Update test_precision_recall_curve.py#2655
Conversation
Update for ddp config for updating in batch way
vfdev-5
left a comment
vfdev-5
left a comment
There was a problem hiding this comment.
Few nits to fix, otherwise lgtm. Thanks !
revert to previous way for generating data
| assert precision_recall_curve(np_y, np_y_pred)[2] == pytest.approx(res[2].cpu().numpy()) | ||
|
|
||
| def get_test_cases(): | ||
| torch.manual_seed(12 + rank) |
There was a problem hiding this comment.
Actually, the problem with setting seed here makes outer loop useless:
https://github.com/pytorch/ignite/pull/2655/files#diff-696528ba9e809a569bf3efba773b235ac7b582dd9dbc5811d62ff046182e0209R183
There was a problem hiding this comment.
@vfdev-5 I don't understand why outer loop does not work when seed is set here. with get_test_cases. It generates as expected. May I ask you to explain a little bit more.
There was a problem hiding this comment.
Outer loop works but we generate the same data 3 times which useless
There was a problem hiding this comment.
Oh, ok let me check again and correct it.
To generate different data
* Update test_precision_recall_curve.py Update for ddp config for updating in batch way * Update test_precision_recall_curve.py revert to previous way for generating data * Update test_precision_recall_curve.py To generate different data * Update test_precision_recall_curve.py Co-authored-by: vfdev <[email protected]>
2 participants
Description: Update for ddp config for updating in batch way
Check list: