Glenn Jocher
e16e9e43e1
new nc=len(names) check
4 anni fa
Glenn Jocher
cb527d3af9
new nc=len(names) check
4 anni fa
Glenn Jocher
603ea0bfdc
update log_dir to runs/exp #107
4 anni fa
Glenn Jocher
72d5b58b9a
disable LR plot to suppress warning message
4 anni fa
Glenn Jocher
24c5a941f0
--resume EMA fix #292
4 anni fa
Alex Stoken
9d631408a2
Move hyp and opt yaml save to top of train()
Fixes bug where scaled values were saved in hyp.yaml, which would cause continuity issues with --resume
4 anni fa
Glenn Jocher
bf6f41567a
hyperparameter printout update
4 anni fa
Glenn Jocher
6b134d93c5
Update train.py
4 anni fa
Glenn Jocher
16f6834486
update train.py and experimental.py
4 anni fa
Alex Stoken
52bac22f09
Add in --resume functionality with option to specify path or to get most recent run
4 anni fa
Laughing
956511dafd
fix LR bug
4 anni fa
Glenn Jocher
bb3c346916
model.yaml nc inherited from dataset.yaml
4 anni fa
Glenn Jocher
df224a0d8f
EMA bug fix #279
4 anni fa
Glenn Jocher
3bdea3f697
strip_optimizer() bug fix #253
4 anni fa
Glenn Jocher
597ed4ce63
Update train.py
4 anni fa
Glenn Jocher
9705c15da1
Update train.py
4 anni fa
ChristopherSTAN
a8d17ce712
Update train.py
4 anni fa
Glenn Jocher
13f69777a6
typo fix
4 anni fa
yxNONG
1aa2b67933
Update train.py
4 anni fa
Glenn Jocher
86784cfdbf
--resume bug fix #252
4 anni fa
Glenn Jocher
ad4c22cbfe
--resume bug fix #187
4 anni fa
Glenn Jocher
3b16c865f0
assert --epochs
4 anni fa
yxNONG
cdb9bde181
Unify the check point of single and multi GPU
save the model.hyp etc to checkpoint when use multi GPU training
4 anni fa
Glenn Jocher
b203c9b7ff
update train.py incompatible model message fix #222
4 anni fa
Glenn Jocher
37e13f8846
update mosaic border
4 anni fa
Glenn Jocher
22fb2b0c25
refactor dataloader
4 anni fa
Glenn Jocher
256a3e89d2
small dataset bug fix #140
4 anni fa
Glenn Jocher
b50fdf16af
model.names multi-GPU bug fix #94
4 anni fa
Alex Stoken
de191655e4
Fix yaml saving (don't sort keys), reorder --opt keys, bug fix hyp dict accessor
4 anni fa
Alex Stoken
2d396bea00
Fix bug in --help from percent sign in help string
4 anni fa
Glenn Jocher
b8557f87e3
add stride to datasets.py
4 anni fa
Alex Stoken
611aacf1bf
Turn opt into dictionary before sending it to yaml
4 anni fa
Alex Stoken
bc4ef4861b
Default optimizer SGD
4 anni fa
Alex Stoken
7abf202cad
Mode all optimizer settings to 'hyp.yaml', integrate proper momentum with Adam optimizer
4 anni fa
Alex Stoken
7edbf6570e
Fix help message for cfg files
4 anni fa
Alex Stoken
d64ad0fbf3
Remove --resume functionality and related checks/logic.
4 anni fa
Glenn Jocher
6c1b87a42e
update google_utils import
4 anni fa
Glenn Jocher
1f1917ef56
remove fast, add merge
4 anni fa
Alex Stoken
e572bb0803
Add plot_results save location to log_dir
4 anni fa
Lornatang
899f1d4bde
Fix DDP bug in single process multiple device use cases
4 anni fa
Glenn Jocher
cce95e744d
backbone as FP16, save default to FP32
4 anni fa
Glenn Jocher
d9b64c27c2
save ckpt in FP16 #119
4 anni fa
Glenn Jocher
9fdb0fbacf
AutoAnchor bug fix # 117
4 anni fa
Alex Stoken
c8152c81a6
Syntax fixes
4 anni fa
Alex Stoken
9b7386f603
Add save_dir arg to test.test, use arg as location for saving batch jpgs
4 anni fa
Alex Stoken
945307beba
Add save_dir to plot_lr_scheduler and plot_labels
Set save_dir = log_dir in train.py
4 anni fa
Alex Stoken
3b2b330872
Move results.txt from weights/ to log_dir
4 anni fa
Alex Stoken
ade023cff2
Fix hyp file read in and dict update.
Add example of hyp yaml
4 anni fa
Lornatang
2368603484
fix refrence bug
In torch==1.5, the import of the API has changed. Although it does not interrupt the operation of the program, it seems to me to be an implicit error and may throw an exception in later versions.
4 anni fa
Glenn Jocher
8db51c7002
tb_writer bug fix
4 anni fa