Upload config.json with huggingface_hub
Browse files- config.json +4 -5
config.json
CHANGED
|
@@ -1,19 +1,18 @@
|
|
| 1 |
{
|
| 2 |
-
"_name_or_path": "/apdcephfs_gy2/share_302508627/darrenhu/llm/model/open_A13B/7b_moe_t1_v53_long_context_stage_daxianwang_fix_renshe_hf",
|
| 3 |
"add_classification_head": false,
|
| 4 |
"anyres_pooling_size": 2,
|
| 5 |
"anyres_vit_max_image_size": null,
|
| 6 |
"anyres_vit_two_views": false,
|
| 7 |
"architectures": [
|
| 8 |
-
"
|
| 9 |
],
|
| 10 |
"attention_bias": false,
|
| 11 |
"attention_dropout": 0.1,
|
| 12 |
"attention_head_dim": 128,
|
| 13 |
"auto_map": {
|
| 14 |
"AutoConfig": "configuration_hunyuan.HunYuanConfig",
|
| 15 |
-
"AutoModel": "hunyuan.
|
| 16 |
-
"AutoModelForCausalLM": "hunyuan.
|
| 17 |
},
|
| 18 |
"bos_token_id": 1,
|
| 19 |
"cla_share_factor": 2,
|
|
@@ -187,7 +186,7 @@
|
|
| 187 |
"topk_group": null,
|
| 188 |
"torch_dtype": "bfloat16",
|
| 189 |
"transformers_version": "4.41.2",
|
| 190 |
-
"use_cache":
|
| 191 |
"use_cla": false,
|
| 192 |
"use_mixed_mlp_moe": true,
|
| 193 |
"use_mla": false,
|
|
|
|
| 1 |
{
|
|
|
|
| 2 |
"add_classification_head": false,
|
| 3 |
"anyres_pooling_size": 2,
|
| 4 |
"anyres_vit_max_image_size": null,
|
| 5 |
"anyres_vit_two_views": false,
|
| 6 |
"architectures": [
|
| 7 |
+
"HunYuanMoEV1ForCausalLM"
|
| 8 |
],
|
| 9 |
"attention_bias": false,
|
| 10 |
"attention_dropout": 0.1,
|
| 11 |
"attention_head_dim": 128,
|
| 12 |
"auto_map": {
|
| 13 |
"AutoConfig": "configuration_hunyuan.HunYuanConfig",
|
| 14 |
+
"AutoModel": "hunyuan.HunYuanModel",
|
| 15 |
+
"AutoModelForCausalLM": "hunyuan.HunYuanMoEV1ForCausalLM"
|
| 16 |
},
|
| 17 |
"bos_token_id": 1,
|
| 18 |
"cla_share_factor": 2,
|
|
|
|
| 186 |
"topk_group": null,
|
| 187 |
"torch_dtype": "bfloat16",
|
| 188 |
"transformers_version": "4.41.2",
|
| 189 |
+
"use_cache": true,
|
| 190 |
"use_cla": false,
|
| 191 |
"use_mixed_mlp_moe": true,
|
| 192 |
"use_mla": false,
|