Felix Marty
commited on
Commit
•
78caf38
1
Parent(s):
f6c4ac6
better readmem
Browse files
README.md
CHANGED
|
@@ -60,6 +60,8 @@ assert torch.allclose(res.logits, res_ort.logits, atol=1e-4)
|
|
| 60 |
|
| 61 |
More than x2 throughput with batch normalization folding and onnxruntime 🔥
|
| 62 |
|
|
|
|
|
|
|
| 63 |
```
|
| 64 |
PyTorch runtime:
|
| 65 |
|
|
|
|
| 60 |
|
| 61 |
More than x2 throughput with batch normalization folding and onnxruntime 🔥
|
| 62 |
|
| 63 |
+
Below you can find latency percentiles and mean (in ms), and the models throughput (in iterations/s).
|
| 64 |
+
|
| 65 |
```
|
| 66 |
PyTorch runtime:
|
| 67 |
|