Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -134,10 +134,10 @@ Benchmarking is one of the most important procedures during model acceleration.
|
|
| 134 |
|
| 135 |
| Metric/Model | S | M | L | XL | Original | W8A8, int8 |
|
| 136 |
|---------------|---|---|---|----|----------|------------|
|
| 137 |
-
| arc_challenge | 41.00 | 40.90 | 42.50 | 42.20 | 42.20 |
|
| 138 |
-
| mmlu | 52.00 | 53.80 | 55.10 | 55.20 | 55.20 |
|
| 139 |
-
| piqa | 68.40 | 70.60 | 70.70 | 70.50 | 70.50 |
|
| 140 |
-
| winogrande | 60.10 | 60.90 | 60.20 | 60.10 | 60.10 |
|
| 141 |
|
| 142 |
|
| 143 |
|
|
@@ -152,7 +152,7 @@ __100 input/300 output; tok/s:__
|
|
| 152 |
|
| 153 |
| GPU/Model | S | M | L | XL | Original | W8A8, int8 |
|
| 154 |
|-----------|-----|---|---|----|----------|------------|
|
| 155 |
-
| H100 |
|
| 156 |
| L40S | 78 | 69 | 60 | 47 | 43 | 78 | - |
|
| 157 |
|
| 158 |
|
|
|
|
| 134 |
|
| 135 |
| Metric/Model | S | M | L | XL | Original | W8A8, int8 |
|
| 136 |
|---------------|---|---|---|----|----------|------------|
|
| 137 |
+
| arc_challenge | 41.00 | 40.90 | 42.50 | 42.20 | 42.20 | 41.00 | - |
|
| 138 |
+
| mmlu | 52.00 | 53.80 | 55.10 | 55.20 | 55.20 | 52.00 | - |
|
| 139 |
+
| piqa | 68.40 | 70.60 | 70.70 | 70.50 | 70.50 | 68.40 | - |
|
| 140 |
+
| winogrande | 60.10 | 60.90 | 60.20 | 60.10 | 60.10 | 60.10 | - |
|
| 141 |
|
| 142 |
|
| 143 |
|
|
|
|
| 152 |
|
| 153 |
| GPU/Model | S | M | L | XL | Original | W8A8, int8 |
|
| 154 |
|-----------|-----|---|---|----|----------|------------|
|
| 155 |
+
| H100 | 204 | 185 | 161 | 135 | 62 | 205 | - |
|
| 156 |
| L40S | 78 | 69 | 60 | 47 | 43 | 78 | - |
|
| 157 |
|
| 158 |
|