Text Generation
Transformers
Safetensors
llama
conversational
Inference Endpoints
text-generation-inference
Files changed (1) hide show
  1. README.md +118 -2
README.md CHANGED
@@ -25,7 +25,110 @@ datasets:
25
  - Intel/orca_dpo_pairs
26
  - unalignment/toxic-dpo-v0.1
27
  - jondurbin/truthy-dpo-v0.1
28
- - allenai/ultrafeedback_binarized_cleaned
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
29
  ---
30
 
31
  # A bagel, with everything
@@ -334,4 +437,17 @@ accelerate launch bagel/tune/dpo.py \
334
  --deepspeed deepspeed.json \
335
  --save_steps 25 \
336
  --save_total_limit 5
337
- ```
 
 
 
 
 
 
 
 
 
 
 
 
 
 
25
  - Intel/orca_dpo_pairs
26
  - unalignment/toxic-dpo-v0.1
27
  - jondurbin/truthy-dpo-v0.1
28
+ - allenai/ultrafeedback_binarized_cleaned
29
+ model-index:
30
+ - name: bagel-dpo-7b-v0.1
31
+ results:
32
+ - task:
33
+ type: text-generation
34
+ name: Text Generation
35
+ dataset:
36
+ name: AI2 Reasoning Challenge (25-Shot)
37
+ type: ai2_arc
38
+ config: ARC-Challenge
39
+ split: test
40
+ args:
41
+ num_few_shot: 25
42
+ metrics:
43
+ - type: acc_norm
44
+ value: 66.72
45
+ name: normalized accuracy
46
+ source:
47
+ url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=jondurbin/bagel-dpo-7b-v0.1
48
+ name: Open LLM Leaderboard
49
+ - task:
50
+ type: text-generation
51
+ name: Text Generation
52
+ dataset:
53
+ name: HellaSwag (10-Shot)
54
+ type: hellaswag
55
+ split: validation
56
+ args:
57
+ num_few_shot: 10
58
+ metrics:
59
+ - type: acc_norm
60
+ value: 84.16
61
+ name: normalized accuracy
62
+ source:
63
+ url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=jondurbin/bagel-dpo-7b-v0.1
64
+ name: Open LLM Leaderboard
65
+ - task:
66
+ type: text-generation
67
+ name: Text Generation
68
+ dataset:
69
+ name: MMLU (5-Shot)
70
+ type: cais/mmlu
71
+ config: all
72
+ split: test
73
+ args:
74
+ num_few_shot: 5
75
+ metrics:
76
+ - type: acc
77
+ value: 64.24
78
+ name: accuracy
79
+ source:
80
+ url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=jondurbin/bagel-dpo-7b-v0.1
81
+ name: Open LLM Leaderboard
82
+ - task:
83
+ type: text-generation
84
+ name: Text Generation
85
+ dataset:
86
+ name: TruthfulQA (0-shot)
87
+ type: truthful_qa
88
+ config: multiple_choice
89
+ split: validation
90
+ args:
91
+ num_few_shot: 0
92
+ metrics:
93
+ - type: mc2
94
+ value: 64.05
95
+ source:
96
+ url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=jondurbin/bagel-dpo-7b-v0.1
97
+ name: Open LLM Leaderboard
98
+ - task:
99
+ type: text-generation
100
+ name: Text Generation
101
+ dataset:
102
+ name: Winogrande (5-shot)
103
+ type: winogrande
104
+ config: winogrande_xl
105
+ split: validation
106
+ args:
107
+ num_few_shot: 5
108
+ metrics:
109
+ - type: acc
110
+ value: 80.9
111
+ name: accuracy
112
+ source:
113
+ url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=jondurbin/bagel-dpo-7b-v0.1
114
+ name: Open LLM Leaderboard
115
+ - task:
116
+ type: text-generation
117
+ name: Text Generation
118
+ dataset:
119
+ name: GSM8k (5-shot)
120
+ type: gsm8k
121
+ config: main
122
+ split: test
123
+ args:
124
+ num_few_shot: 5
125
+ metrics:
126
+ - type: acc
127
+ value: 47.61
128
+ name: accuracy
129
+ source:
130
+ url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=jondurbin/bagel-dpo-7b-v0.1
131
+ name: Open LLM Leaderboard
132
  ---
133
 
134
  # A bagel, with everything
 
437
  --deepspeed deepspeed.json \
438
  --save_steps 25 \
439
  --save_total_limit 5
440
+ ```
441
+ # [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
442
+ Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_jondurbin__bagel-dpo-7b-v0.1)
443
+
444
+ | Metric |Value|
445
+ |---------------------------------|----:|
446
+ |Avg. |67.95|
447
+ |AI2 Reasoning Challenge (25-Shot)|66.72|
448
+ |HellaSwag (10-Shot) |84.16|
449
+ |MMLU (5-Shot) |64.24|
450
+ |TruthfulQA (0-shot) |64.05|
451
+ |Winogrande (5-shot) |80.90|
452
+ |GSM8k (5-shot) |47.61|
453
+