update config
Browse files- config.json +13 -12
config.json
CHANGED
|
@@ -18,7 +18,15 @@
|
|
| 18 |
"num_hidden_layers": 80,
|
| 19 |
"num_key_value_heads": 8,
|
| 20 |
"pretraining_tp": 1,
|
| 21 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
"model.layers.0.mlp.down_proj": {
|
| 23 |
"bias": null,
|
| 24 |
"enable_norm": true,
|
|
@@ -13458,14 +13466,7 @@
|
|
| 13458 |
-1,
|
| 13459 |
16
|
| 13460 |
]
|
| 13461 |
-
}
|
| 13462 |
-
|
| 13463 |
-
|
| 13464 |
-
|
| 13465 |
-
"rope_theta": 500000.0,
|
| 13466 |
-
"tie_word_embeddings": false,
|
| 13467 |
-
"torch_dtype": "bfloat16",
|
| 13468 |
-
"transformers_version": "4.45.1",
|
| 13469 |
-
"use_cache": true,
|
| 13470 |
-
"vocab_size": 128256
|
| 13471 |
-
}
|
|
|
|
| 18 |
"num_hidden_layers": 80,
|
| 19 |
"num_key_value_heads": 8,
|
| 20 |
"pretraining_tp": 1,
|
| 21 |
+
"rms_norm_eps": 1e-05,
|
| 22 |
+
"rope_scaling": null,
|
| 23 |
+
"rope_theta": 500000.0,
|
| 24 |
+
"tie_word_embeddings": false,
|
| 25 |
+
"torch_dtype": "bfloat16",
|
| 26 |
+
"transformers_version": "4.45.1",
|
| 27 |
+
"use_cache": true,
|
| 28 |
+
"vocab_size": 128256,
|
| 29 |
+
"quantization_config": {
|
| 30 |
"model.layers.0.mlp.down_proj": {
|
| 31 |
"bias": null,
|
| 32 |
"enable_norm": true,
|
|
|
|
| 13466 |
-1,
|
| 13467 |
16
|
| 13468 |
]
|
| 13469 |
+
},
|
| 13470 |
+
"quant_method": "vptq"
|
| 13471 |
+
}
|
| 13472 |
+
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|