HongChuang/ultralytics-ascend
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 3

ultralytics 8.0.76 minor fixes and improvements (#2004)

Browse source 
Co-authored-by: Seungtaek Kim <seungtaek.kim.94@gmail.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Laughing <61612323+Laughing-q@users.noreply.github.com>
Co-authored-by: Yonghye Kwon <developer.0hye@gmail.com>
Co-authored-by: Ercalvez <45692523+Ercalvez@users.noreply.github.com>
Co-authored-by: Erwan CALVEZ <ecalvez@enib.fr>
This commit is contained in:
Glenn Jocher 2023-04-13 20:19:42 +02:00 • committed by GitHub
parent 48c4483795
commit 4916014af2
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
15 changed files with 63 additions and 26 deletions
Download patch file Download diff file Expand all files Collapse all files

4
ultralytics/yolo/engine/trainer.py
View file

@ -632,9 +632,9 @@ def check_amp(model):
def amp_allclose(m, im):
# All close FP32 vs AMP results
a = m(im, device=device, verbose=False)[0].boxes.boxes # FP32 inference
a = m(im, device=device, verbose=False)[0].boxes.data # FP32 inference
with torch.cuda.amp.autocast(True):
b = m(im, device=device, verbose=False)[0].boxes.boxes # AMP inference
b = m(im, device=device, verbose=False)[0].boxes.data # AMP inference
del m
return a.shape == b.shape and torch.allclose(a, b.float(), atol=0.5) # close to 0.5 absolute tolerance