Shashata commited on
Commit
2a9fbab
·
verified ·
1 Parent(s): 763f2b3

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +14 -20
README.md CHANGED
@@ -36,12 +36,6 @@ The [GPTQ](https://arxiv.org/abs/2210.17323) algorithm is applied for quantizati
36
  This model was created by using the [llm-compressor](https://github.com/vllm-project/llm-compressor) library as presented in the code snipet below.
37
 
38
  ```python
39
-
40
-
41
-
42
-
43
-
44
-
45
  from transformers import AutoTokenizer
46
  from llmcompressor.transformers import SparseAutoModelForCausalLM, oneshot
47
  from llmcompressor.modifiers.quantization import GPTQModifier
@@ -132,9 +126,9 @@ lm_eval \
132
  <tr>
133
  <td>MMLU (5-shot)
134
  </td>
135
- <td>26.22
136
  </td>
137
- <td>25.20
138
  </td>
139
  <td>96.12%
140
  </td>
@@ -162,41 +156,41 @@ lm_eval \
162
  <tr>
163
  <td>Hellaswag (10-shot)
164
  </td>
165
- <td>71.22
166
  </td>
167
- <td>70.53
168
  </td>
169
- <td>99.03%
170
  </td>
171
  </tr>
172
  <tr>
173
  <td>Winogrande (5-shot)
174
  </td>
175
- <td>68.67
176
  </td>
177
- <td>68.35
178
  </td>
179
- <td>99.53%
180
  </td>
181
  </tr>
182
  <tr>
183
  <td>TruthfulQA (0-shot)
184
  </td>
185
- <td>53.11
186
  </td>
187
- <td>52.57
188
  </td>
189
- <td>98.98%
190
  </td>
191
  </tr>
192
  <tr>
193
  <td><strong>Average</strong>
194
  </td>
195
- <td><strong>58.8</strong>
196
  </td>
197
- <td><strong>57.75</strong>
198
  </td>
199
- <td><strong>98.21%</strong>
200
  </td>
201
  </tr>
202
  </table>
 
36
  This model was created by using the [llm-compressor](https://github.com/vllm-project/llm-compressor) library as presented in the code snipet below.
37
 
38
  ```python
 
 
 
 
 
 
39
  from transformers import AutoTokenizer
40
  from llmcompressor.transformers import SparseAutoModelForCausalLM, oneshot
41
  from llmcompressor.modifiers.quantization import GPTQModifier
 
126
  <tr>
127
  <td>MMLU (5-shot)
128
  </td>
129
+ <td>26.220
130
  </td>
131
+ <td>25.202
132
  </td>
133
  <td>96.12%
134
  </td>
 
156
  <tr>
157
  <td>Hellaswag (10-shot)
158
  </td>
159
+ <td>41.41
160
  </td>
161
+ <td>40.81
162
  </td>
163
+ <td>98.55%
164
  </td>
165
  </tr>
166
  <tr>
167
  <td>Winogrande (5-shot)
168
  </td>
169
+ <td>50.039
170
  </td>
171
+ <td>53.591
172
  </td>
173
+ <td>107.10%
174
  </td>
175
  </tr>
176
  <tr>
177
  <td>TruthfulQA (0-shot)
178
  </td>
179
+ <td>40.38
180
  </td>
181
+ <td>39.87
182
  </td>
183
+ <td>98.74%
184
  </td>
185
  </tr>
186
  <tr>
187
  <td><strong>Average</strong>
188
  </td>
189
+ <td><strong>31.55</strong>
190
  </td>
191
+ <td><strong>31.91</strong>
192
  </td>
193
+ <td><strong>101.16%</strong>
194
  </td>
195
  </tr>
196
  </table>