o_proj.lora_A.vision", |
| "model.layers.16.self_attn.o_proj.lora_A.speech", |
| "model.layers.16.self_attn.o_proj.lora_B.vision", |
| "model.layers.16.self_attn.o_proj.lora_B.speech", |
| "model.layers.16.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.16.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.16.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.16.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.16.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.16.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.16.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.16.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.16.mlp.down_proj.lora_A.vision", |
| "model.layers.16.mlp.down_proj.lora_A.speech", |
| "model.layers.16.mlp.down_proj.lora_B.vision", |
| "model.layers.16.mlp.down_proj.lora_B.speech", |
| "model.layers.17.self_attn.o_proj.lora_A.vision", |
| "model.layers.17.self_attn.o_proj.lora_A.speech", |
| "model.layers.17.self_attn.o_proj.lora_B.vision", |
| "model.layers.17.self_attn.o_proj.lora_B.speech", |
| "model.layers.17.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.17.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.17.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.17.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.17.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.17.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.17.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.17.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.17.mlp.down_proj.lora_A.vision", |
| "model.layers.17.mlp.down_proj.lora_A.speech", |
| "model.layers.17.mlp.down_proj.lora_B.vision", |
| "model.layers.17.mlp.down_proj.lora_B.speech", |
| "model.layers.18.self_attn.o_proj.lora_A.vision", |
| "model.layers.18.self_attn.o_proj.lora_A.speech", |
| "model.layers.18.self_attn.o_proj.lora_B.vision", |
| "model.layers.18.self_attn.o_proj.lora_B.speech", |
| "model.layers.18.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.18.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.18.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.18.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.18.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.18.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.18.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.18.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.18.mlp.down_proj.lora_A.vision", |
| "model.layers.18.mlp.down_proj.lora_A.speech", |
| "model.layers.18.mlp.down_proj.lora_B.vision", |
| "model.layers.18.mlp.down_proj.lora_B.speech", |
| "model.layers.19.self_attn.o_proj.lora_A.vision", |
| "model.layers.19.self_attn.o_proj.lora_A.speech", |
| "model.layers.19.self_attn.o_proj.lora_B.vision", |
| "model.layers.19.self_attn.o_proj.lora_B.speech", |
| "model.layers.19.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.19.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.19.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.19.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.19.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.19.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.19.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.19.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.19.mlp.down_proj.lora_A.vision", |
| "model.layers.19.mlp.down_proj.lora_A.speech", |
| "model.layers.19.mlp.down_proj.lora_B.vision", |
| "model.layers.19.mlp.down_proj.lora_B.speech", |
| "model.layers.20.self_attn.o_proj.lora_A.vision", |
| "model.layers.20.self_attn.o_proj.lora_A.speech", |
| "model.layers.20.self_attn.o_proj.lora_B.vision", |
| "model.layers.20.self_attn.o_proj.lora_B.speech", |
| "model.layers.20.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.20.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.20.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.20.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.20.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.20.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.20.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.20.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.20.mlp.down_proj.lora_A.vision", |
| "model.layers.20.mlp.down_proj.lora_A.speech", |
| "model.layers.20.mlp.down_proj.lora_B.vision", |
| "model.layers.20.mlp.down_proj.lora_B.speech", |
| "model.layers.21.self_attn.o_proj.lora_A.vision", |
| "model.layers.21.self_attn.o_proj.lora_A.speech", |
| "model.layers.21.self_attn.o_proj.lora_B.vision", |
| "model.layers.21.self_attn.o_proj.lora_B.speech", |
| "model.layers.21.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.21.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.21.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.21.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.21.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.21.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.21.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.21.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.21.mlp.down_proj.lora_A.vision", |
| "model.layers.21.mlp.down_proj.lora_A.speech", |
| "model.layers.21.mlp.down_proj.lora_B.vision", |
| "model.layers.21.mlp.down_proj.lora_B.speech", |
| "model.layers.22.self_attn.o_proj.lora_A.vision", |
| "model.layers.22.self_attn.o_proj.lora_A.speech", |
| "model.layers.22.self_attn.o_proj.lora_B.vision", |
| "model.layers.22.self_attn.o_proj.lora_B.speech", |
| "model.layers.22.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.22.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.22.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.22.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.22.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.22.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.22.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.22.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.22.mlp.down_proj.lora_A.vision", |
| "model.layers.22.mlp.down_proj.lora_A.speech", |
| "model.layers.22.mlp.down_proj.lora_B.vision", |
| "model.layers.22.mlp.down_proj.lora_B.speech", |
| "model.layers.23.self_attn.o_proj.lora_A.vision", |
| "model.layers.23.self_attn.o_proj.lora_A.speech", |
| "model.layers.23.self_attn.o_proj.lora_B.vision", |
| "model.layers.23.self_attn.o_proj.lora_B.speech", |
| "model.layers.23.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.23.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.23.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.23.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.23.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.23.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.23.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.23.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.23.mlp.down_proj.lora_A.vision", |
| "model.layers.23.mlp.down_proj.lora_A.speech", |
| "model.layers.23.mlp.down_proj.lora_B.vision", |
| "model.layers.23.mlp.down_proj.lora_B.speech", |
| "model.layers.24.self_attn.o_proj.lora_A.vision", |
| "model.layers.24.self_attn.o_proj.lora_A.speech", |
| "model.layers.24.self_attn.o_proj.lora_B.vision", |
| "model.layers.24.self_attn.o_proj.lora_B.speech", |
| "model.layers.24.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.24.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.24.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.24.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.24.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.24.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.24.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.24.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.24.mlp.down_proj.lora_A.vision", |
| "model.layers.24.mlp.down_proj.lora_A.speech", |
| "model.layers.24.mlp.down_proj.lora_B.vision", |
| "model.layers.24.mlp.down_proj.lora_B.speech", |
| "model.layers.25.self_attn.o_proj.lora_A.vision", |
| "model.layers.25.self_attn.o_proj.lora_A.speech", |
| "model.layers.25.self_attn.o_proj.lora_B.vision", |
| "model.layers.25.self_attn.o_proj.lora_B.speech", |
| "model.layers.25.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.25.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.25.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.25.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.25.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.25.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.25.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.25.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.25.mlp.down_proj.lora_A.vision", |
| "model.layers.25.mlp.down_proj.lora_A.speech", |
| "model.layers.25.mlp.down_proj.lora_B.vision", |
| "model.layers.25.mlp.down_proj.lora_B.speech", |
| "model.layers.26.self_attn.o_proj.lora_A.vision", |
| "model.layers.26.self_attn.o_proj.lora_A.speech", |
| "model.layers.26.self_attn.o_proj.lora_B.vision", |
| "model.layers.26.self_attn.o_proj.lora_B.speech", |
| "model.layers.26.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.26.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.26.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.26.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.26.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.26.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.26.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.26.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.26.mlp.down_proj.lora_A.vision", |
| "model.layers.26.mlp.down_proj.lora_A.speech", |
| "model.layers.26.mlp.down_proj.lora_B.vision", |
| "model.layers.26.mlp.down_proj.lora_B.speech", |
| "model.layers.27.self_attn.o_proj.lora_A.vision", |
| "model.layers.27.self_attn.o_proj.lora_A.speech", |
| "model.layers.27.self_attn.o_proj.lora_B.vision", |
| "model.layers.27.self_attn.o_proj.lora_B.speech", |
| "model.layers.27.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.27.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.27.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.27.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.27.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.27.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.27.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.27.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.27.mlp.down_proj.lora_A.vision", |
| "model.layers.27.mlp.down_proj.lora_A.speech", |
| "model.layers.27.mlp.down_proj.lora_B.vision", |
| "model.layers.27.mlp.down_proj.lora_B.speech", |
| "model.layers.28.self_attn.o_proj.lora_A.vision", |
| "model.layers.28.self_attn.o_proj.lora_A.speech", |
| "model.layers.28.self_attn.o_proj.lora_B.vision", |
| "model.layers.28.self_attn.o_proj.lora_B.speech", |
| "model.layers.28.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.28.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.28.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.28.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.28.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.28.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.28.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.28.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.28.mlp.down_proj.lora_A.vision", |
| "model.layers.28.mlp.down_proj.lora_A.speech", |
| "model.layers.28.mlp.down_proj.lora_B.vision", |
| "model.layers.28.mlp.down_proj.lora_B.speech", |
| "model.layers.29.self_attn.o_proj.lora_A.vision", |
| "model.layers.29.self_attn.o_proj.lora_A.speech", |
| "model.layers.29.self_attn.o_proj.lora_B.vision", |
| "model.layers.29.self_attn.o_proj.lora_B.speech", |
| "model.layers.29.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.29.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.29.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.29.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.29.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.29.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.29.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.29.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.29.mlp.down_proj.lora_A.vision", |
| "model.layers.29.mlp.down_proj.lora_A.speech", |
| "model.layers.29.mlp.down_proj.lora_B.vision", |
| "model.layers.29.mlp.down_proj.lora_B.speech", |
| "model.layers.30.self_attn.o_proj.lora_A.vision", |
| "model.layers.30.self_attn.o_proj.lora_A.speech", |
| "model.layers.30.self_attn.o_proj.lora_B.vision", |
| "model.layers.30.self_attn.o_proj.lora_B.speech", |
| "model.layers.30.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.30.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.30.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.30.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.30.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.30.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.30.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.30.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.30.mlp.down_proj.lora_A.vision", |
| "model.layers.30.mlp.down_proj.lora_A.speech", |
| "model.layers.30.mlp.down_proj.lora_B.vision", |
| "model.layers.30.mlp.down_proj.lora_B.speech", |
| "model.layers.31.self_attn.o_proj.lora_A.vision", |
| "model.layers.31.self_attn.o_proj.lora_A.speech", |
| "model.layers.31.self_attn.o_proj.lora_B.vision", |
| "model.layers.31.self_attn.o_proj.lora_B.speech", |
| "model.layers.31.self_attn.qkv_proj.lora_A.vision", |
| "model.layers.31.self_attn.qkv_proj.lora_A.speech", |
| "model.layers.31.self_attn.qkv_proj.lora_B.vision", |
| "model.layers.31.self_attn.qkv_proj.lora_B.speech", |
| "model.layers.31.mlp.gate_up_proj.lora_A.vision", |
| "model.layers.31.mlp.gate_up_proj.lora_A.speech", |
| "model.layers.31.mlp.gate_up_proj.lora_B.vision", |
| "model.layers.31.mlp.gate_up_proj.lora_B.speech", |
| "model.layers.31.mlp.down_proj.lora_A.vision", |
| "model.layers.31.mlp.down_proj.lora_A.speech", |
| "model.layers.31.mlp.down_proj.lora_B.vision", |
| "model.layers.31.mlp.down_proj.lora_B.speech", |
| "lm_head" |
| ], |
| "kv_cache_scheme": null, |
| "quant_method": "compressed-tensors", |
| "quantization_status": "compressed", |
| "sparsity_config": {}, |
| "transform_config": {}, |
| "version": "0.14.0.1" |
| }, |
| "resid_pdrop": 0.0, |
| "rms_norm_eps": 1e-05, |
| "rope_scaling": { |
| "long_factor": [ |
| 1, |
| 1.118320672, |
| 1.250641126, |
| 1.398617824, |
| 1.564103225, |
| 1.74916897, |
| 1.956131817, |
| 2.187582649, |
| 2.446418898, |
| 2.735880826, |
| 3.059592084, |
| 3.421605075, |
| 3.826451687, |
| 4.279200023, |
| 4.785517845, |
| 5.351743533, |
| 5.984965424, |
| 6.693110555, |
| 7.485043894, |
| 8.370679318, |
| 9.36110372, |
| 10.4687158, |
| 11.70738129, |
| 13.09260651, |
| 14.64173252, |
| 16.37415215, |
| 18.31155283, |
| 20.47818807, |
| 22.90118105, |
| 25.61086418, |
| 28.64115884, |
| 32.03, |
| 32.1, |
| 32.13, |
| 32.23, |
| 32.6, |
| 32.61, |
| 32.64, |
| 32.66, |
| 32.7, |
| 32.71, |
| 32.93, |
| 32.97, |
| 33.28, |
| 33.49, |
| 33.5, |
| 44.16, |
| 47.77 |
| ], |
| "short_factor": [ |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0 |
| ], |
| "type": "longrope" |
| }, |
| "rope_theta": 10000.0, |
| "sliding_window": 262144, |
| "speech_lora": { |
| "dp": 0.01, |
| "layer": "((layers.*self_attn\\.(qkv|o)_proj)|(layers.*mlp\\.(gate_up|down)_proj))", |
| "lora_alpha": 640, |
| "r": 320 |
| }, |
| "tie_word_embeddings": true, |
| "transformers_version": "4.57.6", |
| "use_cache": true, |
| "vision_lora": { |
| "dp": 0.0, |
| "layer": "layers.*((self_attn\\.(qkv_proj|o_proj))|(mlp\\.(gate_up|down)_proj))", |
| "lora_alpha": 512, |
| "r": 256 |
| }, |
| "vocab_size": 200064 |
| } |
