Files changed (1) hide show
  1. README.md +120 -4
README.md CHANGED
@@ -1,11 +1,114 @@
1
  ---
 
 
2
  license: apache-2.0
 
3
  datasets:
4
  - Intel/orca_dpo_pairs
5
  - mlabonne/chatml_dpo_pairs
6
- language:
7
- - en
8
- library_name: transformers
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9
  ---
10
  More information about previous [Neuronovo/neuronovo-9B-v0.2](https://huggingface.co/Neuronovo/neuronovo-9B-v0.2) version available here: 🔗[Don't stop DPOptimizing!](https://www.linkedin.com/pulse/dont-stop-dpoptimizing-jan-koco%2525C5%252584-mq4qf)
11
 
@@ -19,4 +122,17 @@ Changes concerning [Neuronovo/neuronovo-9B-v0.2](https://huggingface.co/Neuronov
19
 
20
  3. **Training Configuration**: The training approach has shifted from using `max_steps=200` to `num_train_epochs=1`. This represents a change in the training strategy, focusing on epoch-based training rather than a fixed number of steps.
21
 
22
- 4. **Learning Rate**: The learning rate has been reduced to a smaller value of `5e-8`. This finer learning rate allows for more precise adjustments during the training process, potentially leading to better model performance.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ language:
3
+ - en
4
  license: apache-2.0
5
+ library_name: transformers
6
  datasets:
7
  - Intel/orca_dpo_pairs
8
  - mlabonne/chatml_dpo_pairs
9
+ model-index:
10
+ - name: neuronovo-9B-v0.4
11
+ results:
12
+ - task:
13
+ type: text-generation
14
+ name: Text Generation
15
+ dataset:
16
+ name: AI2 Reasoning Challenge (25-Shot)
17
+ type: ai2_arc
18
+ config: ARC-Challenge
19
+ split: test
20
+ args:
21
+ num_few_shot: 25
22
+ metrics:
23
+ - type: acc_norm
24
+ value: 72.44
25
+ name: normalized accuracy
26
+ source:
27
+ url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Neuronovo/neuronovo-9B-v0.4
28
+ name: Open LLM Leaderboard
29
+ - task:
30
+ type: text-generation
31
+ name: Text Generation
32
+ dataset:
33
+ name: HellaSwag (10-Shot)
34
+ type: hellaswag
35
+ split: validation
36
+ args:
37
+ num_few_shot: 10
38
+ metrics:
39
+ - type: acc_norm
40
+ value: 88.33
41
+ name: normalized accuracy
42
+ source:
43
+ url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Neuronovo/neuronovo-9B-v0.4
44
+ name: Open LLM Leaderboard
45
+ - task:
46
+ type: text-generation
47
+ name: Text Generation
48
+ dataset:
49
+ name: MMLU (5-Shot)
50
+ type: cais/mmlu
51
+ config: all
52
+ split: test
53
+ args:
54
+ num_few_shot: 5
55
+ metrics:
56
+ - type: acc
57
+ value: 65.24
58
+ name: accuracy
59
+ source:
60
+ url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Neuronovo/neuronovo-9B-v0.4
61
+ name: Open LLM Leaderboard
62
+ - task:
63
+ type: text-generation
64
+ name: Text Generation
65
+ dataset:
66
+ name: TruthfulQA (0-shot)
67
+ type: truthful_qa
68
+ config: multiple_choice
69
+ split: validation
70
+ args:
71
+ num_few_shot: 0
72
+ metrics:
73
+ - type: mc2
74
+ value: 71.07
75
+ source:
76
+ url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Neuronovo/neuronovo-9B-v0.4
77
+ name: Open LLM Leaderboard
78
+ - task:
79
+ type: text-generation
80
+ name: Text Generation
81
+ dataset:
82
+ name: Winogrande (5-shot)
83
+ type: winogrande
84
+ config: winogrande_xl
85
+ split: validation
86
+ args:
87
+ num_few_shot: 5
88
+ metrics:
89
+ - type: acc
90
+ value: 80.66
91
+ name: accuracy
92
+ source:
93
+ url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Neuronovo/neuronovo-9B-v0.4
94
+ name: Open LLM Leaderboard
95
+ - task:
96
+ type: text-generation
97
+ name: Text Generation
98
+ dataset:
99
+ name: GSM8k (5-shot)
100
+ type: gsm8k
101
+ config: main
102
+ split: test
103
+ args:
104
+ num_few_shot: 5
105
+ metrics:
106
+ - type: acc
107
+ value: 62.77
108
+ name: accuracy
109
+ source:
110
+ url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Neuronovo/neuronovo-9B-v0.4
111
+ name: Open LLM Leaderboard
112
  ---
113
  More information about previous [Neuronovo/neuronovo-9B-v0.2](https://huggingface.co/Neuronovo/neuronovo-9B-v0.2) version available here: 🔗[Don't stop DPOptimizing!](https://www.linkedin.com/pulse/dont-stop-dpoptimizing-jan-koco%2525C5%252584-mq4qf)
114
 
 
122
 
123
  3. **Training Configuration**: The training approach has shifted from using `max_steps=200` to `num_train_epochs=1`. This represents a change in the training strategy, focusing on epoch-based training rather than a fixed number of steps.
124
 
125
+ 4. **Learning Rate**: The learning rate has been reduced to a smaller value of `5e-8`. This finer learning rate allows for more precise adjustments during the training process, potentially leading to better model performance.
126
+ # [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
127
+ Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_Neuronovo__neuronovo-9B-v0.4)
128
+
129
+ | Metric |Value|
130
+ |---------------------------------|----:|
131
+ |Avg. |73.42|
132
+ |AI2 Reasoning Challenge (25-Shot)|72.44|
133
+ |HellaSwag (10-Shot) |88.33|
134
+ |MMLU (5-Shot) |65.24|
135
+ |TruthfulQA (0-shot) |71.07|
136
+ |Winogrande (5-shot) |80.66|
137
+ |GSM8k (5-shot) |62.77|
138
+