You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

622 lines
33KB

  1. import argparse
  2. import logging
  3. import math
  4. import os
  5. import random
  6. import time
  7. from copy import deepcopy
  8. from pathlib import Path
  9. from threading import Thread
  10. import numpy as np
  11. import torch.distributed as dist
  12. import torch.nn as nn
  13. import torch.nn.functional as F
  14. import torch.optim as optim
  15. import torch.optim.lr_scheduler as lr_scheduler
  16. import torch.utils.data
  17. import yaml
  18. from torch.cuda import amp
  19. from torch.nn.parallel import DistributedDataParallel as DDP
  20. from torch.utils.tensorboard import SummaryWriter
  21. from tqdm import tqdm
  22. import test # import test.py to get mAP after each epoch
  23. from models.experimental import attempt_load
  24. from models.yolo import Model
  25. from utils.autoanchor import check_anchors
  26. from utils.datasets import create_dataloader
  27. from utils.general import labels_to_class_weights, increment_path, labels_to_image_weights, init_seeds, \
  28. fitness, strip_optimizer, get_latest_run, check_dataset, check_file, check_git_status, check_img_size, \
  29. check_requirements, print_mutation, set_logging, one_cycle, colorstr
  30. from utils.google_utils import attempt_download
  31. from utils.loss import ComputeLoss
  32. from utils.plots import plot_images, plot_labels, plot_results, plot_evolution
  33. from utils.torch_utils import ModelEMA, select_device, intersect_dicts, torch_distributed_zero_first, is_parallel
  34. from utils.wandb_logging.wandb_utils import WandbLogger, resume_and_get_id, check_wandb_config_file
  35. logger = logging.getLogger(__name__)
  36. def train(hyp, opt, device, tb_writer=None):
  37. logger.info(colorstr('hyperparameters: ') + ', '.join(f'{k}={v}' for k, v in hyp.items()))
  38. save_dir, epochs, batch_size, total_batch_size, weights, rank = \
  39. Path(opt.save_dir), opt.epochs, opt.batch_size, opt.total_batch_size, opt.weights, opt.global_rank
  40. # Directories
  41. wdir = save_dir / 'weights'
  42. wdir.mkdir(parents=True, exist_ok=True) # make dir
  43. last = wdir / 'last.pt'
  44. best = wdir / 'best.pt'
  45. results_file = save_dir / 'results.txt'
  46. # Save run settings
  47. with open(save_dir / 'hyp.yaml', 'w') as f:
  48. yaml.dump(hyp, f, sort_keys=False)
  49. with open(save_dir / 'opt.yaml', 'w') as f:
  50. yaml.dump(vars(opt), f, sort_keys=False)
  51. # Configure
  52. plots = not opt.evolve # create plots
  53. cuda = device.type != 'cpu'
  54. init_seeds(2 + rank)
  55. with open(opt.data) as f:
  56. data_dict = yaml.load(f, Loader=yaml.SafeLoader) # data dict
  57. is_coco = opt.data.endswith('coco.yaml')
  58. # Logging- Doing this before checking the dataset. Might update data_dict
  59. loggers = {'wandb': None} # loggers dict
  60. if rank in [-1, 0]:
  61. opt.hyp = hyp # add hyperparameters
  62. run_id = torch.load(weights).get('wandb_id') if weights.endswith('.pt') and os.path.isfile(weights) else None
  63. wandb_logger = WandbLogger(opt, Path(opt.save_dir).stem, run_id, data_dict)
  64. loggers['wandb'] = wandb_logger.wandb
  65. data_dict = wandb_logger.data_dict
  66. if wandb_logger.wandb:
  67. weights, epochs, hyp = opt.weights, opt.epochs, opt.hyp # WandbLogger might update weights, epochs if resuming
  68. nc = 1 if opt.single_cls else int(data_dict['nc']) # number of classes
  69. names = ['item'] if opt.single_cls and len(data_dict['names']) != 1 else data_dict['names'] # class names
  70. assert len(names) == nc, '%g names found for nc=%g dataset in %s' % (len(names), nc, opt.data) # check
  71. # Model
  72. pretrained = weights.endswith('.pt')
  73. if pretrained:
  74. with torch_distributed_zero_first(rank):
  75. attempt_download(weights) # download if not found locally
  76. ckpt = torch.load(weights, map_location=device) # load checkpoint
  77. model = Model(opt.cfg or ckpt['model'].yaml, ch=3, nc=nc, anchors=hyp.get('anchors')).to(device) # create
  78. exclude = ['anchor'] if (opt.cfg or hyp.get('anchors')) and not opt.resume else [] # exclude keys
  79. state_dict = ckpt['model'].float().state_dict() # to FP32
  80. state_dict = intersect_dicts(state_dict, model.state_dict(), exclude=exclude) # intersect
  81. model.load_state_dict(state_dict, strict=False) # load
  82. logger.info('Transferred %g/%g items from %s' % (len(state_dict), len(model.state_dict()), weights)) # report
  83. else:
  84. model = Model(opt.cfg, ch=3, nc=nc, anchors=hyp.get('anchors')).to(device) # create
  85. with torch_distributed_zero_first(rank):
  86. check_dataset(data_dict) # check
  87. train_path = data_dict['train']
  88. test_path = data_dict['val']
  89. # Freeze
  90. freeze = [] # parameter names to freeze (full or partial)
  91. for k, v in model.named_parameters():
  92. v.requires_grad = True # train all layers
  93. if any(x in k for x in freeze):
  94. print('freezing %s' % k)
  95. v.requires_grad = False
  96. # Optimizer
  97. nbs = 64 # nominal batch size
  98. accumulate = max(round(nbs / total_batch_size), 1) # accumulate loss before optimizing
  99. hyp['weight_decay'] *= total_batch_size * accumulate / nbs # scale weight_decay
  100. logger.info(f"Scaled weight_decay = {hyp['weight_decay']}")
  101. pg0, pg1, pg2 = [], [], [] # optimizer parameter groups
  102. for k, v in model.named_modules():
  103. if hasattr(v, 'bias') and isinstance(v.bias, nn.Parameter):
  104. pg2.append(v.bias) # biases
  105. if isinstance(v, nn.BatchNorm2d):
  106. pg0.append(v.weight) # no decay
  107. elif hasattr(v, 'weight') and isinstance(v.weight, nn.Parameter):
  108. pg1.append(v.weight) # apply decay
  109. if opt.adam:
  110. optimizer = optim.Adam(pg0, lr=hyp['lr0'], betas=(hyp['momentum'], 0.999)) # adjust beta1 to momentum
  111. else:
  112. optimizer = optim.SGD(pg0, lr=hyp['lr0'], momentum=hyp['momentum'], nesterov=True)
  113. optimizer.add_param_group({'params': pg1, 'weight_decay': hyp['weight_decay']}) # add pg1 with weight_decay
  114. optimizer.add_param_group({'params': pg2}) # add pg2 (biases)
  115. logger.info('Optimizer groups: %g .bias, %g conv.weight, %g other' % (len(pg2), len(pg1), len(pg0)))
  116. del pg0, pg1, pg2
  117. # Scheduler https://arxiv.org/pdf/1812.01187.pdf
  118. # https://pytorch.org/docs/stable/_modules/torch/optim/lr_scheduler.html#OneCycleLR
  119. if opt.linear_lr:
  120. lf = lambda x: (1 - x / (epochs - 1)) * (1.0 - hyp['lrf']) + hyp['lrf'] # linear
  121. else:
  122. lf = one_cycle(1, hyp['lrf'], epochs) # cosine 1->hyp['lrf']
  123. scheduler = lr_scheduler.LambdaLR(optimizer, lr_lambda=lf)
  124. # plot_lr_scheduler(optimizer, scheduler, epochs)
  125. # EMA
  126. ema = ModelEMA(model) if rank in [-1, 0] else None
  127. # Resume
  128. start_epoch, best_fitness = 0, 0.0
  129. if pretrained:
  130. # Optimizer
  131. if ckpt['optimizer'] is not None:
  132. optimizer.load_state_dict(ckpt['optimizer'])
  133. best_fitness = ckpt['best_fitness']
  134. # EMA
  135. if ema and ckpt.get('ema'):
  136. ema.ema.load_state_dict(ckpt['ema'].float().state_dict())
  137. ema.updates = ckpt['updates']
  138. # Results
  139. if ckpt.get('training_results') is not None:
  140. results_file.write_text(ckpt['training_results']) # write results.txt
  141. # Epochs
  142. start_epoch = ckpt['epoch'] + 1
  143. if opt.resume:
  144. assert start_epoch > 0, '%s training to %g epochs is finished, nothing to resume.' % (weights, epochs)
  145. if epochs < start_epoch:
  146. logger.info('%s has been trained for %g epochs. Fine-tuning for %g additional epochs.' %
  147. (weights, ckpt['epoch'], epochs))
  148. epochs += ckpt['epoch'] # finetune additional epochs
  149. del ckpt, state_dict
  150. # Image sizes
  151. gs = max(int(model.stride.max()), 32) # grid size (max stride)
  152. nl = model.model[-1].nl # number of detection layers (used for scaling hyp['obj'])
  153. imgsz, imgsz_test = [check_img_size(x, gs) for x in opt.img_size] # verify imgsz are gs-multiples
  154. # DP mode
  155. if cuda and rank == -1 and torch.cuda.device_count() > 1:
  156. model = torch.nn.DataParallel(model)
  157. # SyncBatchNorm
  158. if opt.sync_bn and cuda and rank != -1:
  159. model = torch.nn.SyncBatchNorm.convert_sync_batchnorm(model).to(device)
  160. logger.info('Using SyncBatchNorm()')
  161. # Trainloader
  162. dataloader, dataset = create_dataloader(train_path, imgsz, batch_size, gs, opt,
  163. hyp=hyp, augment=True, cache=opt.cache_images, rect=opt.rect, rank=rank,
  164. world_size=opt.world_size, workers=opt.workers,
  165. image_weights=opt.image_weights, quad=opt.quad, prefix=colorstr('train: '))
  166. mlc = np.concatenate(dataset.labels, 0)[:, 0].max() # max label class
  167. nb = len(dataloader) # number of batches
  168. assert mlc < nc, 'Label class %g exceeds nc=%g in %s. Possible class labels are 0-%g' % (mlc, nc, opt.data, nc - 1)
  169. # Process 0
  170. if rank in [-1, 0]:
  171. testloader = create_dataloader(test_path, imgsz_test, batch_size * 2, gs, opt, # testloader
  172. hyp=hyp, cache=opt.cache_images and not opt.notest, rect=True, rank=-1,
  173. world_size=opt.world_size, workers=opt.workers,
  174. pad=0.5, prefix=colorstr('val: '))[0]
  175. if not opt.resume:
  176. labels = np.concatenate(dataset.labels, 0)
  177. c = torch.tensor(labels[:, 0]) # classes
  178. # cf = torch.bincount(c.long(), minlength=nc) + 1. # frequency
  179. # model._initialize_biases(cf.to(device))
  180. if plots:
  181. plot_labels(labels, names, save_dir, loggers)
  182. if tb_writer:
  183. tb_writer.add_histogram('classes', c, 0)
  184. # Anchors
  185. if not opt.noautoanchor:
  186. check_anchors(dataset, model=model, thr=hyp['anchor_t'], imgsz=imgsz)
  187. model.half().float() # pre-reduce anchor precision
  188. # DDP mode
  189. if cuda and rank != -1:
  190. model = DDP(model, device_ids=[opt.local_rank], output_device=opt.local_rank)
  191. # Model parameters
  192. hyp['box'] *= 3. / nl # scale to layers
  193. hyp['cls'] *= nc / 80. * 3. / nl # scale to classes and layers
  194. hyp['obj'] *= (imgsz / 640) ** 2 * 3. / nl # scale to image size and layers
  195. model.nc = nc # attach number of classes to model
  196. model.hyp = hyp # attach hyperparameters to model
  197. model.gr = 1.0 # iou loss ratio (obj_loss = 1.0 or iou)
  198. model.class_weights = labels_to_class_weights(dataset.labels, nc).to(device) * nc # attach class weights
  199. model.names = names
  200. # Start training
  201. t0 = time.time()
  202. nw = max(round(hyp['warmup_epochs'] * nb), 1000) # number of warmup iterations, max(3 epochs, 1k iterations)
  203. # nw = min(nw, (epochs - start_epoch) / 2 * nb) # limit warmup to < 1/2 of training
  204. maps = np.zeros(nc) # mAP per class
  205. results = (0, 0, 0, 0, 0, 0, 0) # P, R, mAP@.5, mAP@.5-.95, val_loss(box, obj, cls)
  206. scheduler.last_epoch = start_epoch - 1 # do not move
  207. scaler = amp.GradScaler(enabled=cuda)
  208. compute_loss = ComputeLoss(model) # init loss class
  209. logger.info(f'Image sizes {imgsz} train, {imgsz_test} test\n'
  210. f'Using {dataloader.num_workers} dataloader workers\n'
  211. f'Logging results to {save_dir}\n'
  212. f'Starting training for {epochs} epochs...')
  213. for epoch in range(start_epoch, epochs): # epoch ------------------------------------------------------------------
  214. model.train()
  215. # Update image weights (optional)
  216. if opt.image_weights:
  217. # Generate indices
  218. if rank in [-1, 0]:
  219. cw = model.class_weights.cpu().numpy() * (1 - maps) ** 2 / nc # class weights
  220. iw = labels_to_image_weights(dataset.labels, nc=nc, class_weights=cw) # image weights
  221. dataset.indices = random.choices(range(dataset.n), weights=iw, k=dataset.n) # rand weighted idx
  222. # Broadcast if DDP
  223. if rank != -1:
  224. indices = (torch.tensor(dataset.indices) if rank == 0 else torch.zeros(dataset.n)).int()
  225. dist.broadcast(indices, 0)
  226. if rank != 0:
  227. dataset.indices = indices.cpu().numpy()
  228. # Update mosaic border
  229. # b = int(random.uniform(0.25 * imgsz, 0.75 * imgsz + gs) // gs * gs)
  230. # dataset.mosaic_border = [b - imgsz, -b] # height, width borders
  231. mloss = torch.zeros(4, device=device) # mean losses
  232. if rank != -1:
  233. dataloader.sampler.set_epoch(epoch)
  234. pbar = enumerate(dataloader)
  235. logger.info(('\n' + '%10s' * 8) % ('Epoch', 'gpu_mem', 'box', 'obj', 'cls', 'total', 'labels', 'img_size'))
  236. if rank in [-1, 0]:
  237. pbar = tqdm(pbar, total=nb) # progress bar
  238. optimizer.zero_grad()
  239. for i, (imgs, targets, paths, _) in pbar: # batch -------------------------------------------------------------
  240. ni = i + nb * epoch # number integrated batches (since train start)
  241. imgs = imgs.to(device, non_blocking=True).float() / 255.0 # uint8 to float32, 0-255 to 0.0-1.0
  242. # Warmup
  243. if ni <= nw:
  244. xi = [0, nw] # x interp
  245. # model.gr = np.interp(ni, xi, [0.0, 1.0]) # iou loss ratio (obj_loss = 1.0 or iou)
  246. accumulate = max(1, np.interp(ni, xi, [1, nbs / total_batch_size]).round())
  247. for j, x in enumerate(optimizer.param_groups):
  248. # bias lr falls from 0.1 to lr0, all other lrs rise from 0.0 to lr0
  249. x['lr'] = np.interp(ni, xi, [hyp['warmup_bias_lr'] if j == 2 else 0.0, x['initial_lr'] * lf(epoch)])
  250. if 'momentum' in x:
  251. x['momentum'] = np.interp(ni, xi, [hyp['warmup_momentum'], hyp['momentum']])
  252. # Multi-scale
  253. if opt.multi_scale:
  254. sz = random.randrange(imgsz * 0.5, imgsz * 1.5 + gs) // gs * gs # size
  255. sf = sz / max(imgs.shape[2:]) # scale factor
  256. if sf != 1:
  257. ns = [math.ceil(x * sf / gs) * gs for x in imgs.shape[2:]] # new shape (stretched to gs-multiple)
  258. imgs = F.interpolate(imgs, size=ns, mode='bilinear', align_corners=False)
  259. # Forward
  260. with amp.autocast(enabled=cuda):
  261. pred = model(imgs) # forward
  262. loss, loss_items = compute_loss(pred, targets.to(device)) # loss scaled by batch_size
  263. if rank != -1:
  264. loss *= opt.world_size # gradient averaged between devices in DDP mode
  265. if opt.quad:
  266. loss *= 4.
  267. # Backward
  268. scaler.scale(loss).backward()
  269. # Optimize
  270. if ni % accumulate == 0:
  271. scaler.step(optimizer) # optimizer.step
  272. scaler.update()
  273. optimizer.zero_grad()
  274. if ema:
  275. ema.update(model)
  276. # Print
  277. if rank in [-1, 0]:
  278. mloss = (mloss * i + loss_items) / (i + 1) # update mean losses
  279. mem = '%.3gG' % (torch.cuda.memory_reserved() / 1E9 if torch.cuda.is_available() else 0) # (GB)
  280. s = ('%10s' * 2 + '%10.4g' * 6) % (
  281. '%g/%g' % (epoch, epochs - 1), mem, *mloss, targets.shape[0], imgs.shape[-1])
  282. pbar.set_description(s)
  283. # Plot
  284. if plots and ni < 3:
  285. f = save_dir / f'train_batch{ni}.jpg' # filename
  286. Thread(target=plot_images, args=(imgs, targets, paths, f), daemon=True).start()
  287. # if tb_writer:
  288. # tb_writer.add_image(f, result, dataformats='HWC', global_step=epoch)
  289. # tb_writer.add_graph(model, imgs) # add model to tensorboard
  290. elif plots and ni == 10 and wandb_logger.wandb:
  291. wandb_logger.log({"Mosaics": [wandb_logger.wandb.Image(str(x), caption=x.name) for x in
  292. save_dir.glob('train*.jpg') if x.exists()]})
  293. # end batch ------------------------------------------------------------------------------------------------
  294. # end epoch ----------------------------------------------------------------------------------------------------
  295. # Scheduler
  296. lr = [x['lr'] for x in optimizer.param_groups] # for tensorboard
  297. scheduler.step()
  298. # DDP process 0 or single-GPU
  299. if rank in [-1, 0]:
  300. # mAP
  301. ema.update_attr(model, include=['yaml', 'nc', 'hyp', 'gr', 'names', 'stride', 'class_weights'])
  302. final_epoch = epoch + 1 == epochs
  303. if not opt.notest or final_epoch: # Calculate mAP
  304. wandb_logger.current_epoch = epoch + 1
  305. results, maps, times = test.test(data_dict,
  306. batch_size=total_batch_size,
  307. imgsz=imgsz_test,
  308. model=ema.ema,
  309. single_cls=opt.single_cls,
  310. dataloader=testloader,
  311. save_dir=save_dir,
  312. verbose=nc < 50 and final_epoch,
  313. plots=plots and final_epoch,
  314. wandb_logger=wandb_logger,
  315. compute_loss=compute_loss,
  316. is_coco=is_coco)
  317. # Write
  318. with open(results_file, 'a') as f:
  319. f.write(s + '%10.4g' * 7 % results + '\n') # append metrics, val_loss
  320. if len(opt.name) and opt.bucket:
  321. os.system('gsutil cp %s gs://%s/results/results%s.txt' % (results_file, opt.bucket, opt.name))
  322. # Log
  323. tags = ['train/box_loss', 'train/obj_loss', 'train/cls_loss', # train loss
  324. 'metrics/precision', 'metrics/recall', 'metrics/mAP_0.5', 'metrics/mAP_0.5:0.95',
  325. 'val/box_loss', 'val/obj_loss', 'val/cls_loss', # val loss
  326. 'x/lr0', 'x/lr1', 'x/lr2'] # params
  327. for x, tag in zip(list(mloss[:-1]) + list(results) + lr, tags):
  328. if tb_writer:
  329. tb_writer.add_scalar(tag, x, epoch) # tensorboard
  330. if wandb_logger.wandb:
  331. wandb_logger.log({tag: x}) # W&B
  332. # Update best mAP
  333. fi = fitness(np.array(results).reshape(1, -1)) # weighted combination of [P, R, mAP@.5, mAP@.5-.95]
  334. if fi > best_fitness:
  335. best_fitness = fi
  336. wandb_logger.end_epoch(best_result=best_fitness == fi)
  337. # Save model
  338. if (not opt.nosave) or (final_epoch and not opt.evolve): # if save
  339. ckpt = {'epoch': epoch,
  340. 'best_fitness': best_fitness,
  341. 'training_results': results_file.read_text(),
  342. 'model': deepcopy(model.module if is_parallel(model) else model).half(),
  343. 'ema': deepcopy(ema.ema).half(),
  344. 'updates': ema.updates,
  345. 'optimizer': optimizer.state_dict(),
  346. 'wandb_id': wandb_logger.wandb_run.id if wandb_logger.wandb else None}
  347. # Save last, best and delete
  348. torch.save(ckpt, last)
  349. if best_fitness == fi:
  350. torch.save(ckpt, best)
  351. if wandb_logger.wandb:
  352. if ((epoch + 1) % opt.save_period == 0 and not final_epoch) and opt.save_period != -1:
  353. wandb_logger.log_model(
  354. last.parent, opt, epoch, fi, best_model=best_fitness == fi)
  355. del ckpt
  356. # end epoch ----------------------------------------------------------------------------------------------------
  357. # end training
  358. if rank in [-1, 0]:
  359. # Plots
  360. if plots:
  361. plot_results(save_dir=save_dir) # save as results.png
  362. if wandb_logger.wandb:
  363. files = ['results.png', 'confusion_matrix.png', *[f'{x}_curve.png' for x in ('F1', 'PR', 'P', 'R')]]
  364. wandb_logger.log({"Results": [wandb_logger.wandb.Image(str(save_dir / f), caption=f) for f in files
  365. if (save_dir / f).exists()]})
  366. # Test best.pt
  367. logger.info('%g epochs completed in %.3f hours.\n' % (epoch - start_epoch + 1, (time.time() - t0) / 3600))
  368. if opt.data.endswith('coco.yaml') and nc == 80: # if COCO
  369. for m in (last, best) if best.exists() else (last): # speed, mAP tests
  370. results, _, _ = test.test(opt.data,
  371. batch_size=batch_size * 2,
  372. imgsz=imgsz_test,
  373. conf_thres=0.001,
  374. iou_thres=0.7,
  375. model=attempt_load(m, device).half(),
  376. single_cls=opt.single_cls,
  377. dataloader=testloader,
  378. save_dir=save_dir,
  379. save_json=True,
  380. plots=False,
  381. is_coco=is_coco)
  382. # Strip optimizers
  383. final = best if best.exists() else last # final model
  384. for f in last, best:
  385. if f.exists():
  386. strip_optimizer(f) # strip optimizers
  387. if opt.bucket:
  388. os.system(f'gsutil cp {final} gs://{opt.bucket}/weights') # upload
  389. if wandb_logger.wandb: # Log the stripped model
  390. wandb_logger.wandb.log_artifact(str(final), type='model',
  391. name='run_' + wandb_logger.wandb_run.id + '_model',
  392. aliases=['last', 'best', 'stripped'])
  393. wandb_logger.finish_run()
  394. else:
  395. dist.destroy_process_group()
  396. torch.cuda.empty_cache()
  397. return results
  398. if __name__ == '__main__':
  399. parser = argparse.ArgumentParser()
  400. parser.add_argument('--weights', type=str, default='yolov5s.pt', help='initial weights path')
  401. parser.add_argument('--cfg', type=str, default='', help='model.yaml path')
  402. parser.add_argument('--data', type=str, default='data/coco128.yaml', help='data.yaml path')
  403. parser.add_argument('--hyp', type=str, default='data/hyp.scratch.yaml', help='hyperparameters path')
  404. parser.add_argument('--epochs', type=int, default=300)
  405. parser.add_argument('--batch-size', type=int, default=16, help='total batch size for all GPUs')
  406. parser.add_argument('--img-size', nargs='+', type=int, default=[640, 640], help='[train, test] image sizes')
  407. parser.add_argument('--rect', action='store_true', help='rectangular training')
  408. parser.add_argument('--resume', nargs='?', const=True, default=False, help='resume most recent training')
  409. parser.add_argument('--nosave', action='store_true', help='only save final checkpoint')
  410. parser.add_argument('--notest', action='store_true', help='only test final epoch')
  411. parser.add_argument('--noautoanchor', action='store_true', help='disable autoanchor check')
  412. parser.add_argument('--evolve', action='store_true', help='evolve hyperparameters')
  413. parser.add_argument('--bucket', type=str, default='', help='gsutil bucket')
  414. parser.add_argument('--cache-images', action='store_true', help='cache images for faster training')
  415. parser.add_argument('--image-weights', action='store_true', help='use weighted image selection for training')
  416. parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
  417. parser.add_argument('--multi-scale', action='store_true', help='vary img-size +/- 50%%')
  418. parser.add_argument('--single-cls', action='store_true', help='train multi-class data as single-class')
  419. parser.add_argument('--adam', action='store_true', help='use torch.optim.Adam() optimizer')
  420. parser.add_argument('--sync-bn', action='store_true', help='use SyncBatchNorm, only available in DDP mode')
  421. parser.add_argument('--local_rank', type=int, default=-1, help='DDP parameter, do not modify')
  422. parser.add_argument('--workers', type=int, default=8, help='maximum number of dataloader workers')
  423. parser.add_argument('--project', default='runs/train', help='save to project/name')
  424. parser.add_argument('--entity', default=None, help='W&B entity')
  425. parser.add_argument('--name', default='exp', help='save to project/name')
  426. parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment')
  427. parser.add_argument('--quad', action='store_true', help='quad dataloader')
  428. parser.add_argument('--linear-lr', action='store_true', help='linear LR')
  429. parser.add_argument('--upload_dataset', action='store_true', help='Upload dataset as W&B artifact table')
  430. parser.add_argument('--bbox_interval', type=int, default=-1, help='Set bounding-box image logging interval for W&B')
  431. parser.add_argument('--save_period', type=int, default=-1, help='Log model after every "save_period" epoch')
  432. parser.add_argument('--artifact_alias', type=str, default="latest", help='version of dataset artifact to be used')
  433. opt = parser.parse_args()
  434. # Set DDP variables
  435. opt.world_size = int(os.environ['WORLD_SIZE']) if 'WORLD_SIZE' in os.environ else 1
  436. opt.global_rank = int(os.environ['RANK']) if 'RANK' in os.environ else -1
  437. set_logging(opt.global_rank)
  438. if opt.global_rank in [-1, 0]:
  439. check_git_status()
  440. check_requirements()
  441. # Resume
  442. wandb_run = resume_and_get_id(opt)
  443. if opt.resume and not wandb_run: # resume an interrupted run
  444. ckpt = opt.resume if isinstance(opt.resume, str) else get_latest_run() # specified or most recent path
  445. assert os.path.isfile(ckpt), 'ERROR: --resume checkpoint does not exist'
  446. apriori = opt.global_rank, opt.local_rank
  447. with open(Path(ckpt).parent.parent / 'opt.yaml') as f:
  448. opt = argparse.Namespace(**yaml.load(f, Loader=yaml.SafeLoader)) # replace
  449. opt.cfg, opt.weights, opt.resume, opt.batch_size, opt.global_rank, opt.local_rank = '', ckpt, True, opt.total_batch_size, *apriori # reinstate
  450. logger.info('Resuming training from %s' % ckpt)
  451. else:
  452. # opt.hyp = opt.hyp or ('hyp.finetune.yaml' if opt.weights else 'hyp.scratch.yaml')
  453. opt.data, opt.cfg, opt.hyp = check_file(opt.data), check_file(opt.cfg), check_file(opt.hyp) # check files
  454. assert len(opt.cfg) or len(opt.weights), 'either --cfg or --weights must be specified'
  455. opt.img_size.extend([opt.img_size[-1]] * (2 - len(opt.img_size))) # extend to 2 sizes (train, test)
  456. opt.name = 'evolve' if opt.evolve else opt.name
  457. opt.save_dir = increment_path(Path(opt.project) / opt.name, exist_ok=opt.exist_ok | opt.evolve) # increment run
  458. # DDP mode
  459. opt.total_batch_size = opt.batch_size
  460. device = select_device(opt.device, batch_size=opt.batch_size)
  461. if opt.local_rank != -1:
  462. assert torch.cuda.device_count() > opt.local_rank
  463. torch.cuda.set_device(opt.local_rank)
  464. device = torch.device('cuda', opt.local_rank)
  465. dist.init_process_group(backend='nccl', init_method='env://') # distributed backend
  466. assert opt.batch_size % opt.world_size == 0, '--batch-size must be multiple of CUDA device count'
  467. opt.batch_size = opt.total_batch_size // opt.world_size
  468. # Hyperparameters
  469. with open(opt.hyp) as f:
  470. hyp = yaml.load(f, Loader=yaml.SafeLoader) # load hyps
  471. # Train
  472. logger.info(opt)
  473. if not opt.evolve:
  474. tb_writer = None # init loggers
  475. if opt.global_rank in [-1, 0]:
  476. logger.info(f'Start Tensorboard with "tensorboard --logdir {opt.project}", view at http://localhost:6006/')
  477. tb_writer = SummaryWriter(opt.save_dir) # Tensorboard
  478. train(hyp, opt, device, tb_writer)
  479. # Evolve hyperparameters (optional)
  480. else:
  481. # Hyperparameter evolution metadata (mutation scale 0-1, lower_limit, upper_limit)
  482. meta = {'lr0': (1, 1e-5, 1e-1), # initial learning rate (SGD=1E-2, Adam=1E-3)
  483. 'lrf': (1, 0.01, 1.0), # final OneCycleLR learning rate (lr0 * lrf)
  484. 'momentum': (0.3, 0.6, 0.98), # SGD momentum/Adam beta1
  485. 'weight_decay': (1, 0.0, 0.001), # optimizer weight decay
  486. 'warmup_epochs': (1, 0.0, 5.0), # warmup epochs (fractions ok)
  487. 'warmup_momentum': (1, 0.0, 0.95), # warmup initial momentum
  488. 'warmup_bias_lr': (1, 0.0, 0.2), # warmup initial bias lr
  489. 'box': (1, 0.02, 0.2), # box loss gain
  490. 'cls': (1, 0.2, 4.0), # cls loss gain
  491. 'cls_pw': (1, 0.5, 2.0), # cls BCELoss positive_weight
  492. 'obj': (1, 0.2, 4.0), # obj loss gain (scale with pixels)
  493. 'obj_pw': (1, 0.5, 2.0), # obj BCELoss positive_weight
  494. 'iou_t': (0, 0.1, 0.7), # IoU training threshold
  495. 'anchor_t': (1, 2.0, 8.0), # anchor-multiple threshold
  496. 'anchors': (2, 2.0, 10.0), # anchors per output grid (0 to ignore)
  497. 'fl_gamma': (0, 0.0, 2.0), # focal loss gamma (efficientDet default gamma=1.5)
  498. 'hsv_h': (1, 0.0, 0.1), # image HSV-Hue augmentation (fraction)
  499. 'hsv_s': (1, 0.0, 0.9), # image HSV-Saturation augmentation (fraction)
  500. 'hsv_v': (1, 0.0, 0.9), # image HSV-Value augmentation (fraction)
  501. 'degrees': (1, 0.0, 45.0), # image rotation (+/- deg)
  502. 'translate': (1, 0.0, 0.9), # image translation (+/- fraction)
  503. 'scale': (1, 0.0, 0.9), # image scale (+/- gain)
  504. 'shear': (1, 0.0, 10.0), # image shear (+/- deg)
  505. 'perspective': (0, 0.0, 0.001), # image perspective (+/- fraction), range 0-0.001
  506. 'flipud': (1, 0.0, 1.0), # image flip up-down (probability)
  507. 'fliplr': (0, 0.0, 1.0), # image flip left-right (probability)
  508. 'mosaic': (1, 0.0, 1.0), # image mixup (probability)
  509. 'mixup': (1, 0.0, 1.0)} # image mixup (probability)
  510. assert opt.local_rank == -1, 'DDP mode not implemented for --evolve'
  511. opt.notest, opt.nosave = True, True # only test/save final epoch
  512. # ei = [isinstance(x, (int, float)) for x in hyp.values()] # evolvable indices
  513. yaml_file = Path(opt.save_dir) / 'hyp_evolved.yaml' # save best result here
  514. if opt.bucket:
  515. os.system('gsutil cp gs://%s/evolve.txt .' % opt.bucket) # download evolve.txt if exists
  516. for _ in range(300): # generations to evolve
  517. if Path('evolve.txt').exists(): # if evolve.txt exists: select best hyps and mutate
  518. # Select parent(s)
  519. parent = 'single' # parent selection method: 'single' or 'weighted'
  520. x = np.loadtxt('evolve.txt', ndmin=2)
  521. n = min(5, len(x)) # number of previous results to consider
  522. x = x[np.argsort(-fitness(x))][:n] # top n mutations
  523. w = fitness(x) - fitness(x).min() # weights
  524. if parent == 'single' or len(x) == 1:
  525. # x = x[random.randint(0, n - 1)] # random selection
  526. x = x[random.choices(range(n), weights=w)[0]] # weighted selection
  527. elif parent == 'weighted':
  528. x = (x * w.reshape(n, 1)).sum(0) / w.sum() # weighted combination
  529. # Mutate
  530. mp, s = 0.8, 0.2 # mutation probability, sigma
  531. npr = np.random
  532. npr.seed(int(time.time()))
  533. g = np.array([x[0] for x in meta.values()]) # gains 0-1
  534. ng = len(meta)
  535. v = np.ones(ng)
  536. while all(v == 1): # mutate until a change occurs (prevent duplicates)
  537. v = (g * (npr.random(ng) < mp) * npr.randn(ng) * npr.random() * s + 1).clip(0.3, 3.0)
  538. for i, k in enumerate(hyp.keys()): # plt.hist(v.ravel(), 300)
  539. hyp[k] = float(x[i + 7] * v[i]) # mutate
  540. # Constrain to limits
  541. for k, v in meta.items():
  542. hyp[k] = max(hyp[k], v[1]) # lower limit
  543. hyp[k] = min(hyp[k], v[2]) # upper limit
  544. hyp[k] = round(hyp[k], 5) # significant digits
  545. # Train mutation
  546. results = train(hyp.copy(), opt, device)
  547. # Write mutation results
  548. print_mutation(hyp.copy(), results, yaml_file, opt.bucket)
  549. # Plot results
  550. plot_evolution(yaml_file)
  551. print(f'Hyperparameter evolution complete. Best results saved as: {yaml_file}\n'
  552. f'Command to train a new model with these hyperparameters: $ python train.py --hyp {yaml_file}')