Glenn Jocher
19d03a955c
Remove DDP process group timeout ( #4422 )
2021-08-15 18:32:41 +02:00
Glenn Jocher
24bea5e4b7
Standardize headers and docstrings ( #4417 )
...
* Implement new headers
* Reformat 1
* Reformat 2
* Reformat 3 - math
* Reformat 4 - yaml
2021-08-14 21:17:51 +02:00
Glenn Jocher
ce7deec440
int(mlc) (#4385 )
2021-08-11 17:32:13 +02:00
Glenn Jocher
86c7150cfd
Update newline ( #4308 )
2021-08-04 17:41:38 +02:00
Glenn Jocher
e78aeac973
Evolve in CSV format ( #4307 )
...
* Update evolution to CSV format
* Update
* Update
* Update
* Update
* Update
* reset args
* reset args
* reset args
* plot_results() fix
* Cleanup
* Cleanup2
2021-08-04 17:13:38 +02:00
junji hashimoto
2d99063201
Feature python train.py --cache disk ( #4049 )
...
* Add cache-on-disk and cache-directory to cache images on disk
* Fix load_image with cache_on_disk
* Add no_cache flag for load_image
* Revert the parts('logging' and a new line) that do not need to be modified
* Add the assertion for shapes of cached images
* Add a suffix string for cached images
* Fix boundary-error of letterbox for load_mosaic
* Add prefix as cache-key of cache-on-disk
* Update cache-function on disk
* Add psutil in requirements.txt
* Update train.py
* Cleanup1
* Cleanup2
* Skip existing npy
* Include re-space
* Export return character fix
Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>
2021-08-02 18:47:24 +02:00
Kalen Michael
b74929c910
Add train.py and val.py callbacks ( #4220 )
...
* added callbacks
* Update callbacks.py
* Update train.py
* Update val.py
* Fix CamlCase add staticmethod
* Refactor logger into callbacks
* Cleanup
* New callback on_val_image_end()
* Add curves and results images to TensorBoard
Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>
2021-08-01 00:18:07 +02:00
IneovaAI
bceb57b910
Add python train.py --freeze N argument ( #4238 )
...
* Add freeze as an argument
I train on different platforms and sometimes I want to freeze some layers. I have to go into the code and change it and also keep track of how many layers I froze on what platform. Please add the number of layers to freeze as an argument in future versions thanks.
* Update train.py
* Update train.py
* Cleanup
Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>
2021-07-30 17:39:48 +02:00
Glenn Jocher
8d3c3ef45c
Fix weight decay comment ( #4228 )
2021-07-30 01:35:39 +02:00
Glenn Jocher
c2c958c350
Explicit requirements.txt location ( #4225 )
2021-07-29 17:29:39 +02:00
Glenn Jocher
b60b62e874
PyCharm reformat ( #4209 )
...
* PyCharm reformat
* YAML reformat
* Markdown reformat
2021-07-28 23:35:14 +02:00
Ayush Chaurasia
e88e8f7a98
W&B: Restructure code to support the new dataset_check() feature ( #4197 )
...
* Improve docstrings and run names
* default wandb login prompt with timeout
* return key
* Update api_key check logic
* Properly support zipped dataset feature
* update docstring
* Revert tuorial change
* extend changes to log_dataset
* add run name
* bug fix
* bug fix
* Update comment
* fix import check
* remove unused import
* Hardcore .yaml file extension
* reduce code
* Reformat using pycharm
Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>
2021-07-28 17:40:08 +02:00
Glenn Jocher
5d66e48723
Train from --data path/to/dataset.zip feature ( #4185 )
...
* Train from `--data path/to/dataset.zip` feature
* Update dataset_stats()
* cleanup
* cleanup2
2021-07-28 02:04:10 +02:00
Glenn Jocher
0ad6301c96
Update script headers ( #4163 )
...
* Update download script headers
* cleanup
* bug fix attempt
* bug fix attempt2
* bug fix attempt3
* cleanup
2021-07-26 15:23:33 +02:00
Glenn Jocher
96e36a7c91
New CSV Logger ( #4148 )
...
* New CSV Logger
* cleanup
* move batch plots into Logger
* rename comment
* Remove total loss from progress bar
* mloss :-1 bug fix
* Update plot_results()
* Update plot_results()
* plot_results bug fix
2021-07-25 19:06:37 +02:00
Glenn Jocher
efe60b5681
Refactor train.py and val.py loggers ( #4137 )
...
* Update loggers
* Config
* Update val.py
* cleanup
* fix1
* fix2
* fix3 and reformat
* format sweep.py
* Logger() class
* cleanup
* cleanup2
* wandb package import fix
* wandb package import fix2
* txt fix
* fix4
* fix5
* fix6
* drop wandb into utils/loggers
* fix 7
* rename loggers/wandb_logging to loggers/wandb
* Update message
* Update message
* Update message
* cleanup
* Fix x axis bug
* fix rank 0 issue
* cleanup
2021-07-25 01:18:39 +02:00
Glenn Jocher
63dd65e7ed
Update train.py ( #4136 )
...
* Refactor train.py
* Update imports
* Update imports
* Update optimizer
* cleanup
2021-07-24 16:11:39 +02:00
Glenn Jocher
2c073cd207
Add train.py `--img-size floor ( #4099 )
2021-07-21 16:50:47 +02:00
Glenn Jocher
f7d8562060
val.py refactor (#4053 )
...
* val.py refactor
* cleanup
* cleanup
* cleanup
* cleanup
* save after eval
* opt.imgsz bug fix
* wandb refactor
* dataloader to train_loader
* capitalize global variables
* runs/hub/exp to runs/detect/exp
* refactor wandb logging
* Refactor wandb operations (#4061 )
Co-authored-by: Ayush Chaurasia <ayush.chaurarsia@gmail.com>
2021-07-19 10:43:01 +02:00
Glenn Jocher
951922c735
Add --sync-bn known issue ( #4032 )
...
* Add `--sync-bn` known issue
* Update train.py
2021-07-17 13:07:19 +02:00
Glenn Jocher
720aaa65c8
Rename test.py to val.py ( #4000 )
2021-07-14 15:43:54 +02:00
Eldar Kurtic
e7888af94c
Fix inconsistent NMS IoU value for COCO ( #3934 )
...
Evaluation of 'best' and 'last' models will use the same params as the evaluation during the training phase.
This PR fixes https://github.com/ultralytics/yolov5/issues/3907
2021-07-08 15:29:02 +02:00
Glenn Jocher
8930e22cce
Evolution commented hyp['anchors'] fix ( #3887 )
...
Fix for `KeyError: 'anchors'` error when start hyperparameter evolution:
```bash
python train.py --evolve
```
```bash
Traceback (most recent call last):
File "E:\yolov5\train.py", line 623, in <module>
hyp[k] = max(hyp[k], v[1]) # lower limit
KeyError: 'anchors'
```
2021-07-05 12:48:27 +02:00
san-soucie
d3e9d69850
--evolve 300 generations CLI argument (#3863 )
...
* evolve command accepts argument for number of generations
* evolve generations argument used in evolve for loop
* evolve argument boolean fixes
* default to 300 evolve generations
* Update train.py
Co-authored-by: John San Soucie <jsansoucie@whoi.edu>
Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>
2021-07-04 12:14:35 +02:00
Glenn Jocher
c6c88dc601
Copy-Paste augmentation for YOLOv5 ( #3845 )
...
* Copy-paste augmentation initial commit
* if any segments
* Add obscuration rejection
* Add copy_paste hyperparameter
* Update comments
2021-07-01 00:35:04 +02:00
yellowdolphin
3974d725b6
Fix warmup accumulate ( #3722 )
...
* gradient accumulation during warmup in train.py
Context:
`accumulate` is the number of batches/gradients accumulated before calling the next optimizer.step().
During warmup, it is ramped up from 1 to the final value nbs / batch_size.
Although I have not seen this in other libraries, I like the idea. During warmup, as grads are large, too large steps are more of on issue than gradient noise due to small steps.
The bug:
The condition to perform the opt step is wrong
> if ni % accumulate == 0:
This produces irregular step sizes if `accumulate` is not constant. It becomes relevant when batch_size is small and `accumulate` changes many times during warmup.
This demo also shows the proposed solution, to use a ">=" condition instead:
https://colab.research.google.com/drive/1MA2z2eCXYB_BC5UZqgXueqL_y1Tz_XVq?usp=sharing
Further, I propose not to restrict the number of warmup iterations to >= 1000. If the user changes hyp['warmup_epochs'], this causes unexpected behavior. Also, it makes evolution unstable if this parameter was to be optimized.
* replace last_opt_step tracking by do_step(ni)
* add docstrings
* move down nw
* Update train.py
* revert math import move
Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>
2021-06-28 12:25:13 +02:00
Glenn Jocher
92d49fde35
Update seeds for single-GPU reproducibility ( #3789 )
...
For seed=0 on single-GPU.
2021-06-26 15:42:40 +02:00
Piotr Skalski
09246a5a33
fix/incorrect_fitness_import ( #3770 )
2021-06-25 16:16:18 +02:00
Glenn Jocher
f2d97ebb25
Remove DDP MultiHeadAttention fix ( #3768 )
2021-06-25 12:52:05 +02:00
Glenn Jocher
f79d7479da
Add optional dataset.yaml path attribute ( #3753 )
...
* Add optional dataset.yaml `path` attribute
@KalenMike
* pass locals to python scripts
* handle lists
* update coco128.yaml
* Capitalize first letter
* add test key
* finalize GlobalWheat2020.yaml
* finalize objects365.yaml
* finalize SKU-110K.yaml
* finalize SKU-110K.yaml
* finalize VisDrone.yaml
* NoneType fix
* update download comment
* voc to VOC
* update
* update VOC.yaml
* update VOC.yaml
* remove dashes
* delete get_voc.sh
* force coco and coco128 to ../datasets
* Capitalize Argoverse_HD.yaml
* Capitalize Objects365.yaml
* update Argoverse_HD.yaml
* coco segments fix
* VOC single-thread
* update Argoverse_HD.yaml
* update data_dict in test handling
* create root
2021-06-25 01:25:03 +02:00
Glenn Jocher
ae4261c774
Force non-zero hyp evolution weights w ( #3748 )
...
Fix for https://github.com/ultralytics/yolov5/issues/3741
2021-06-23 12:56:22 +02:00
Glenn Jocher
fdc22398fa
Create data/hyps directory ( #3747 )
2021-06-23 12:49:38 +02:00
Glenn Jocher
1f69d12591
Update 4 main ops for paths and .run() ( #3715 )
...
* Add yolov5/ to path
* rename functions to run()
* cleanup
* rename fix
* CI fix
* cleanup find models/export.py
2021-06-21 17:25:04 +02:00
Ayush Chaurasia
75c0ff43af
[x]W&B: Don't resume transfer learning runs ( #3604 )
...
* Allow config cahnge
* Allow val change in wandb config
* Don't resume transfer learning runs
* Add entity in log dataset
2021-06-21 14:00:25 +02:00
Glenn Jocher
e8810a53e8
Update DDP backend if dist.is_nccl_available() ( #3705 )
2021-06-20 17:15:42 +02:00
Glenn Jocher
fbf41e0913
Add train.run() method ( #3700 )
...
* Update train.py explicit arguments
* Update train.py
* Add run method
2021-06-20 15:06:58 +02:00
Glenn Jocher
c1af67dcd4
Add torch DP warning ( #3698 )
2021-06-19 19:50:46 +02:00
Glenn Jocher
b3e2f4e08d
Eliminate total_batch_size variable ( #3697 )
...
* Eliminate `total_batch_size` variable
* cleanup
* Update train.py
2021-06-19 19:14:59 +02:00
Glenn Jocher
fad27c0046
Update DDP for torch.distributed.run with gloo backend ( #3680 )
...
* Update DDP for `torch.distributed.run`
* Add LOCAL_RANK
* remove opt.local_rank
* backend="gloo|nccl"
* print
* print
* debug
* debug
* os.getenv
* gloo
* gloo
* gloo
* cleanup
* fix getenv
* cleanup
* cleanup destroy
* try nccl
* return opt
* add --local_rank
* add timeout
* add init_method
* gloo
* move destroy
* move destroy
* move print(opt) under if RANK
* destroy only RANK 0
* move destroy inside train()
* restore destroy outside train()
* update print(opt)
* cleanup
* nccl
* gloo with 60 second timeout
* update namespace printing
2021-06-19 16:30:25 +02:00
lb-desupervised
bfb2276b1d
Slightly modify CLI execution ( #3687 )
...
* Slightly modify CLI execution
This simple change makes it easier to run the primary functions of this
repo (train/detect/test) from within Python. An object which represents
`opt` can be constructed and fed to the `main` function of each of these
modules, rather than having to call the lower level functions directly,
or run the module as a script.
* Update export.py
Add CLI parsing update for more convenient module usage within Python.
Co-authored-by: Lewis Belcher <lb@desupervised.io>
2021-06-19 12:06:59 +02:00
Glenn Jocher
2296f1546f
Update WORLD_SIZE and RANK retrieval ( #3670 )
2021-06-17 23:24:30 +02:00
Glenn Jocher
045d5d8629
Update TensorBoard ( #3669 )
2021-06-17 22:12:42 +02:00
Glenn Jocher
fa201f968e
Update train(hyp, *args) to accept hyp file or dict ( #3668 )
2021-06-17 22:03:25 +02:00
Glenn Jocher
6d6e2ca65f
Update train.py ( #3667 )
2021-06-17 21:32:39 +02:00
Wei Quan
4c5d9bff80
Fix incorrect end epoch comment ( #3612 )
2021-06-15 11:24:56 +02:00
Glenn Jocher
4984cf54be
train.py GPU memory fix ( #3590 )
...
* train.py GPU memory fix
* ema
* cuda
* cuda
* zeros input
* to device
* batch index 0
2021-06-11 20:24:03 +02:00
Glenn Jocher
4695ca8314
Refactoring cleanup ( #3565 )
...
* Refactoring cleanup
* Update test.py
2021-06-09 22:50:27 +02:00
Glenn Jocher
5948f20a3d
Update test.py profiling ( #3555 )
...
* Update test.py profiling
* half_precision to half
* inplace
2021-06-09 16:25:17 +02:00
Glenn Jocher
63157d214d
Remove is_coco argument from test() ( #3553 )
2021-06-09 15:09:51 +02:00
Glenn Jocher
958ab92dc1
Remove opt from create_dataloader()` ( #3552 )
2021-06-09 13:14:56 +02:00
