Updated compression_config to quantization_config
Browse files- config.json +75 -75
config.json
CHANGED
|
@@ -10,80 +10,6 @@
|
|
| 10 |
"AutoModelForCausalLM": "modeling_phimoe.PhiMoEForCausalLM"
|
| 11 |
},
|
| 12 |
"bos_token_id": 1,
|
| 13 |
-
"compression_config": {
|
| 14 |
-
"config_groups": {
|
| 15 |
-
"group_0": {
|
| 16 |
-
"input_activations": {
|
| 17 |
-
"actorder": null,
|
| 18 |
-
"block_structure": null,
|
| 19 |
-
"dynamic": false,
|
| 20 |
-
"group_size": null,
|
| 21 |
-
"num_bits": 8,
|
| 22 |
-
"observer": "minmax",
|
| 23 |
-
"observer_kwargs": {},
|
| 24 |
-
"strategy": "tensor",
|
| 25 |
-
"symmetric": true,
|
| 26 |
-
"type": "float"
|
| 27 |
-
},
|
| 28 |
-
"output_activations": null,
|
| 29 |
-
"targets": [
|
| 30 |
-
"Linear"
|
| 31 |
-
],
|
| 32 |
-
"weights": {
|
| 33 |
-
"actorder": null,
|
| 34 |
-
"block_structure": null,
|
| 35 |
-
"dynamic": false,
|
| 36 |
-
"group_size": null,
|
| 37 |
-
"num_bits": 8,
|
| 38 |
-
"observer": "minmax",
|
| 39 |
-
"observer_kwargs": {},
|
| 40 |
-
"strategy": "tensor",
|
| 41 |
-
"symmetric": true,
|
| 42 |
-
"type": "float"
|
| 43 |
-
}
|
| 44 |
-
}
|
| 45 |
-
},
|
| 46 |
-
"format": "float-quantized",
|
| 47 |
-
"global_compression_ratio": 1.56406776155077,
|
| 48 |
-
"ignore": [
|
| 49 |
-
"model.layers.0.block_sparse_moe.gate",
|
| 50 |
-
"model.layers.1.block_sparse_moe.gate",
|
| 51 |
-
"model.layers.2.block_sparse_moe.gate",
|
| 52 |
-
"model.layers.3.block_sparse_moe.gate",
|
| 53 |
-
"model.layers.4.block_sparse_moe.gate",
|
| 54 |
-
"model.layers.5.block_sparse_moe.gate",
|
| 55 |
-
"model.layers.6.block_sparse_moe.gate",
|
| 56 |
-
"model.layers.7.block_sparse_moe.gate",
|
| 57 |
-
"model.layers.8.block_sparse_moe.gate",
|
| 58 |
-
"model.layers.9.block_sparse_moe.gate",
|
| 59 |
-
"model.layers.10.block_sparse_moe.gate",
|
| 60 |
-
"model.layers.11.block_sparse_moe.gate",
|
| 61 |
-
"model.layers.12.block_sparse_moe.gate",
|
| 62 |
-
"model.layers.13.block_sparse_moe.gate",
|
| 63 |
-
"model.layers.14.block_sparse_moe.gate",
|
| 64 |
-
"model.layers.15.block_sparse_moe.gate",
|
| 65 |
-
"model.layers.16.block_sparse_moe.gate",
|
| 66 |
-
"model.layers.17.block_sparse_moe.gate",
|
| 67 |
-
"model.layers.18.block_sparse_moe.gate",
|
| 68 |
-
"model.layers.19.block_sparse_moe.gate",
|
| 69 |
-
"model.layers.20.block_sparse_moe.gate",
|
| 70 |
-
"model.layers.21.block_sparse_moe.gate",
|
| 71 |
-
"model.layers.22.block_sparse_moe.gate",
|
| 72 |
-
"model.layers.23.block_sparse_moe.gate",
|
| 73 |
-
"model.layers.24.block_sparse_moe.gate",
|
| 74 |
-
"model.layers.25.block_sparse_moe.gate",
|
| 75 |
-
"model.layers.26.block_sparse_moe.gate",
|
| 76 |
-
"model.layers.27.block_sparse_moe.gate",
|
| 77 |
-
"model.layers.28.block_sparse_moe.gate",
|
| 78 |
-
"model.layers.29.block_sparse_moe.gate",
|
| 79 |
-
"model.layers.30.block_sparse_moe.gate",
|
| 80 |
-
"model.layers.31.block_sparse_moe.gate",
|
| 81 |
-
"lm_head"
|
| 82 |
-
],
|
| 83 |
-
"kv_cache_scheme": null,
|
| 84 |
-
"quant_method": "compressed-tensors",
|
| 85 |
-
"quantization_status": "compressed"
|
| 86 |
-
},
|
| 87 |
"eos_token_id": 32000,
|
| 88 |
"hidden_act": "silu",
|
| 89 |
"hidden_dropout": 0.0,
|
|
@@ -248,5 +174,79 @@
|
|
| 248 |
"torch_dtype": "bfloat16",
|
| 249 |
"transformers_version": "4.44.2",
|
| 250 |
"use_cache": true,
|
| 251 |
-
"vocab_size": 32064
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 252 |
}
|
|
|
|
| 10 |
"AutoModelForCausalLM": "modeling_phimoe.PhiMoEForCausalLM"
|
| 11 |
},
|
| 12 |
"bos_token_id": 1,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 13 |
"eos_token_id": 32000,
|
| 14 |
"hidden_act": "silu",
|
| 15 |
"hidden_dropout": 0.0,
|
|
|
|
| 174 |
"torch_dtype": "bfloat16",
|
| 175 |
"transformers_version": "4.44.2",
|
| 176 |
"use_cache": true,
|
| 177 |
+
"vocab_size": 32064,
|
| 178 |
+
"quantization_config": {
|
| 179 |
+
"config_groups": {
|
| 180 |
+
"group_0": {
|
| 181 |
+
"input_activations": {
|
| 182 |
+
"actorder": null,
|
| 183 |
+
"block_structure": null,
|
| 184 |
+
"dynamic": false,
|
| 185 |
+
"group_size": null,
|
| 186 |
+
"num_bits": 8,
|
| 187 |
+
"observer": "minmax",
|
| 188 |
+
"observer_kwargs": {},
|
| 189 |
+
"strategy": "tensor",
|
| 190 |
+
"symmetric": true,
|
| 191 |
+
"type": "float"
|
| 192 |
+
},
|
| 193 |
+
"output_activations": null,
|
| 194 |
+
"targets": [
|
| 195 |
+
"Linear"
|
| 196 |
+
],
|
| 197 |
+
"weights": {
|
| 198 |
+
"actorder": null,
|
| 199 |
+
"block_structure": null,
|
| 200 |
+
"dynamic": false,
|
| 201 |
+
"group_size": null,
|
| 202 |
+
"num_bits": 8,
|
| 203 |
+
"observer": "minmax",
|
| 204 |
+
"observer_kwargs": {},
|
| 205 |
+
"strategy": "tensor",
|
| 206 |
+
"symmetric": true,
|
| 207 |
+
"type": "float"
|
| 208 |
+
}
|
| 209 |
+
}
|
| 210 |
+
},
|
| 211 |
+
"format": "float-quantized",
|
| 212 |
+
"global_compression_ratio": 1.56406776155077,
|
| 213 |
+
"ignore": [
|
| 214 |
+
"model.layers.0.block_sparse_moe.gate",
|
| 215 |
+
"model.layers.1.block_sparse_moe.gate",
|
| 216 |
+
"model.layers.2.block_sparse_moe.gate",
|
| 217 |
+
"model.layers.3.block_sparse_moe.gate",
|
| 218 |
+
"model.layers.4.block_sparse_moe.gate",
|
| 219 |
+
"model.layers.5.block_sparse_moe.gate",
|
| 220 |
+
"model.layers.6.block_sparse_moe.gate",
|
| 221 |
+
"model.layers.7.block_sparse_moe.gate",
|
| 222 |
+
"model.layers.8.block_sparse_moe.gate",
|
| 223 |
+
"model.layers.9.block_sparse_moe.gate",
|
| 224 |
+
"model.layers.10.block_sparse_moe.gate",
|
| 225 |
+
"model.layers.11.block_sparse_moe.gate",
|
| 226 |
+
"model.layers.12.block_sparse_moe.gate",
|
| 227 |
+
"model.layers.13.block_sparse_moe.gate",
|
| 228 |
+
"model.layers.14.block_sparse_moe.gate",
|
| 229 |
+
"model.layers.15.block_sparse_moe.gate",
|
| 230 |
+
"model.layers.16.block_sparse_moe.gate",
|
| 231 |
+
"model.layers.17.block_sparse_moe.gate",
|
| 232 |
+
"model.layers.18.block_sparse_moe.gate",
|
| 233 |
+
"model.layers.19.block_sparse_moe.gate",
|
| 234 |
+
"model.layers.20.block_sparse_moe.gate",
|
| 235 |
+
"model.layers.21.block_sparse_moe.gate",
|
| 236 |
+
"model.layers.22.block_sparse_moe.gate",
|
| 237 |
+
"model.layers.23.block_sparse_moe.gate",
|
| 238 |
+
"model.layers.24.block_sparse_moe.gate",
|
| 239 |
+
"model.layers.25.block_sparse_moe.gate",
|
| 240 |
+
"model.layers.26.block_sparse_moe.gate",
|
| 241 |
+
"model.layers.27.block_sparse_moe.gate",
|
| 242 |
+
"model.layers.28.block_sparse_moe.gate",
|
| 243 |
+
"model.layers.29.block_sparse_moe.gate",
|
| 244 |
+
"model.layers.30.block_sparse_moe.gate",
|
| 245 |
+
"model.layers.31.block_sparse_moe.gate",
|
| 246 |
+
"lm_head"
|
| 247 |
+
],
|
| 248 |
+
"kv_cache_scheme": null,
|
| 249 |
+
"quant_method": "compressed-tensors",
|
| 250 |
+
"quantization_status": "compressed"
|
| 251 |
+
}
|
| 252 |
}
|