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.

624 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, de_parallel
  34. from utils.wandb_logging.wandb_utils import WandbLogger, check_wandb_resume
  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.safe_dump(hyp, f, sort_keys=False)
  49. with open(save_dir / 'opt.yaml', 'w') as f:
  50. yaml.safe_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.safe_load(f) # data dict
  57. # Logging- Doing this before checking the dataset. Might update data_dict
  58. loggers = {'wandb': None} # loggers dict
  59. if rank in [-1, 0]:
  60. opt.hyp = hyp # add hyperparameters
  61. run_id = torch.load(weights).get('wandb_id') if weights.endswith('.pt') and os.path.isfile(weights) else None
  62. wandb_logger = WandbLogger(opt, save_dir.stem, run_id, data_dict)
  63. loggers['wandb'] = wandb_logger.wandb
  64. data_dict = wandb_logger.data_dict
  65. if wandb_logger.wandb:
  66. weights, epochs, hyp = opt.weights, opt.epochs, opt.hyp # WandbLogger might update weights, epochs if resuming
  67. nc = 1 if opt.single_cls else int(data_dict['nc']) # number of classes
  68. names = ['item'] if opt.single_cls and len(data_dict['names']) != 1 else data_dict['names'] # class names
  69. assert len(names) == nc, '%g names found for nc=%g dataset in %s' % (len(names), nc, opt.data) # check
  70. is_coco = opt.data.endswith('coco.yaml') and nc == 80 # COCO dataset
  71. # Model
  72. pretrained = weights.endswith('.pt')
  73. if pretrained:
  74. with torch_distributed_zero_first(rank):
  75. weights = 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. # nn.MultiheadAttention incompatibility with DDP https://github.com/pytorch/pytorch/issues/26698
  192. find_unused_parameters=any(isinstance(layer, nn.MultiheadAttention) for layer in model.modules()))
  193. # Model parameters
  194. hyp['box'] *= 3. / nl # scale to layers
  195. hyp['cls'] *= nc / 80. * 3. / nl # scale to classes and layers
  196. hyp['obj'] *= (imgsz / 640) ** 2 * 3. / nl # scale to image size and layers
  197. hyp['label_smoothing'] = opt.label_smoothing
  198. model.nc = nc # attach number of classes to model
  199. model.hyp = hyp # attach hyperparameters to model
  200. model.gr = 1.0 # iou loss ratio (obj_loss = 1.0 or iou)
  201. model.class_weights = labels_to_class_weights(dataset.labels, nc).to(device) * nc # attach class weights
  202. model.names = names
  203. # Start training
  204. t0 = time.time()
  205. nw = max(round(hyp['warmup_epochs'] * nb), 1000) # number of warmup iterations, max(3 epochs, 1k iterations)
  206. # nw = min(nw, (epochs - start_epoch) / 2 * nb) # limit warmup to < 1/2 of training
  207. maps = np.zeros(nc) # mAP per class
  208. results = (0, 0, 0, 0, 0, 0, 0) # P, R, mAP@.5, mAP@.5-.95, val_loss(box, obj, cls)
  209. scheduler.last_epoch = start_epoch - 1 # do not move
  210. scaler = amp.GradScaler(enabled=cuda)
  211. compute_loss = ComputeLoss(model) # init loss class
  212. logger.info(f'Image sizes {imgsz} train, {imgsz_test} test\n'
  213. f'Using {dataloader.num_workers} dataloader workers\n'
  214. f'Logging results to {save_dir}\n'
  215. f'Starting training for {epochs} epochs...')
  216. for epoch in range(start_epoch, epochs): # epoch ------------------------------------------------------------------
  217. model.train()
  218. # Update image weights (optional)
  219. if opt.image_weights:
  220. # Generate indices
  221. if rank in [-1, 0]:
  222. cw = model.class_weights.cpu().numpy() * (1 - maps) ** 2 / nc # class weights
  223. iw = labels_to_image_weights(dataset.labels, nc=nc, class_weights=cw) # image weights
  224. dataset.indices = random.choices(range(dataset.n), weights=iw, k=dataset.n) # rand weighted idx
  225. # Broadcast if DDP
  226. if rank != -1:
  227. indices = (torch.tensor(dataset.indices) if rank == 0 else torch.zeros(dataset.n)).int()
  228. dist.broadcast(indices, 0)
  229. if rank != 0:
  230. dataset.indices = indices.cpu().numpy()
  231. # Update mosaic border
  232. # b = int(random.uniform(0.25 * imgsz, 0.75 * imgsz + gs) // gs * gs)
  233. # dataset.mosaic_border = [b - imgsz, -b] # height, width borders
  234. mloss = torch.zeros(4, device=device) # mean losses
  235. if rank != -1:
  236. dataloader.sampler.set_epoch(epoch)
  237. pbar = enumerate(dataloader)
  238. logger.info(('\n' + '%10s' * 8) % ('Epoch', 'gpu_mem', 'box', 'obj', 'cls', 'total', 'labels', 'img_size'))
  239. if rank in [-1, 0]:
  240. pbar = tqdm(pbar, total=nb) # progress bar
  241. optimizer.zero_grad()
  242. for i, (imgs, targets, paths, _) in pbar: # batch -------------------------------------------------------------
  243. ni = i + nb * epoch # number integrated batches (since train start)
  244. imgs = imgs.to(device, non_blocking=True).float() / 255.0 # uint8 to float32, 0-255 to 0.0-1.0
  245. # Warmup
  246. if ni <= nw:
  247. xi = [0, nw] # x interp
  248. # model.gr = np.interp(ni, xi, [0.0, 1.0]) # iou loss ratio (obj_loss = 1.0 or iou)
  249. accumulate = max(1, np.interp(ni, xi, [1, nbs / total_batch_size]).round())
  250. for j, x in enumerate(optimizer.param_groups):
  251. # bias lr falls from 0.1 to lr0, all other lrs rise from 0.0 to lr0
  252. x['lr'] = np.interp(ni, xi, [hyp['warmup_bias_lr'] if j == 2 else 0.0, x['initial_lr'] * lf(epoch)])
  253. if 'momentum' in x:
  254. x['momentum'] = np.interp(ni, xi, [hyp['warmup_momentum'], hyp['momentum']])
  255. # Multi-scale
  256. if opt.multi_scale:
  257. sz = random.randrange(imgsz * 0.5, imgsz * 1.5 + gs) // gs * gs # size
  258. sf = sz / max(imgs.shape[2:]) # scale factor
  259. if sf != 1:
  260. ns = [math.ceil(x * sf / gs) * gs for x in imgs.shape[2:]] # new shape (stretched to gs-multiple)
  261. imgs = F.interpolate(imgs, size=ns, mode='bilinear', align_corners=False)
  262. # Forward
  263. with amp.autocast(enabled=cuda):
  264. pred = model(imgs) # forward
  265. loss, loss_items = compute_loss(pred, targets.to(device)) # loss scaled by batch_size
  266. if rank != -1:
  267. loss *= opt.world_size # gradient averaged between devices in DDP mode
  268. if opt.quad:
  269. loss *= 4.
  270. # Backward
  271. scaler.scale(loss).backward()
  272. # Optimize
  273. if ni % accumulate == 0:
  274. scaler.step(optimizer) # optimizer.step
  275. scaler.update()
  276. optimizer.zero_grad()
  277. if ema:
  278. ema.update(model)
  279. # Print
  280. if rank in [-1, 0]:
  281. mloss = (mloss * i + loss_items) / (i + 1) # update mean losses
  282. mem = '%.3gG' % (torch.cuda.memory_reserved() / 1E9 if torch.cuda.is_available() else 0) # (GB)
  283. s = ('%10s' * 2 + '%10.4g' * 6) % (
  284. '%g/%g' % (epoch, epochs - 1), mem, *mloss, targets.shape[0], imgs.shape[-1])
  285. pbar.set_description(s)
  286. # Plot
  287. if plots and ni < 3:
  288. f = save_dir / f'train_batch{ni}.jpg' # filename
  289. Thread(target=plot_images, args=(imgs, targets, paths, f), daemon=True).start()
  290. if tb_writer:
  291. tb_writer.add_graph(torch.jit.trace(de_parallel(model), imgs, strict=False), []) # model graph
  292. # tb_writer.add_image(f, result, dataformats='HWC', global_step=epoch)
  293. elif plots and ni == 10 and wandb_logger.wandb:
  294. wandb_logger.log({"Mosaics": [wandb_logger.wandb.Image(str(x), caption=x.name) for x in
  295. save_dir.glob('train*.jpg') if x.exists()]})
  296. # end batch ------------------------------------------------------------------------------------------------
  297. # end epoch ----------------------------------------------------------------------------------------------------
  298. # Scheduler
  299. lr = [x['lr'] for x in optimizer.param_groups] # for tensorboard
  300. scheduler.step()
  301. # DDP process 0 or single-GPU
  302. if rank in [-1, 0]:
  303. # mAP
  304. ema.update_attr(model, include=['yaml', 'nc', 'hyp', 'gr', 'names', 'stride', 'class_weights'])
  305. final_epoch = epoch + 1 == epochs
  306. if not opt.notest or final_epoch: # Calculate mAP
  307. wandb_logger.current_epoch = epoch + 1
  308. results, maps, times = test.test(data_dict,
  309. batch_size=batch_size * 2,
  310. imgsz=imgsz_test,
  311. model=ema.ema,
  312. single_cls=opt.single_cls,
  313. dataloader=testloader,
  314. save_dir=save_dir,
  315. save_json=is_coco and final_epoch,
  316. verbose=nc < 50 and final_epoch,
  317. plots=plots and final_epoch,
  318. wandb_logger=wandb_logger,
  319. compute_loss=compute_loss,
  320. is_coco=is_coco)
  321. # Write
  322. with open(results_file, 'a') as f:
  323. f.write(s + '%10.4g' * 7 % results + '\n') # append metrics, val_loss
  324. # Log
  325. tags = ['train/box_loss', 'train/obj_loss', 'train/cls_loss', # train loss
  326. 'metrics/precision', 'metrics/recall', 'metrics/mAP_0.5', 'metrics/mAP_0.5:0.95',
  327. 'val/box_loss', 'val/obj_loss', 'val/cls_loss', # val loss
  328. 'x/lr0', 'x/lr1', 'x/lr2'] # params
  329. for x, tag in zip(list(mloss[:-1]) + list(results) + lr, tags):
  330. if tb_writer:
  331. tb_writer.add_scalar(tag, x, epoch) # tensorboard
  332. if wandb_logger.wandb:
  333. wandb_logger.log({tag: x}) # W&B
  334. # Update best mAP
  335. fi = fitness(np.array(results).reshape(1, -1)) # weighted combination of [P, R, mAP@.5, mAP@.5-.95]
  336. if fi > best_fitness:
  337. best_fitness = fi
  338. wandb_logger.end_epoch(best_result=best_fitness == fi)
  339. # Save model
  340. if (not opt.nosave) or (final_epoch and not opt.evolve): # if save
  341. ckpt = {'epoch': epoch,
  342. 'best_fitness': best_fitness,
  343. 'training_results': results_file.read_text(),
  344. 'model': deepcopy(de_parallel(model)).half(),
  345. 'ema': deepcopy(ema.ema).half(),
  346. 'updates': ema.updates,
  347. 'optimizer': optimizer.state_dict(),
  348. 'wandb_id': wandb_logger.wandb_run.id if wandb_logger.wandb else None}
  349. # Save last, best and delete
  350. torch.save(ckpt, last)
  351. if best_fitness == fi:
  352. torch.save(ckpt, best)
  353. if wandb_logger.wandb:
  354. if ((epoch + 1) % opt.save_period == 0 and not final_epoch) and opt.save_period != -1:
  355. wandb_logger.log_model(
  356. last.parent, opt, epoch, fi, best_model=best_fitness == fi)
  357. del ckpt
  358. # end epoch ----------------------------------------------------------------------------------------------------
  359. # end training
  360. if rank in [-1, 0]:
  361. logger.info(f'{epoch - start_epoch + 1} epochs completed in {(time.time() - t0) / 3600:.3f} hours.\n')
  362. if plots:
  363. plot_results(save_dir=save_dir) # save as results.png
  364. if wandb_logger.wandb:
  365. files = ['results.png', 'confusion_matrix.png', *[f'{x}_curve.png' for x in ('F1', 'PR', 'P', 'R')]]
  366. wandb_logger.log({"Results": [wandb_logger.wandb.Image(str(save_dir / f), caption=f) for f in files
  367. if (save_dir / f).exists()]})
  368. if not opt.evolve:
  369. if is_coco: # COCO dataset
  370. for m in [last, best] if best.exists() else [last]: # speed, mAP tests
  371. results, _, _ = test.test(opt.data,
  372. batch_size=batch_size * 2,
  373. imgsz=imgsz_test,
  374. conf_thres=0.001,
  375. iou_thres=0.7,
  376. model=attempt_load(m, device).half(),
  377. single_cls=opt.single_cls,
  378. dataloader=testloader,
  379. save_dir=save_dir,
  380. save_json=True,
  381. plots=False,
  382. is_coco=is_coco)
  383. # Strip optimizers
  384. for f in last, best:
  385. if f.exists():
  386. strip_optimizer(f) # strip optimizers
  387. if wandb_logger.wandb: # Log the stripped model
  388. wandb_logger.wandb.log_artifact(str(best if best.exists() else last), type='model',
  389. name='run_' + wandb_logger.wandb_run.id + '_model',
  390. aliases=['latest', 'best', 'stripped'])
  391. wandb_logger.finish_run()
  392. else:
  393. dist.destroy_process_group()
  394. torch.cuda.empty_cache()
  395. return results
  396. if __name__ == '__main__':
  397. parser = argparse.ArgumentParser()
  398. parser.add_argument('--weights', type=str, default='yolov5s.pt', help='initial weights path')
  399. parser.add_argument('--cfg', type=str, default='', help='model.yaml path')
  400. parser.add_argument('--data', type=str, default='data/coco128.yaml', help='data.yaml path')
  401. parser.add_argument('--hyp', type=str, default='data/hyp.scratch.yaml', help='hyperparameters path')
  402. parser.add_argument('--epochs', type=int, default=300)
  403. parser.add_argument('--batch-size', type=int, default=16, help='total batch size for all GPUs')
  404. parser.add_argument('--img-size', nargs='+', type=int, default=[640, 640], help='[train, test] image sizes')
  405. parser.add_argument('--rect', action='store_true', help='rectangular training')
  406. parser.add_argument('--resume', nargs='?', const=True, default=False, help='resume most recent training')
  407. parser.add_argument('--nosave', action='store_true', help='only save final checkpoint')
  408. parser.add_argument('--notest', action='store_true', help='only test final epoch')
  409. parser.add_argument('--noautoanchor', action='store_true', help='disable autoanchor check')
  410. parser.add_argument('--evolve', action='store_true', help='evolve hyperparameters')
  411. parser.add_argument('--bucket', type=str, default='', help='gsutil bucket')
  412. parser.add_argument('--cache-images', action='store_true', help='cache images for faster training')
  413. parser.add_argument('--image-weights', action='store_true', help='use weighted image selection for training')
  414. parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
  415. parser.add_argument('--multi-scale', action='store_true', help='vary img-size +/- 50%%')
  416. parser.add_argument('--single-cls', action='store_true', help='train multi-class data as single-class')
  417. parser.add_argument('--adam', action='store_true', help='use torch.optim.Adam() optimizer')
  418. parser.add_argument('--sync-bn', action='store_true', help='use SyncBatchNorm, only available in DDP mode')
  419. parser.add_argument('--local_rank', type=int, default=-1, help='DDP parameter, do not modify')
  420. parser.add_argument('--workers', type=int, default=8, help='maximum number of dataloader workers')
  421. parser.add_argument('--project', default='runs/train', help='save to project/name')
  422. parser.add_argument('--entity', default=None, help='W&B entity')
  423. parser.add_argument('--name', default='exp', help='save to project/name')
  424. parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment')
  425. parser.add_argument('--quad', action='store_true', help='quad dataloader')
  426. parser.add_argument('--linear-lr', action='store_true', help='linear LR')
  427. parser.add_argument('--label-smoothing', type=float, default=0.0, help='Label smoothing epsilon')
  428. parser.add_argument('--upload_dataset', action='store_true', help='Upload dataset as W&B artifact table')
  429. parser.add_argument('--bbox_interval', type=int, default=-1, help='Set bounding-box image logging interval for W&B')
  430. parser.add_argument('--save_period', type=int, default=-1, help='Log model after every "save_period" epoch')
  431. parser.add_argument('--artifact_alias', type=str, default="latest", help='version of dataset artifact to be used')
  432. opt = parser.parse_args()
  433. # Set DDP variables
  434. opt.world_size = int(os.environ['WORLD_SIZE']) if 'WORLD_SIZE' in os.environ else 1
  435. opt.global_rank = int(os.environ['RANK']) if 'RANK' in os.environ else -1
  436. set_logging(opt.global_rank)
  437. if opt.global_rank in [-1, 0]:
  438. check_git_status()
  439. check_requirements(exclude=('pycocotools', 'thop'))
  440. # Resume
  441. wandb_run = check_wandb_resume(opt)
  442. if opt.resume and not wandb_run: # resume an interrupted run
  443. ckpt = opt.resume if isinstance(opt.resume, str) else get_latest_run() # specified or most recent path
  444. assert os.path.isfile(ckpt), 'ERROR: --resume checkpoint does not exist'
  445. apriori = opt.global_rank, opt.local_rank
  446. with open(Path(ckpt).parent.parent / 'opt.yaml') as f:
  447. opt = argparse.Namespace(**yaml.safe_load(f)) # replace
  448. opt.cfg, opt.weights, opt.resume, opt.batch_size, opt.global_rank, opt.local_rank = \
  449. '', 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 = str(increment_path(Path(opt.project) / opt.name, exist_ok=opt.exist_ok | opt.evolve))
  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. assert not opt.image_weights, '--image-weights argument is not compatible with DDP training'
  468. opt.batch_size = opt.total_batch_size // opt.world_size
  469. # Hyperparameters
  470. with open(opt.hyp) as f:
  471. hyp = yaml.safe_load(f) # load hyps
  472. # Train
  473. logger.info(opt)
  474. if not opt.evolve:
  475. tb_writer = None # init loggers
  476. if opt.global_rank in [-1, 0]:
  477. prefix = colorstr('tensorboard: ')
  478. logger.info(f"{prefix}Start with 'tensorboard --logdir {opt.project}', view at http://localhost:6006/")
  479. tb_writer = SummaryWriter(opt.save_dir) # Tensorboard
  480. train(hyp, opt, device, tb_writer)
  481. # Evolve hyperparameters (optional)
  482. else:
  483. # Hyperparameter evolution metadata (mutation scale 0-1, lower_limit, upper_limit)
  484. meta = {'lr0': (1, 1e-5, 1e-1), # initial learning rate (SGD=1E-2, Adam=1E-3)
  485. 'lrf': (1, 0.01, 1.0), # final OneCycleLR learning rate (lr0 * lrf)
  486. 'momentum': (0.3, 0.6, 0.98), # SGD momentum/Adam beta1
  487. 'weight_decay': (1, 0.0, 0.001), # optimizer weight decay
  488. 'warmup_epochs': (1, 0.0, 5.0), # warmup epochs (fractions ok)
  489. 'warmup_momentum': (1, 0.0, 0.95), # warmup initial momentum
  490. 'warmup_bias_lr': (1, 0.0, 0.2), # warmup initial bias lr
  491. 'box': (1, 0.02, 0.2), # box loss gain
  492. 'cls': (1, 0.2, 4.0), # cls loss gain
  493. 'cls_pw': (1, 0.5, 2.0), # cls BCELoss positive_weight
  494. 'obj': (1, 0.2, 4.0), # obj loss gain (scale with pixels)
  495. 'obj_pw': (1, 0.5, 2.0), # obj BCELoss positive_weight
  496. 'iou_t': (0, 0.1, 0.7), # IoU training threshold
  497. 'anchor_t': (1, 2.0, 8.0), # anchor-multiple threshold
  498. 'anchors': (2, 2.0, 10.0), # anchors per output grid (0 to ignore)
  499. 'fl_gamma': (0, 0.0, 2.0), # focal loss gamma (efficientDet default gamma=1.5)
  500. 'hsv_h': (1, 0.0, 0.1), # image HSV-Hue augmentation (fraction)
  501. 'hsv_s': (1, 0.0, 0.9), # image HSV-Saturation augmentation (fraction)
  502. 'hsv_v': (1, 0.0, 0.9), # image HSV-Value augmentation (fraction)
  503. 'degrees': (1, 0.0, 45.0), # image rotation (+/- deg)
  504. 'translate': (1, 0.0, 0.9), # image translation (+/- fraction)
  505. 'scale': (1, 0.0, 0.9), # image scale (+/- gain)
  506. 'shear': (1, 0.0, 10.0), # image shear (+/- deg)
  507. 'perspective': (0, 0.0, 0.001), # image perspective (+/- fraction), range 0-0.001
  508. 'flipud': (1, 0.0, 1.0), # image flip up-down (probability)
  509. 'fliplr': (0, 0.0, 1.0), # image flip left-right (probability)
  510. 'mosaic': (1, 0.0, 1.0), # image mixup (probability)
  511. 'mixup': (1, 0.0, 1.0)} # image mixup (probability)
  512. assert opt.local_rank == -1, 'DDP mode not implemented for --evolve'
  513. opt.notest, opt.nosave = True, True # only test/save final epoch
  514. # ei = [isinstance(x, (int, float)) for x in hyp.values()] # evolvable indices
  515. yaml_file = Path(opt.save_dir) / 'hyp_evolved.yaml' # save best result here
  516. if opt.bucket:
  517. os.system('gsutil cp gs://%s/evolve.txt .' % opt.bucket) # download evolve.txt if exists
  518. for _ in range(300): # generations to evolve
  519. if Path('evolve.txt').exists(): # if evolve.txt exists: select best hyps and mutate
  520. # Select parent(s)
  521. parent = 'single' # parent selection method: 'single' or 'weighted'
  522. x = np.loadtxt('evolve.txt', ndmin=2)
  523. n = min(5, len(x)) # number of previous results to consider
  524. x = x[np.argsort(-fitness(x))][:n] # top n mutations
  525. w = fitness(x) - fitness(x).min() # weights
  526. if parent == 'single' or len(x) == 1:
  527. # x = x[random.randint(0, n - 1)] # random selection
  528. x = x[random.choices(range(n), weights=w)[0]] # weighted selection
  529. elif parent == 'weighted':
  530. x = (x * w.reshape(n, 1)).sum(0) / w.sum() # weighted combination
  531. # Mutate
  532. mp, s = 0.8, 0.2 # mutation probability, sigma
  533. npr = np.random
  534. npr.seed(int(time.time()))
  535. g = np.array([x[0] for x in meta.values()]) # gains 0-1
  536. ng = len(meta)
  537. v = np.ones(ng)
  538. while all(v == 1): # mutate until a change occurs (prevent duplicates)
  539. v = (g * (npr.random(ng) < mp) * npr.randn(ng) * npr.random() * s + 1).clip(0.3, 3.0)
  540. for i, k in enumerate(hyp.keys()): # plt.hist(v.ravel(), 300)
  541. hyp[k] = float(x[i + 7] * v[i]) # mutate
  542. # Constrain to limits
  543. for k, v in meta.items():
  544. hyp[k] = max(hyp[k], v[1]) # lower limit
  545. hyp[k] = min(hyp[k], v[2]) # upper limit
  546. hyp[k] = round(hyp[k], 5) # significant digits
  547. # Train mutation
  548. results = train(hyp.copy(), opt, device)
  549. # Write mutation results
  550. print_mutation(hyp.copy(), results, yaml_file, opt.bucket)
  551. # Plot results
  552. plot_evolution(yaml_file)
  553. print(f'Hyperparameter evolution complete. Best results saved as: {yaml_file}\n'
  554. f'Command to train a new model with these hyperparameters: $ python train.py --hyp {yaml_file}')