alphaXiv commited on
Commit
d10468d
·
verified ·
1 Parent(s): 859106b

perf: update dtype to bfloat16

Browse files
Files changed (1) hide show
  1. config.json +4 -4
config.json CHANGED
@@ -2,7 +2,7 @@
2
  "architectures": [
3
  "Qwen3_5ForConditionalGeneration"
4
  ],
5
- "dtype": "float32",
6
  "image_token_id": 248056,
7
  "model_type": "qwen3_5",
8
  "text_config": {
@@ -10,7 +10,7 @@
10
  "attention_dropout": 0.0,
11
  "attn_output_gate": true,
12
  "bos_token_id": null,
13
- "dtype": "float32",
14
  "eos_token_id": 248044,
15
  "full_attention_interval": 4,
16
  "head_dim": 256,
@@ -91,7 +91,7 @@
91
  "vision_config": {
92
  "deepstack_visual_indexes": [],
93
  "depth": 24,
94
- "dtype": "float32",
95
  "hidden_act": "gelu_pytorch_tanh",
96
  "hidden_size": 1024,
97
  "in_channels": 3,
@@ -107,4 +107,4 @@
107
  },
108
  "vision_end_token_id": 248054,
109
  "vision_start_token_id": 248053
110
- }
 
2
  "architectures": [
3
  "Qwen3_5ForConditionalGeneration"
4
  ],
5
+ "dtype": "bfloat16",
6
  "image_token_id": 248056,
7
  "model_type": "qwen3_5",
8
  "text_config": {
 
10
  "attention_dropout": 0.0,
11
  "attn_output_gate": true,
12
  "bos_token_id": null,
13
+ "dtype": "bfloat16",
14
  "eos_token_id": 248044,
15
  "full_attention_interval": 4,
16
  "head_dim": 256,
 
91
  "vision_config": {
92
  "deepstack_visual_indexes": [],
93
  "depth": 24,
94
+ "dtype": "bfloat16",
95
  "hidden_act": "gelu_pytorch_tanh",
96
  "hidden_size": 1024,
97
  "in_channels": 3,
 
107
  },
108
  "vision_end_token_id": 248054,
109
  "vision_start_token_id": 248053
110
+ }