File size: 4,851 Bytes
c94c8c9 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 |
from tqdm import tqdm
import torch
from trainer.build import TRAINER_REGISTRY, BaseTrainer
@TRAINER_REGISTRY.register()
class OpenVocabTrainer(BaseTrainer):
def __init__(self, cfg):
super().__init__(cfg)
self.best_metric = -1
def forward(self, data_dict, mode):
return self.model(data_dict, mode)
def backward(self, loss, mode=None):
self.accelerator.backward(loss)
if self.grad_norm is not None and self.accelerator.sync_gradients:
self.accelerator.clip_grad_norm_(self.model.parameters(), self.grad_norm)
self.optimizer.step()
self.scheduler.step()
self.optimizer.zero_grad()
def train_step(self, epoch, mode = None):
self.model.train()
loader = self.data_loaders[self.mode]
is_main = self.accelerator.is_main_process
pbar = tqdm(loader, disable=not is_main, desc=f"[Epoch {epoch + 1}/{self.epochs}]")
for data_dict in pbar:
with self.accelerator.accumulate(self.model):
# Forward pass
data_dict = self.forward(data_dict, mode=mode)
# Compute loss
loss, losses = self.loss(data_dict)
# Backward + optimizer step
self.backward(loss, mode=mode)
# Logging
self.global_step += 1
log_dict = {'step': self.global_step, **losses}
if mode == 'qa':
metrics = self.evaluator["train"].batch_metrics(data_dict)
log_dict.update(metrics)
self.log(log_dict, mode="train")
@torch.no_grad()
def eval_step(self, epoch, mode):
self.model.eval()
loader = self.data_loaders["val"]
pbar = tqdm(range(len(loader)), disable=(not self.accelerator.is_main_process))
for i, data_dict in enumerate(loader):
data_dict = self.forward(data_dict, mode = mode)
loss, losses = self.loss(data_dict)
log_dict = {'epoch': epoch, **losses}
pbar.update(1)
self.log(log_dict, mode="val")
@torch.no_grad()
def test_step(self):
self.model.eval()
loader = self.data_loaders["test"]
pbar = tqdm(range(len(loader)), disable=(not self.accelerator.is_main_process))
for i, data_dict in enumerate(loader):
data_dict = self.forward(data_dict)
self.evaluator["val"].update(data_dict)
pbar.update(1)
is_best, results = self.evaluator["val"].record()
self.log(results, mode="test")
self.evaluator["val"].reset()
return results
def run(self):
num_trainable_params = 0
for name, param in self.model.named_parameters():
if param.requires_grad:
num_trainable_params += param.numel()
print(name)
print(f"Total number of trainable parameters: {num_trainable_params:,}")
if self.mode in ['warmup', "pretrain"]:
start_epoch = self.exp_tracker.epoch
self.global_step = start_epoch * len(self.data_loaders[self.mode])
for epoch in range(start_epoch, self.epochs):
self.exp_tracker.step()
self.train_step(epoch, mode=self.mode)
self.accelerator.wait_for_everyone()
if self.accelerator.is_main_process:
if self.epochs_per_save and (epoch + 1) % self.epochs_per_save == 0:
self.save(f"ckpt_{epoch+1}.pth")
self.save(f"ckpt_{epoch+1}.pth")
self.accelerator.end_training()
else:
if self.mode == "train":
start_epoch = self.exp_tracker.epoch
self.global_step = start_epoch * len(self.data_loaders["train"])
for epoch in range(start_epoch, self.epochs):
self.exp_tracker.step()
self.train_step(epoch)
if self.epochs_per_eval and (epoch + 1) % self.epochs_per_eval == 0:
is_best = self.eval_step(epoch)
self.accelerator.print(f"[Epoch {epoch + 1}/{self.epochs}] finished eval, is_best: {is_best}")
else:
is_best = False
self.accelerator.wait_for_everyone()
if self.accelerator.is_main_process:
if is_best:
self.save("best.pth")
if self.epochs_per_save and (epoch + 1) % self.epochs_per_save == 0:
self.save(f"ckpt_{epoch+1}.pth")
self.test_step()
if self.mode == "train":
self.accelerator.end_training() |