Zero-Shot Classification
MLX
Safetensors
English
multilingual
qwen3_5
jang
quantized
apple-silicon
osaurus
decision-model
classification
multimodal
vision
video
8-bit precision
Instructions to use OsaurusAI/d3-JANGH2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use OsaurusAI/d3-JANGH2 with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] hf download OsaurusAI/d3-JANGH2 --local-dir d3-JANGH2
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
Download config.json from OsaurusAI/d3-JANGH2: direct link, hf CLI and curl.
- Browser
- Download file 150 kB
-
https://huggingface.co/OsaurusAI/d3-JANGH2/resolve/main/config.json
- Command line
-
hf download hf://OsaurusAI/d3-JANGH2/config.json
-
curl -L -o config.json https://huggingface.co/OsaurusAI/d3-JANGH2/resolve/main/config.json
150 kB
| { | |
| "architectures": [ | |
| "Qwen3_5Model" | |
| ], | |
| "dtype": "bfloat16", | |
| "image_token_id": 248056, | |
| "language_model_only": false, | |
| "model_type": "qwen3_5", | |
| "text_config": { | |
| "attention_bias": false, | |
| "attention_dropout": 0.0, | |
| "attn_output_gate": true, | |
| "bos_token_id": 248044, | |
| "dtype": "bfloat16", | |
| "eos_token_id": 248044, | |
| "full_attention_interval": 4, | |
| "head_dim": 256, | |
| "hidden_act": "silu", | |
| "hidden_size": 5120, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 17408, | |
| "layer_types": [ | |
| "linear_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "full_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "full_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "full_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "full_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "full_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "full_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "full_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "full_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "full_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "full_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "full_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "full_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "full_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "full_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "full_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "linear_attention", | |
| "full_attention" | |
| ], | |
| "linear_conv_kernel_dim": 4, | |
| "linear_key_head_dim": 128, | |
| "linear_num_key_heads": 16, | |
| "linear_num_value_heads": 48, | |
| "linear_value_head_dim": 128, | |
| "mamba_ssm_dtype": "float32", | |
| "max_position_embeddings": 262144, | |
| "model_type": "qwen3_5_text", | |
| "mtp_num_hidden_layers": 0, | |
| "mtp_use_dedicated_embeddings": false, | |
| "num_attention_heads": 24, | |
| "num_hidden_layers": 64, | |
| "num_key_value_heads": 4, | |
| "output_gate_type": "swish", | |
| "pad_token_id": null, | |
| "partial_rotary_factor": 0.25, | |
| "rms_norm_eps": 1e-06, | |
| "rope_parameters": { | |
| "mrope_interleaved": true, | |
| "mrope_section": [ | |
| 11, | |
| 11, | |
| 10 | |
| ], | |
| "partial_rotary_factor": 0.25, | |
| "rope_theta": 10000000, | |
| "rope_type": "default" | |
| }, | |
| "tie_word_embeddings": false, | |
| "use_cache": true, | |
| "vocab_size": 248320, | |
| "norm_convention": "qwen3_5_language_mlx_plus_one" | |
| }, | |
| "tie_word_embeddings": false, | |
| "transformers_version": "5.17.0", | |
| "video_token_id": 248057, | |
| "vision_config": { | |
| "deepstack_visual_indexes": [], | |
| "depth": 27, | |
| "hidden_act": "gelu_pytorch_tanh", | |
| "hidden_size": 1152, | |
| "in_channels": 3, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 4304, | |
| "model_type": "qwen3_5_vision", | |
| "num_heads": 16, | |
| "num_position_embeddings": 2304, | |
| "out_hidden_size": 5120, | |
| "patch_size": 16, | |
| "rope_parameters": { | |
| "rope_theta": 10000.0, | |
| "rope_type": "axial" | |
| }, | |
| "spatial_merge_size": 2, | |
| "temporal_patch_size": 2 | |
| }, | |
| "vision_end_token_id": 248054, | |
| "vision_start_token_id": 248053, | |
| "norm_convention": "qwen3_5_language_mlx_plus_one", | |
| "quantization": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine", | |
| "language_model.model.layers.0.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.0.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.0.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.0.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.0.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.0.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.0.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.0.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.1.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.1.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.1.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.1.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.1.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.1.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.1.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.1.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.2.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.2.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.2.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.2.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.2.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.2.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.2.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.2.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.3.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.3.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.3.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.3.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.3.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.3.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.3.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.4.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.4.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.4.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.4.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.4.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.4.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.4.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.4.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.5.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.5.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.5.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.5.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.5.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.5.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.5.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.5.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.6.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.6.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.6.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.6.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.6.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.6.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.6.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.6.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.7.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.7.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.7.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.7.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.7.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.7.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.7.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.8.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.8.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.8.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.8.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.8.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.8.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.8.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.8.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.9.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.9.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.9.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.9.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.9.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.9.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.9.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.9.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.10.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.10.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.10.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.10.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.10.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.10.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.10.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.10.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.11.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.11.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.11.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.11.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.11.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.11.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.11.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.12.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.12.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.12.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.12.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.12.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.12.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.12.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.12.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.13.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.13.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.13.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.13.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.13.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.13.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.13.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.13.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.14.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.14.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.14.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.14.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.14.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.14.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.14.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.14.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.15.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.15.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.15.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.15.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.15.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.15.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.15.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.16.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.16.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.16.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.16.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.16.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.16.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.16.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.16.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.17.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.17.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.17.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.17.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.17.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.17.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.17.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.17.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.18.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.18.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.18.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.18.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.18.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.18.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.18.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.18.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.19.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.19.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.19.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.19.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.19.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.19.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.19.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.20.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.20.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.20.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.20.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.20.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.20.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.20.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.20.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.21.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.21.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.21.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.21.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.21.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.21.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.21.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.21.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.22.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.22.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.22.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.22.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.22.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.22.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.22.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.22.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.23.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.23.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.23.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.23.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.23.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.23.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.23.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.24.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.24.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.24.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.24.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.24.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.24.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.24.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.24.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.25.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.25.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.25.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.25.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.25.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.25.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.25.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.25.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.26.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.26.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.26.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.26.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.26.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.26.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.26.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.26.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.27.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.27.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.27.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.27.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.27.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.27.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.27.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.28.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.28.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.28.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.28.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.28.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.28.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.28.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.28.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.29.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.29.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.29.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.29.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.29.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.29.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.29.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.29.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.30.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.30.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.30.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.30.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.30.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.30.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.30.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.30.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.31.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.31.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.31.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.31.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.31.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.31.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.31.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.32.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.32.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.32.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.32.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.32.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.32.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.32.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.32.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.33.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.33.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.33.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.33.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.33.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.33.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.33.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.33.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.34.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.34.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.34.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.34.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.34.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.34.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.34.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.34.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.35.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.35.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.35.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.35.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.35.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.35.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.35.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.36.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.36.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.36.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.36.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.36.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.36.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.36.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.36.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.37.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.37.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.37.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.37.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.37.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.37.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.37.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.37.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.38.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.38.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.38.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.38.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.38.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.38.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.38.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.38.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.39.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.39.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.39.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.39.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.39.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.39.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.39.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.40.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.40.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.40.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.40.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.40.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.40.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.40.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.40.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.41.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.41.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.41.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.41.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.41.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.41.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.41.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.41.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.42.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.42.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.42.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.42.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.42.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.42.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.42.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.42.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.43.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.43.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.43.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.43.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.43.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.43.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.43.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.44.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.44.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.44.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.44.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.44.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.44.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.44.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.44.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.45.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.45.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.45.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.45.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.45.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.45.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.45.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.45.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.46.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.46.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.46.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.46.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.46.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.46.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.46.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.46.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.47.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.47.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.47.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.47.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.47.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.47.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.47.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.48.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.48.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.48.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.48.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.48.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.48.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.48.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.48.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.49.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.49.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.49.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.49.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.49.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.49.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.49.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.49.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.50.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.50.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.50.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.50.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.50.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.50.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.50.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.50.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.51.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.51.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.51.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.51.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.51.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.51.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.51.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.52.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.52.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.52.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.52.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.52.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.52.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.52.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.52.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.53.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.53.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.53.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.53.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.53.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.53.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.53.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.53.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.54.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.54.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.54.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.54.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.54.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.54.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.54.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.54.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.55.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.55.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.55.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.55.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.55.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.55.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.55.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.56.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.56.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.56.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.56.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.56.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.56.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.56.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.56.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.57.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.57.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.57.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.57.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.57.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.57.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.57.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.57.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.58.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.58.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.58.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.58.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.58.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.58.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.58.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.58.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.59.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.59.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.59.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.59.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.59.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.59.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.59.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.60.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.60.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.60.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.60.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.60.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.60.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.60.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.60.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.61.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.61.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.61.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.61.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.61.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.61.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.61.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.61.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.62.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.62.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.62.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.62.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.62.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.62.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.62.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.62.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.63.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.63.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.63.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.63.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.63.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.63.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.63.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.embed_tokens": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.0.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.0.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.0.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.1.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.1.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.1.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.10.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.10.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.10.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.11.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.11.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.11.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.12.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.12.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.12.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.13.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.13.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.13.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.14.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.14.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.14.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.15.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.15.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.15.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.16.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.16.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.16.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.17.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.17.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.17.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.18.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.18.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.18.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.19.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.19.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.19.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.2.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.2.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.2.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.20.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.20.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.20.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.21.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.21.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.21.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.22.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.22.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.22.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.23.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.23.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.23.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.24.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.24.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.24.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.25.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.25.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.25.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.26.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.26.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.26.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.3.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.3.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.3.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.4.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.4.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.4.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.5.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.5.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.5.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.6.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.6.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.6.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.7.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.7.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.7.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.8.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.8.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.8.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.9.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.9.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.9.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.merger.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.merger.linear_fc2": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.pos_embed": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| } | |
| }, | |
| "quantization_config": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine", | |
| "language_model.model.layers.0.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.0.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.0.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.0.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.0.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.0.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.0.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.0.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.1.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.1.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.1.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.1.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.1.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.1.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.1.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.1.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.2.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.2.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.2.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.2.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.2.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.2.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.2.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.2.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.3.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.3.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.3.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.3.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.3.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.3.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.3.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.4.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.4.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.4.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.4.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.4.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.4.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.4.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.4.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.5.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.5.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.5.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.5.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.5.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.5.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.5.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.5.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.6.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.6.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.6.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.6.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.6.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.6.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.6.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.6.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.7.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.7.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.7.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.7.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.7.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.7.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.7.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.8.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.8.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.8.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.8.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.8.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.8.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.8.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.8.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.9.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.9.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.9.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.9.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.9.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.9.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.9.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.9.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.10.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.10.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.10.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.10.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.10.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.10.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.10.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.10.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.11.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.11.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.11.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.11.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.11.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.11.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.11.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.12.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.12.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.12.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.12.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.12.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.12.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.12.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.12.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.13.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.13.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.13.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.13.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.13.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.13.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.13.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.13.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.14.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.14.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.14.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.14.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.14.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.14.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.14.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.14.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.15.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.15.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.15.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.15.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.15.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.15.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.15.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.16.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.16.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.16.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.16.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.16.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.16.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.16.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.16.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.17.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.17.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.17.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.17.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.17.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.17.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.17.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.17.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.18.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.18.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.18.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.18.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.18.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.18.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.18.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.18.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.19.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.19.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.19.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.19.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.19.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.19.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.19.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.20.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.20.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.20.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.20.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.20.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.20.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.20.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.20.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.21.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.21.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.21.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.21.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.21.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.21.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.21.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.21.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.22.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.22.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.22.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.22.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.22.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.22.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.22.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.22.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.23.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.23.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.23.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.23.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.23.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.23.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.23.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.24.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.24.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.24.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.24.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.24.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.24.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.24.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.24.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.25.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.25.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.25.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.25.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.25.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.25.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.25.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.25.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.26.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.26.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.26.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.26.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.26.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.26.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.26.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.26.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.27.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.27.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.27.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.27.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.27.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.27.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.27.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.28.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.28.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.28.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.28.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.28.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.28.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.28.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.28.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.29.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.29.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.29.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.29.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.29.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.29.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.29.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.29.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.30.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.30.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.30.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.30.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.30.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.30.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.30.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.30.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.31.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.31.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.31.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.31.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.31.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.31.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.31.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.32.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.32.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.32.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.32.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.32.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.32.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.32.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.32.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.33.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.33.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.33.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.33.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.33.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.33.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.33.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.33.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.34.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.34.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.34.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.34.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.34.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.34.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.34.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.34.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.35.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.35.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.35.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.35.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.35.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.35.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.35.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.36.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.36.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.36.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.36.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.36.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.36.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.36.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.36.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.37.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.37.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.37.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.37.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.37.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.37.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.37.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.37.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.38.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.38.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.38.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.38.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.38.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.38.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.38.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.38.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.39.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.39.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.39.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.39.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.39.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.39.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.39.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.40.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.40.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.40.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.40.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.40.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.40.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.40.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.40.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.41.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.41.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.41.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.41.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.41.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.41.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.41.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.41.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.42.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.42.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.42.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.42.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.42.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.42.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.42.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.42.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.43.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.43.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.43.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.43.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.43.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.43.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.43.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.44.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.44.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.44.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.44.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.44.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.44.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.44.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.44.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.45.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.45.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.45.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.45.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.45.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.45.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.45.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.45.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.46.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.46.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.46.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.46.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.46.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.46.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.46.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.46.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.47.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.47.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.47.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.47.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.47.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.47.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.47.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.48.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.48.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.48.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.48.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.48.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.48.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.48.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.48.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.49.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.49.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.49.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.49.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.49.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.49.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.49.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.49.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.50.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.50.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.50.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.50.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.50.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.50.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.50.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.50.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.51.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.51.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.51.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.51.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.51.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.51.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.51.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.52.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.52.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.52.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.52.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.52.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.52.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.52.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.52.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.53.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.53.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.53.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.53.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.53.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.53.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.53.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.53.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.54.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.54.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.54.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.54.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.54.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.54.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.54.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.54.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.55.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.55.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.55.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.55.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.55.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.55.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.55.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.56.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.56.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.56.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.56.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.56.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.56.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.56.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.56.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.57.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.57.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.57.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.57.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.57.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.57.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.57.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.57.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.58.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.58.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.58.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.58.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.58.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.58.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.58.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.58.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.59.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.59.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.59.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.59.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.59.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.59.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.59.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.60.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.60.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.60.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.60.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.60.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.60.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.60.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.60.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.61.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.61.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.61.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.61.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.61.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.61.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.61.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.61.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.62.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.62.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.62.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 2, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.62.linear_attn.in_proj_qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.62.linear_attn.in_proj_z": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.62.linear_attn.in_proj_a": false, | |
| "language_model.model.layers.62.linear_attn.in_proj_b": false, | |
| "language_model.model.layers.62.linear_attn.out_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.63.mlp.gate_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.63.mlp.up_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.63.mlp.down_proj": { | |
| "mode": "jangtq2", | |
| "bits": 4, | |
| "rotation": "hadamard32" | |
| }, | |
| "language_model.model.layers.63.self_attn.q_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.63.self_attn.k_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.63.self_attn.v_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.layers.63.self_attn.o_proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "language_model.model.embed_tokens": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.0.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.0.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.0.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.1.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.1.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.1.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.10.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.10.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.10.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.11.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.11.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.11.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.12.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.12.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.12.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.13.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.13.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.13.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.14.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.14.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.14.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.15.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.15.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.15.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.16.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.16.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.16.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.17.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.17.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.17.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.18.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.18.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.18.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.19.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.19.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.19.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.2.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.2.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.2.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.20.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.20.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.20.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.21.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.21.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.21.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.22.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.22.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.22.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.23.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.23.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.23.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.24.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.24.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.24.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.25.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.25.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.25.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.26.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.26.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.26.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.3.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.3.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.3.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.4.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.4.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.4.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.5.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.5.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.5.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.6.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.6.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.6.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.7.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.7.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.7.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.8.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.8.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.8.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.9.attn.proj": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.9.attn.qkv": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.blocks.9.mlp.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.merger.linear_fc1": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.merger.linear_fc2": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| }, | |
| "vision_tower.pos_embed": { | |
| "group_size": 64, | |
| "bits": 8, | |
| "mode": "affine" | |
| } | |
| }, | |
| "jangtq": { | |
| "version": 2, | |
| "packing": "lsb-bitstream", | |
| "scale_dtype": "float16", | |
| "rotation": "hadamard32", | |
| "codebook_family": "odd-cubic", | |
| "codebooks": { | |
| "2": { | |
| "alpha": 0.8929999999999999, | |
| "beta": 0.05065, | |
| "levels": [ | |
| -1.5104438066482544, | |
| -0.4528312385082245, | |
| 0.4528312385082245, | |
| 1.5104438066482544 | |
| ] | |
| }, | |
| "3": { | |
| "alpha": 0.47124999999999995, | |
| "beta": 0.011599999999999997, | |
| "levels": [ | |
| -2.1467249393463135, | |
| -1.359375, | |
| -0.746025025844574, | |
| -0.23707500100135803, | |
| 0.23707500100135803, | |
| 0.746025025844574, | |
| 1.359375, | |
| 2.1467249393463135 | |
| ] | |
| }, | |
| "4": { | |
| "alpha": 0.2405, | |
| "beta": 0.002100000000000002, | |
| "levels": [ | |
| -2.689687490463257, | |
| -2.1399624347686768, | |
| -1.6721374988555908, | |
| -1.2736124992370605, | |
| -0.9317874908447266, | |
| -0.6340625286102295, | |
| -0.36783748865127563, | |
| -0.12051250040531158, | |
| 0.12051250040531158, | |
| 0.36783748865127563, | |
| 0.6340625286102295, | |
| 0.9317874908447266, | |
| 1.2736124992370605, | |
| 1.6721374988555908, | |
| 2.1399624347686768, | |
| 2.689687490463257 | |
| ] | |
| } | |
| } | |
| }, | |
| "awq": { | |
| "folded_into": "producing RMSNorm (gate/up input, attention / GDN input) and up rows (down input)", | |
| "formula": "(stored+1)/s - 1", | |
| "per_layer_alpha": { | |
| "0": { | |
| "ffn_in": 0.4, | |
| "down_in": 0.1, | |
| "gdn_in": 0.3 | |
| }, | |
| "1": { | |
| "ffn_in": 0.5, | |
| "down_in": 0.5, | |
| "gdn_in": 0.4 | |
| }, | |
| "2": { | |
| "ffn_in": 0.5, | |
| "down_in": 0.0, | |
| "gdn_in": 0.4 | |
| }, | |
| "3": { | |
| "ffn_in": 0.5, | |
| "down_in": 0.4, | |
| "attn_in": 0.4 | |
| }, | |
| "4": { | |
| "ffn_in": 0.5, | |
| "down_in": 0.4, | |
| "gdn_in": 0.4 | |
| }, | |
| "5": { | |
| "ffn_in": 0.5, | |
| "down_in": 0.4, | |
| "gdn_in": 0.4 | |
| }, | |
| "6": { | |
| "ffn_in": 0.3, | |
| "down_in": 0.2, | |
| "gdn_in": 0.4 | |
| }, | |
| "7": { | |
| "ffn_in": 0.5, | |
| "down_in": 0.4, | |
| "attn_in": 0.4 | |
| }, | |
| "8": { | |
| "ffn_in": 0.5, | |
| "down_in": 0.5, | |
| "gdn_in": 0.4 | |
| }, | |
| "9": { | |
| "ffn_in": 0.5, | |
| "down_in": 0.4, | |
| "gdn_in": 0.4 | |
| }, | |
| "10": { | |
| "ffn_in": 0.4, | |
| "down_in": 0.4, | |
| "gdn_in": 0.4 | |
| }, | |
| "11": { | |
| "ffn_in": 0.4, | |
| "down_in": 0.4, | |
| "attn_in": 0.4 | |
| }, | |
| "12": { | |
| "ffn_in": 0.5, | |
| "down_in": 0.4, | |
| "gdn_in": 0.4 | |
| }, | |
| "13": { | |
| "ffn_in": 0.5, | |
| "down_in": 0.4, | |
| "gdn_in": 0.4 | |
| }, | |
| "14": { | |
| "ffn_in": 0.5, | |
| "down_in": 0.4, | |
| "gdn_in": 0.4 | |
| }, | |
| "15": { | |
| "ffn_in": 0.4, | |
| "down_in": 0.4, | |
| "attn_in": 0.4 | |
| }, | |
| "16": { | |
| "ffn_in": 0.4, | |
| "down_in": 0.5, | |
| "gdn_in": 0.4 | |
| }, | |
| "17": { | |
| "ffn_in": 0.4, | |
| "down_in": 0.5, | |
| "gdn_in": 0.4 | |
| }, | |
| "18": { | |
| "ffn_in": 0.4, | |
| "down_in": 0.3, | |
| "gdn_in": 0.4 | |
| }, | |
| "19": { | |
| "ffn_in": 0.4, | |
| "down_in": 0.5, | |
| "attn_in": 0.3 | |
| }, | |
| "20": { | |
| "ffn_in": 0.4, | |
| "down_in": 0.5, | |
| "gdn_in": 0.4 | |
| }, | |
| "21": { | |
| "ffn_in": 0.4, | |
| "down_in": 0.5, | |
| "gdn_in": 0.4 | |
| }, | |
| "22": { | |
| "ffn_in": 0.0, | |
| "down_in": 0.4, | |
| "gdn_in": 0.4 | |
| }, | |
| "23": { | |
| "ffn_in": 0.3, | |
| "down_in": 0.5, | |
| "attn_in": 0.3 | |
| }, | |
| "24": { | |
| "ffn_in": 0.4, | |
| "down_in": 0.5, | |
| "gdn_in": 0.4 | |
| }, | |
| "25": { | |
| "ffn_in": 0.4, | |
| "down_in": 0.4, | |
| "gdn_in": 0.4 | |
| }, | |
| "26": { | |
| "ffn_in": 0.3, | |
| "down_in": 0.1, | |
| "gdn_in": 0.4 | |
| }, | |
| "27": { | |
| "ffn_in": 0.3, | |
| "down_in": 0.4, | |
| "attn_in": 0.4 | |
| }, | |
| "28": { | |
| "ffn_in": 0.3, | |
| "down_in": 0.5, | |
| "gdn_in": 0.4 | |
| }, | |
| "29": { | |
| "ffn_in": 0.3, | |
| "down_in": 0.5, | |
| "gdn_in": 0.4 | |
| }, | |
| "30": { | |
| "ffn_in": 0.3, | |
| "down_in": 0.0, | |
| "gdn_in": 0.4 | |
| }, | |
| "31": { | |
| "ffn_in": 0.3, | |
| "down_in": 0.4, | |
| "attn_in": 0.4 | |
| }, | |
| "32": { | |
| "ffn_in": 0.3, | |
| "down_in": 0.5, | |
| "gdn_in": 0.4 | |
| }, | |
| "33": { | |
| "ffn_in": 0.3, | |
| "down_in": 0.5, | |
| "gdn_in": 0.3 | |
| }, | |
| "34": { | |
| "ffn_in": 0.3, | |
| "down_in": 0.4, | |
| "gdn_in": 0.4 | |
| }, | |
| "35": { | |
| "ffn_in": 0.2, | |
| "down_in": 0.0, | |
| "attn_in": 0.3 | |
| }, | |
| "36": { | |
| "ffn_in": 0.2, | |
| "down_in": 0.5, | |
| "gdn_in": 0.3 | |
| }, | |
| "37": { | |
| "ffn_in": 0.3, | |
| "down_in": 0.5, | |
| "gdn_in": 0.3 | |
| }, | |
| "38": { | |
| "ffn_in": 0.2, | |
| "down_in": 0.5, | |
| "gdn_in": 0.3 | |
| }, | |
| "39": { | |
| "ffn_in": 0.2, | |
| "down_in": 0.4, | |
| "attn_in": 0.3 | |
| }, | |
| "40": { | |
| "ffn_in": 0.2, | |
| "down_in": 0.5, | |
| "gdn_in": 0.4 | |
| }, | |
| "41": { | |
| "ffn_in": 0.3, | |
| "down_in": 0.4, | |
| "gdn_in": 0.4 | |
| }, | |
| "42": { | |
| "ffn_in": 0.3, | |
| "down_in": 0.5, | |
| "gdn_in": 0.4 | |
| }, | |
| "43": { | |
| "ffn_in": 0.3, | |
| "down_in": 0.3, | |
| "attn_in": 0.4 | |
| }, | |
| "44": { | |
| "ffn_in": 0.3, | |
| "down_in": 0.4, | |
| "gdn_in": 0.4 | |
| }, | |
| "45": { | |
| "ffn_in": 0.3, | |
| "down_in": 0.5, | |
| "gdn_in": 0.3 | |
| }, | |
| "46": { | |
| "ffn_in": 0.3, | |
| "down_in": 0.0, | |
| "gdn_in": 0.4 | |
| }, | |
| "47": { | |
| "ffn_in": 0.3, | |
| "down_in": 0.2, | |
| "attn_in": 0.4 | |
| }, | |
| "48": { | |
| "ffn_in": 0.3, | |
| "down_in": 0.4, | |
| "gdn_in": 0.3 | |
| }, | |
| "49": { | |
| "ffn_in": 0.3, | |
| "down_in": 0.3, | |
| "gdn_in": 0.3 | |
| }, | |
| "50": { | |
| "ffn_in": 0.2, | |
| "down_in": 0.4, | |
| "gdn_in": 0.4 | |
| }, | |
| "51": { | |
| "ffn_in": 0.2, | |
| "down_in": 0.0, | |
| "attn_in": 0.3 | |
| }, | |
| "52": { | |
| "ffn_in": 0.2, | |
| "down_in": 0.1, | |
| "gdn_in": 0.3 | |
| }, | |
| "53": { | |
| "ffn_in": 0.2, | |
| "down_in": 0.4, | |
| "gdn_in": 0.3 | |
| }, | |
| "54": { | |
| "ffn_in": 0.2, | |
| "down_in": 0.0, | |
| "gdn_in": 0.3 | |
| }, | |
| "55": { | |
| "ffn_in": 0.2, | |
| "down_in": 0.0, | |
| "attn_in": 0.2 | |
| }, | |
| "56": { | |
| "ffn_in": 0.2, | |
| "down_in": 0.3, | |
| "gdn_in": 0.3 | |
| }, | |
| "57": { | |
| "ffn_in": 0.2, | |
| "down_in": 0.5, | |
| "gdn_in": 0.3 | |
| }, | |
| "58": { | |
| "ffn_in": 0.2, | |
| "down_in": 0.0, | |
| "gdn_in": 0.3 | |
| }, | |
| "59": { | |
| "ffn_in": 0.2, | |
| "down_in": 0.1, | |
| "attn_in": 0.2 | |
| }, | |
| "60": { | |
| "ffn_in": 0.2, | |
| "down_in": 0.4, | |
| "gdn_in": 0.3 | |
| }, | |
| "61": { | |
| "ffn_in": 0.2, | |
| "down_in": 0.2, | |
| "gdn_in": 0.3 | |
| }, | |
| "62": { | |
| "ffn_in": 0.2, | |
| "down_in": 0.4, | |
| "gdn_in": 0.3 | |
| }, | |
| "63": { | |
| "ffn_in": 0.2, | |
| "down_in": 0.2, | |
| "attn_in": 0.2 | |
| } | |
| } | |
| } | |
| } |