Commit 5bc99c52 authored by dzy7e's avatar dzy7e
Browse files

auc, map

parent 1fb954ee

requirements-zoo.txt

+1 −0
Original line number Diff line number Diff line
@@ -2,6 +2,7 @@ torch<2 
lpips

matplotlib

torchvision

torchmetrics>=0.11.4

tqdm

onnx

onnxoptimizer

zoo/ccip/train_.py

+28 −21
Original line number Diff line number Diff line
@@ -75,7 +75,7 @@ def _sample_analysis(poss, negs, svm_samples: int = 10000): 
def train(dataset_dir: str, session_name: Optional[str] = None, from_ckpt: Optional[str] = None,

          train_ratio: float = 0.8, max_epochs: int = 500, group_size: int = 30,

          learning_rate: float = 0.001, weight_decay: float = 1e-2, tau: float = 0.15,

          save_per_epoch: int = 10, eval_epoch: int = 5, log_iter: int = 500, num_workers=8,

          save_per_epoch: int = 10, eval_epoch: int = 5, loss_log_iter:int = 20, log_iter: int = 500, num_workers=8,

          model_name: str = 'clip/ViT-B/32', seed: Optional[int] = 0):

    if seed is not None:

        # native random, numpy, torch and faker's seeds are includes
@@ -171,22 +171,29 @@ def train(dataset_dir: str, session_name: Optional[str] = None, from_ckpt: Optio 
            running_loss += loss.item()*len(char_ids)

            train_pos_total += len(char_ids)



            mask = torch.ones_like(outputs)

            mask -= torch.diag_embed(torch.diag(mask))

            mask = torch.ones_like(outputs).bool()

            mask ^= torch.diag_embed(torch.diag(mask))

            outputs = outputs.detach().cpu()

            gt_same = (char_ids.view(-1, 1) == char_ids.view(1, -1)).detach().cpu()

            pred_list.append(outputs[mask])

            gt_list.append(gt_same.long()[mask])



            with torch.no_grad():

                if (i+1)%loss_log_iter == 0:

                    mean_loss = running_loss/train_pos_total

                    if writer:

                        writer.add_scalar('train/loss', mean_loss, epoch)



                if (i+1)%log_iter == 0:

                pred_t, gt_t = accelerator.gather_for_metrics((torch.cat(pred_list), torch.cat(gt_list)))

                    pred_t = torch.cat(pred_list).to(accelerator.device)

                    gt_t = torch.cat(gt_list).to(accelerator.device)

                    pred_t, gt_t = accelerator.gather_for_metrics((pred_t, gt_t))

                    if accelerator.is_local_main_process:

                        auc = metric_auroc(pred_t, gt_t).item()

                        ap = metric_ap(pred_t, gt_t).item()

                    mean_loss = running_loss/train_pos_total

                        logging.info(f'Epoch [{epoch}/{max_epochs}], loss: {mean_loss:.6f}, AUC: {auc:.3e}, AP: {ap:.3e}.')

                        if writer:

                        writer.add_scalar('train/loss', mean_loss, epoch)

                            #writer.add_scalar('train/loss', mean_loss, epoch)

                            writer.add_scalar('train/auc', auc, epoch)

                            writer.add_scalar('train/ap', auc, epoch)
@@ -205,21 +212,21 @@ def train(dataset_dir: str, session_name: Optional[str] = None, from_ckpt: Optio 


                    outputs = model(inputs)  # BxB



                    mask = torch.ones_like(outputs)

                    mask -= torch.diag_embed(torch.diag(mask))

                    mask = torch.ones_like(outputs).bool()

                    mask ^= torch.diag_embed(torch.diag(mask))

                    outputs = outputs.detach().cpu()

                    gt_same = (char_ids.view(-1, 1) == char_ids.view(1, -1)).detach().cpu()

                    pred_list.append(outputs[mask])

                    gt_list.append(gt_same.long()[mask])



                pred_t, gt_t = accelerator.gather_for_metrics((torch.cat(pred_list), torch.cat(gt_list)))

                pred_t = torch.cat(pred_list).to(accelerator.device)

                gt_t = torch.cat(gt_list).to(accelerator.device)

                pred_t, gt_t = accelerator.gather_for_metrics((pred_t, gt_t))

                if accelerator.is_local_main_process:

                    auc = metric_auroc(pred_t, gt_t).item()

                    ap = metric_ap(pred_t, gt_t).item()

                    mean_loss = running_loss/train_pos_total

                    logging.info(f'Epoch [{epoch}/{max_epochs}], loss: {mean_loss:.6f}, AUC: {auc:.3e}, AP: {ap:.3e}.')

                    logging.info(f'Epoch [{epoch}/{max_epochs}], AUC: {auc:.3e}, AP: {ap:.3e}.')

                    if writer:

                        writer.add_scalar('test/loss', mean_loss, epoch)

                        writer.add_scalar('test/auc', auc, epoch)

                        writer.add_scalar('test/ap', auc, epoch)