Hugging Face's logo

opencsg
/
opencsg-CodeLlama-34b-v0.2

Text Generation
Transformers
Safetensors
code
llama
llama-2
text-generation-inference
Model card Files
xet
Community
AmeliaYin commited on
Commit
3427e63
·
verified ·
1 Parent(s): ea6a304

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +9 -0
README.md CHANGED
@@ -42,9 +42,12 @@ This is the repository for the base 34B version finetuned based on [CodeLlama-34
42
  | 7B | [opencsg/Opencsg-CodeLlama-7b-v0.1](https://huggingface.co/opencsg/opencsg-CodeLlama-7b-v0.1) |
43
  | 13B | [opencsg/Opencsg-CodeLlama-13b-v0.1](https://huggingface.co/opencsg/opencsg-CodeLlama-13b-v0.1) |
44
  | 34B | [opencsg/Opencsg-CodeLlama-34b-v0.1](https://huggingface.co/opencsg/opencsg-CodeLlama-34b-v0.1) |
 
 
45
  | 34B | [opencsg/Opencsg-CodeLlama-34b-v0.2](https://huggingface.co/opencsg/opencsg-CodeLlama-34b-v0.2) |
46
 
47
 
 
48
  ## Model Eval
49
 
50
  HumanEval is the most common code generation benchmark for evaluating model performance, especially on the compeltion of code exercise cases.
@@ -62,8 +65,10 @@ To simplify the comparison, we chosed the Pass@1 metric for the Python language,
62
  | --- |----------------------------------------------------------------------------- |
63
  | CodeLlama-7b-hf | 30.5%|
64
  | **opencsg-CodeLlama-7b-v0.1** | **43.9%** |
 
65
  | CodeLlama-13b-hf | 36.0%|
66
  | **opencsg-CodeLlama-13b-v0.1** | **51.2%** |
 
67
  | CodeLlama-34b-hf | 48.2%|
68
  | **opencsg-CodeLlama-34b-v0.1**| **56.1%** |
69
  | **opencsg-CodeLlama-34b-v0.2**| **64.0%** |
@@ -161,6 +166,8 @@ opencsg-CodeLlama-v0.2是一系列基于CodeLlama的通过全参数微调方法
161
  | 7B | [opencsg/Opencsg-CodeLlama-7b-v0.1](https://huggingface.co/opencsg/opencsg-CodeLlama-7b-v0.1) |
162
  | 13B | [opencsg/Opencsg-CodeLlama-13b-v0.1](https://huggingface.co/opencsg/opencsg-CodeLlama-13b-v0.1) |
163
  | 34B | [opencsg/Opencsg-CodeLlama-34b-v0.1](https://huggingface.co/opencsg/opencsg-CodeLlama-34b-v0.1) |
 
 
164
  | 34B | [opencsg/Opencsg-CodeLlama-34b-v0.2](https://huggingface.co/opencsg/opencsg-CodeLlama-34b-v0.2) |
165
 
166
 
@@ -181,8 +188,10 @@ HumanEval 是评估模型在代码生成方面性能的最常见的基准，尤
181
  | --- |----------------------------------------------------------------------------- |
182
  | CodeLlama-7b-hf | 30.5%|
183
  | **opencsg-CodeLlama-7b-v0.1** | **43.9%** |
 
184
  | CodeLlama-13b-hf | 36.0%|
185
  | **opencsg-CodeLlama-13b-v0.1** | **51.2%** |
 
186
  | CodeLlama-34b-hf | 48.2%|
187
  | **opencsg-CodeLlama-34b-v0.1**| **56.1%** |
188
  | **opencsg-CodeLlama-34b-v0.2**| **64.0%** |
 
42
  | 7B | [opencsg/Opencsg-CodeLlama-7b-v0.1](https://huggingface.co/opencsg/opencsg-CodeLlama-7b-v0.1) |
43
  | 13B | [opencsg/Opencsg-CodeLlama-13b-v0.1](https://huggingface.co/opencsg/opencsg-CodeLlama-13b-v0.1) |
44
  | 34B | [opencsg/Opencsg-CodeLlama-34b-v0.1](https://huggingface.co/opencsg/opencsg-CodeLlama-34b-v0.1) |
45
+ | 7B | [opencsg/Opencsg-CodeLlama-7b-v0.2](https://huggingface.co/opencsg/opencsg-CodeLlama-7b-v0.2) |
46
+ | 13B | [opencsg/Opencsg-CodeLlama-13b-v0.2](https://huggingface.co/opencsg/opencsg-CodeLlama-13b-v0.2) |
47
  | 34B | [opencsg/Opencsg-CodeLlama-34b-v0.2](https://huggingface.co/opencsg/opencsg-CodeLlama-34b-v0.2) |
48
 
49
 
50
+
51
  ## Model Eval
52
 
53
  HumanEval is the most common code generation benchmark for evaluating model performance, especially on the compeltion of code exercise cases.
 
65
  | --- |----------------------------------------------------------------------------- |
66
  | CodeLlama-7b-hf | 30.5%|
67
  | **opencsg-CodeLlama-7b-v0.1** | **43.9%** |
68
+ | **opencsg-CodeLlama-7b-v0.2** | **50.0%** |
69
  | CodeLlama-13b-hf | 36.0%|
70
  | **opencsg-CodeLlama-13b-v0.1** | **51.2%** |
71
+ | **opencsg-CodeLlama-13b-v0.2** | **55.5%** |
72
  | CodeLlama-34b-hf | 48.2%|
73
  | **opencsg-CodeLlama-34b-v0.1**| **56.1%** |
74
  | **opencsg-CodeLlama-34b-v0.2**| **64.0%** |
 
166
  | 7B | [opencsg/Opencsg-CodeLlama-7b-v0.1](https://huggingface.co/opencsg/opencsg-CodeLlama-7b-v0.1) |
167
  | 13B | [opencsg/Opencsg-CodeLlama-13b-v0.1](https://huggingface.co/opencsg/opencsg-CodeLlama-13b-v0.1) |
168
  | 34B | [opencsg/Opencsg-CodeLlama-34b-v0.1](https://huggingface.co/opencsg/opencsg-CodeLlama-34b-v0.1) |
169
+ | 7B | [opencsg/Opencsg-CodeLlama-7b-v0.2](https://huggingface.co/opencsg/opencsg-CodeLlama-7b-v0.2) |
170
+ | 13B | [opencsg/Opencsg-CodeLlama-13b-v0.2](https://huggingface.co/opencsg/opencsg-CodeLlama-13b-v0.2) |
171
  | 34B | [opencsg/Opencsg-CodeLlama-34b-v0.2](https://huggingface.co/opencsg/opencsg-CodeLlama-34b-v0.2) |
172
 
173
 
 
188
  | --- |----------------------------------------------------------------------------- |
189
  | CodeLlama-7b-hf | 30.5%|
190
  | **opencsg-CodeLlama-7b-v0.1** | **43.9%** |
191
+ | **opencsg-CodeLlama-7b-v0.2** | **50.0%** |
192
  | CodeLlama-13b-hf | 36.0%|
193
  | **opencsg-CodeLlama-13b-v0.1** | **51.2%** |
194
+ | **opencsg-CodeLlama-13b-v0.2** | **55.5%** |
195
  | CodeLlama-34b-hf | 48.2%|
196
  | **opencsg-CodeLlama-34b-v0.1**| **56.1%** |
197
  | **opencsg-CodeLlama-34b-v0.2**| **64.0%** |