feihu.hf
commited on
Commit
·
d616ac6
1
Parent(s):
30c5e98
update README
Browse files
README.md
CHANGED
|
@@ -318,10 +318,12 @@ YaRN is currently supported by several inference frameworks, e.g., `transformers
|
|
| 318 |
|
| 319 |
## Performance
|
| 320 |
|
| 321 |
-
| QUANTIZATION TYPE | AIME24 |
|
| 322 |
-
| --- | --- |
|
| 323 |
-
| bf16 | 76.0 |
|
| 324 |
-
| AWQ-int4 | 71.3 |
|
|
|
|
|
|
|
| 325 |
|
| 326 |
## Best Practices
|
| 327 |
|
|
|
|
| 318 |
|
| 319 |
## Performance
|
| 320 |
|
| 321 |
+
| Mode | QUANTIZATION TYPE | LiveBench 2024-11-25 | GPQA | MMLU-Redux | AIME24 |
|
| 322 |
+
| --- | --- | --- |
|
| 323 |
+
| Thinking | bf16 | 67.1 | 62.0 | 87.5 | 76.0 |
|
| 324 |
+
| Thinking | AWQ-int4 | 65.5 | 59.0 | 86.4 | 71.3 |
|
| 325 |
+
| Non-Thinking | bf16 | 53.5 | 39.3 | 79.5 | - |
|
| 326 |
+
| Non-Thinking | AWQ-int4 | 48.9 | 35.9 | 79.1 | - |
|
| 327 |
|
| 328 |
## Best Practices
|
| 329 |
|