Upload Apollo-72B/config.json with huggingface_hub
Browse files- Apollo-72B/config.json +43 -0
Apollo-72B/config.json
ADDED
|
@@ -0,0 +1,43 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "/sds_wangby/models/Qwen-72B",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"QWenLMHeadModel"
|
| 5 |
+
],
|
| 6 |
+
"attn_dropout_prob": 0.0,
|
| 7 |
+
"auto_map": {
|
| 8 |
+
"AutoConfig": "configuration_qwen.QWenConfig",
|
| 9 |
+
"AutoModelForCausalLM": "modeling_qwen.QWenLMHeadModel"
|
| 10 |
+
},
|
| 11 |
+
"bf16": true,
|
| 12 |
+
"emb_dropout_prob": 0.0,
|
| 13 |
+
"fp16": false,
|
| 14 |
+
"fp32": false,
|
| 15 |
+
"hidden_size": 8192,
|
| 16 |
+
"initializer_range": 0.02,
|
| 17 |
+
"intermediate_size": 49152,
|
| 18 |
+
"kv_channels": 128,
|
| 19 |
+
"layer_norm_epsilon": 1e-06,
|
| 20 |
+
"max_position_embeddings": 32768,
|
| 21 |
+
"model_type": "qwen",
|
| 22 |
+
"no_bias": true,
|
| 23 |
+
"num_attention_heads": 64,
|
| 24 |
+
"num_hidden_layers": 80,
|
| 25 |
+
"onnx_safe": null,
|
| 26 |
+
"rope_theta": 1000000,
|
| 27 |
+
"rotary_emb_base": 1000000,
|
| 28 |
+
"rotary_pct": 1.0,
|
| 29 |
+
"scale_attn_weights": true,
|
| 30 |
+
"seq_length": 32768,
|
| 31 |
+
"softmax_in_fp32": false,
|
| 32 |
+
"tie_word_embeddings": false,
|
| 33 |
+
"tokenizer_class": "QWenTokenizer",
|
| 34 |
+
"torch_dtype": "bfloat16",
|
| 35 |
+
"transformers_version": "4.35.2",
|
| 36 |
+
"use_cache": true,
|
| 37 |
+
"use_cache_kernel": false,
|
| 38 |
+
"use_cache_quantization": false,
|
| 39 |
+
"use_dynamic_ntk": false,
|
| 40 |
+
"use_flash_attn": true,
|
| 41 |
+
"use_logn_attn": false,
|
| 42 |
+
"vocab_size": 152064
|
| 43 |
+
}
|