Update for Transformers AWQ support
Browse files- config.json +34 -27
config.json
CHANGED
|
@@ -1,29 +1,36 @@
|
|
| 1 |
{
|
| 2 |
-
|
| 3 |
-
|
| 4 |
-
|
| 5 |
-
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
|
| 13 |
-
|
| 14 |
-
|
| 15 |
-
|
| 16 |
-
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
|
| 20 |
-
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
|
| 24 |
-
|
| 25 |
-
|
| 26 |
-
|
| 27 |
-
|
| 28 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 29 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"_name_or_path": "/mnt/user/qumu/download_models/codellama/CodeLlama-34b-Python-hf",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"LlamaForCausalLM"
|
| 5 |
+
],
|
| 6 |
+
"bos_token_id": 1,
|
| 7 |
+
"eos_token": "</s>",
|
| 8 |
+
"eos_token_id": 2,
|
| 9 |
+
"hidden_act": "silu",
|
| 10 |
+
"hidden_size": 8192,
|
| 11 |
+
"initializer_range": 0.02,
|
| 12 |
+
"intermediate_size": 22016,
|
| 13 |
+
"max_position_embeddings": 16384,
|
| 14 |
+
"model_type": "llama",
|
| 15 |
+
"num_attention_heads": 64,
|
| 16 |
+
"num_hidden_layers": 48,
|
| 17 |
+
"num_key_value_heads": 8,
|
| 18 |
+
"pad_token": "<unk>",
|
| 19 |
+
"pad_token_id": 0,
|
| 20 |
+
"pretraining_tp": 1,
|
| 21 |
+
"rms_norm_eps": 1e-05,
|
| 22 |
+
"rope_scaling": null,
|
| 23 |
+
"rope_theta": 1000000,
|
| 24 |
+
"tie_word_embeddings": false,
|
| 25 |
+
"torch_dtype": "bfloat16",
|
| 26 |
+
"transformers_version": "4.32.0",
|
| 27 |
+
"use_cache": true,
|
| 28 |
+
"vocab_size": 32000,
|
| 29 |
+
"quantization_config": {
|
| 30 |
+
"quant_method": "awq",
|
| 31 |
+
"zero_point": true,
|
| 32 |
+
"group_size": 128,
|
| 33 |
+
"bits": 4,
|
| 34 |
+
"version": "gemm"
|
| 35 |
+
}
|
| 36 |
}
|