NullSense commited on
Commit
cd252a8
·
verified ·
1 Parent(s): 58fb2ad

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +67 -1
README.md CHANGED
@@ -1,7 +1,73 @@
1
  ---
2
  license: apache-2.0
3
  base_model: Nanbeige/Nanbeige4.2-3B
4
- tags: [fp8, quantized, compressed-tensors, vllm, looped-transformer, nanbeige]
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5
  ---
6
 
7
  # Nanbeige4.2-3B-FP8-Dynamic
 
1
  ---
2
  license: apache-2.0
3
  base_model: Nanbeige/Nanbeige4.2-3B
4
+ base_model_relation: quantized
5
+ pipeline_tag: text-generation
6
+ language: [en, zh]
7
+ tags: [fp8, nvfp4, quantized, compressed-tensors, vllm, looped-transformer, nanbeige]
8
+ model-index:
9
+ - name: Nanbeige4.2-3B-FP8-Dynamic
10
+ results:
11
+ - task:
12
+ type: text-generation
13
+ dataset:
14
+ name: GSM8K
15
+ type: gsm8k
16
+ metrics:
17
+ - name: GSM8K (strict, thinking, n=100)
18
+ type: exact_match
19
+ value: 89.0
20
+ verified: false
21
+ - task:
22
+ type: text-generation
23
+ dataset:
24
+ name: GSM8K
25
+ type: gsm8k
26
+ metrics:
27
+ - name: GSM8K (flexible)
28
+ type: exact_match
29
+ value: 93.0
30
+ verified: false
31
+ - task:
32
+ type: text-generation
33
+ dataset:
34
+ name: IFEval
35
+ type: ifeval
36
+ metrics:
37
+ - name: IFEval prompt-strict (non-thinking, n=250)
38
+ type: exact_match
39
+ value: 76.4
40
+ verified: false
41
+ - task:
42
+ type: text-generation
43
+ dataset:
44
+ name: MMLU-Pro
45
+ type: mmlu_pro
46
+ metrics:
47
+ - name: MMLU-Pro (non-thinking, 25/category)
48
+ type: exact_match
49
+ value: 62.6
50
+ verified: false
51
+ - task:
52
+ type: text-generation
53
+ dataset:
54
+ name: BBH
55
+ type: bbh
56
+ metrics:
57
+ - name: BBH CoT few-shot (non-thinking)
58
+ type: exact_match
59
+ value: 64.9
60
+ verified: false
61
+ - task:
62
+ type: text-generation
63
+ dataset:
64
+ name: MultiHop-RAG
65
+ type: multihop_rag
66
+ metrics:
67
+ - name: MultiHop-RAG generator-only, gold evidence (n=248)
68
+ type: exact_match
69
+ value: 74.2
70
+ verified: false
71
  ---
72
 
73
  # Nanbeige4.2-3B-FP8-Dynamic