Update config.json
Browse files- config.json +0 -81
config.json
CHANGED
|
@@ -79,87 +79,6 @@
|
|
| 79 |
"num_hidden_layers": 28,
|
| 80 |
"num_key_value_heads": 2,
|
| 81 |
"pad_token_id": 151654,
|
| 82 |
-
"quantization_config": {
|
| 83 |
-
"_load_in_4bit": true,
|
| 84 |
-
"_load_in_8bit": false,
|
| 85 |
-
"bnb_4bit_compute_dtype": "bfloat16",
|
| 86 |
-
"bnb_4bit_quant_storage": "uint8",
|
| 87 |
-
"bnb_4bit_quant_type": "nf4",
|
| 88 |
-
"bnb_4bit_use_double_quant": true,
|
| 89 |
-
"llm_int8_enable_fp32_cpu_offload": false,
|
| 90 |
-
"llm_int8_has_fp16_weight": false,
|
| 91 |
-
"llm_int8_skip_modules": [
|
| 92 |
-
"lm_head",
|
| 93 |
-
"multi_modal_projector",
|
| 94 |
-
"merger",
|
| 95 |
-
"modality_projection",
|
| 96 |
-
"model.layers.1.mlp",
|
| 97 |
-
"visual.blocks.29.mlp",
|
| 98 |
-
"visual.blocks.21.mlp",
|
| 99 |
-
"visual.blocks.24.attn",
|
| 100 |
-
"visual.blocks.26.mlp",
|
| 101 |
-
"visual.blocks.23.attn",
|
| 102 |
-
"visual.blocks.18.mlp",
|
| 103 |
-
"visual.blocks.19.mlp",
|
| 104 |
-
"visual.blocks.22.attn",
|
| 105 |
-
"visual.blocks.17.mlp",
|
| 106 |
-
"visual.blocks.25.attn",
|
| 107 |
-
"visual.blocks.24.mlp",
|
| 108 |
-
"visual.blocks.26.attn",
|
| 109 |
-
"visual.blocks.29.attn",
|
| 110 |
-
"visual.blocks.21.attn",
|
| 111 |
-
"visual.blocks.22.mlp",
|
| 112 |
-
"visual.blocks.14.mlp",
|
| 113 |
-
"visual.blocks.23.mlp",
|
| 114 |
-
"visual.blocks.14.attn",
|
| 115 |
-
"visual.merger.mlp",
|
| 116 |
-
"visual.blocks.16.mlp",
|
| 117 |
-
"visual.blocks.20.mlp",
|
| 118 |
-
"visual.blocks.25.mlp",
|
| 119 |
-
"visual.blocks.13.mlp",
|
| 120 |
-
"visual.blocks.18.attn",
|
| 121 |
-
"visual.blocks.20.attn",
|
| 122 |
-
"visual.blocks.17.attn",
|
| 123 |
-
"visual.blocks.15.mlp",
|
| 124 |
-
"visual.blocks.27.attn",
|
| 125 |
-
"visual.blocks.28.attn",
|
| 126 |
-
"visual.blocks.12.mlp",
|
| 127 |
-
"visual.blocks.16.attn",
|
| 128 |
-
"visual.blocks.19.attn",
|
| 129 |
-
"visual.blocks.15.attn",
|
| 130 |
-
"visual.blocks.11.attn",
|
| 131 |
-
"visual.blocks.13.attn",
|
| 132 |
-
"visual.blocks.12.attn",
|
| 133 |
-
"visual.blocks.10.mlp",
|
| 134 |
-
"visual.blocks.11.mlp",
|
| 135 |
-
"visual.blocks.10.attn",
|
| 136 |
-
"visual.blocks.9.mlp",
|
| 137 |
-
"visual.blocks.5.attn",
|
| 138 |
-
"visual.blocks.9.attn",
|
| 139 |
-
"visual.blocks.4.attn",
|
| 140 |
-
"visual.blocks.6.mlp",
|
| 141 |
-
"visual.blocks.7.mlp",
|
| 142 |
-
"visual.blocks.1.attn",
|
| 143 |
-
"visual.blocks.6.attn",
|
| 144 |
-
"visual.blocks.7.attn",
|
| 145 |
-
"visual.blocks.5.mlp",
|
| 146 |
-
"visual.blocks.8.attn",
|
| 147 |
-
"visual.blocks.8.mlp",
|
| 148 |
-
"visual.blocks.4.mlp",
|
| 149 |
-
"visual.blocks.3.attn",
|
| 150 |
-
"visual.blocks.3.mlp",
|
| 151 |
-
"visual.blocks.1.mlp",
|
| 152 |
-
"visual.blocks.2.mlp",
|
| 153 |
-
"visual.blocks.2.attn",
|
| 154 |
-
"visual.blocks.0.attn",
|
| 155 |
-
"visual.blocks.0.mlp",
|
| 156 |
-
"visual.blocks.31.mlp"
|
| 157 |
-
],
|
| 158 |
-
"llm_int8_threshold": 6.0,
|
| 159 |
-
"load_in_4bit": true,
|
| 160 |
-
"load_in_8bit": false,
|
| 161 |
-
"quant_method": "bitsandbytes"
|
| 162 |
-
},
|
| 163 |
"rms_norm_eps": 1e-06,
|
| 164 |
"rope_scaling": {
|
| 165 |
"mrope_section": [
|
|
|
|
| 79 |
"num_hidden_layers": 28,
|
| 80 |
"num_key_value_heads": 2,
|
| 81 |
"pad_token_id": 151654,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 82 |
"rms_norm_eps": 1e-06,
|
| 83 |
"rope_scaling": {
|
| 84 |
"mrope_section": [
|