Create config.json
Browse files- config.json +41 -0
config.json
ADDED
|
@@ -0,0 +1,41 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "openbmb/MiniCPM3-4B",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"MiniCPM3ForCausalLM"
|
| 5 |
+
],
|
| 6 |
+
"auto_map": {
|
| 7 |
+
"AutoConfig": "configuration_minicpm.MiniCPM3Config",
|
| 8 |
+
"AutoModel": "modeling_minicpm.MiniCPM3Model",
|
| 9 |
+
"AutoModelForCausalLM": "modeling_minicpm.MiniCPM3ForCausalLM",
|
| 10 |
+
"AutoModelForSeq2SeqLM": "modeling_minicpm.MiniCPM3ForCausalLM",
|
| 11 |
+
"AutoModelForSequenceClassification": "modeling_minicpm.MiniCPM3ForSequenceClassification"
|
| 12 |
+
},
|
| 13 |
+
"bos_token_id": 1,
|
| 14 |
+
"eos_token_id": [2, 73440],
|
| 15 |
+
"hidden_act": "silu",
|
| 16 |
+
"initializer_range": 0.1,
|
| 17 |
+
"hidden_size": 2560,
|
| 18 |
+
"num_hidden_layers": 62,
|
| 19 |
+
"intermediate_size": 6400,
|
| 20 |
+
"max_position_embeddings": 32768,
|
| 21 |
+
"num_attention_heads": 40,
|
| 22 |
+
"num_key_value_heads": 40,
|
| 23 |
+
"qk_nope_head_dim": 64,
|
| 24 |
+
"qk_rope_head_dim": 32,
|
| 25 |
+
"q_lora_rank": 768,
|
| 26 |
+
"kv_lora_rank": 256,
|
| 27 |
+
"rms_norm_eps": 1e-05,
|
| 28 |
+
"rope_scaling": {
|
| 29 |
+
"type": "longrope",
|
| 30 |
+
"long_factor": [1.0591234137867171, 1.1241891283591912, 1.2596935748670968, 1.5380380402321725, 2.093982484148734, 3.1446935121267696, 4.937952647693647, 7.524541999994549, 10.475458000005451, 13.062047352306353, 14.85530648787323, 15.906017515851266, 16.461961959767827, 16.740306425132907, 16.87581087164081, 16.940876586213285],
|
| 31 |
+
"short_factor": [1.0591234137867171, 1.1241891283591912, 1.2596935748670968, 1.5380380402321725, 2.093982484148734, 3.1446935121267696, 4.937952647693647, 7.524541999994549, 10.475458000005451, 13.062047352306353, 14.85530648787323, 15.906017515851266, 16.461961959767827, 16.740306425132907, 16.87581087164081, 16.940876586213285],
|
| 32 |
+
"original_max_position_embeddings": 32768
|
| 33 |
+
},
|
| 34 |
+
"torch_dtype": "bfloat16",
|
| 35 |
+
"transformers_version": "4.41.0",
|
| 36 |
+
"use_cache": true,
|
| 37 |
+
"vocab_size": 73448,
|
| 38 |
+
"scale_emb": 12,
|
| 39 |
+
"dim_model_base": 256,
|
| 40 |
+
"scale_depth": 1.4
|
| 41 |
+
}
|