search

huggingface / transformers

🤗 Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX.
https://huggingface.co/transformers
Apache License 2.0
135.13k stars 27.04k forks source link

Training "microsoft/beit-large-finetuned-ade-640-640" on my dataset present some worry can not to solve #23924

Closed 813BoZai closed 1 year ago

813BoZai commented 1 year ago

System Info

win10 anaconda python3.8

Who can help?

PyTorch: @sgugger

Information

Tasks

Reproduction

"import timm import torch from easydict import EasyDict from torch import nn

import pandas as pd

metadata = pd.read_json("metadata.json", orient="colunms")

dict_all = metadata.sample_pairs.loc[metadata.index[22:]] index_organ_pd = pd.DataFrame() index_list = [] organ_list = [] for num in range(len(dict_all)): index_list.append(dict_all.index[num]) organ_list.append(dict_all[num]['organ']) index_organ_pd["index"] = index_list index_organ_pd["organ"] = organ_list

key_list = list(index_organ_pd.organ.value_counts().index)

id2label = {int(num): key_list[num] for num in range(len(key_list))}

id2label = {1:"BG", 2:"CA", 255:"UNK"} label2id = {v: k for k, v in id2label.items()}

import glob import os

train_path = 'D:\workplace\workplace_python\jupyter_lab\timm_model\dataset_oce\train\tissue\img' train_lablepath = 'D:\workplace\workplace_python\jupyter_lab\timm_model\dataset_oce\train\tissue\label' test_path = 'D:\workplace\workplace_python\jupyter_lab\timm_model\dataset_oce\test\tissue\img' test_lablepath = 'D:\workplace\workplace_python\jupyter_lab\timm_model\dataset_oce\test\tissue\label'

train_tissuePath = glob.glob(os.path.join(train_path, '*.jpg'))

train_tissuelabelPath = glob.glob(os.path.join(train_lablepath, '*.jpg'))

test_tissuePath = glob.glob(os.path.join(test_path, '*.jpg'))

test_tissuelabelPath = glob.glob(os.path.join(test_lablepath, '*.jpg'))

import PIL import cv2 import numpy as np

def label_imgpro(tmp_pic): if len(tmp_pic[tmp_pic[:]==2].reshape(-1,1))/len(tmp_pic.reshape(-1,1))>=0.2: return 2 else: return 1

def get_data(train_tissuePath,train_tissueLabelPath,index_organ_pd): train_ds={} image=[] annotation=[] scene_category =[] for num in range(len(train_tissuePath)): image.append(PIL.Image.open(train_tissuePath[num])) annotation.append(PIL.Image.open(train_tissueLabelPath[num]).convert('L'))

tmp_key = np.array(index_organ_pd[index_organ_pd["index"]==train_tissuePath[num].split("\")[-1].split(".")[0]]["organ"])[0]

    # print(tmp_key)
    # tmp_value = label2id.get(tmp_key)
    # print(tmp_value)
    scene_category.append(label_imgpro(np.array(PIL.Image.open(train_tissueLabelPath[num]).convert('L'))))
train_ds["image"]=image
train_ds["annotation"]=annotation
train_ds["scene_category"]=scene_category
return train_ds

train_dict = get_data(train_tissuePath,train_tissuelabelPath,index_organ_pd) test_dict = get_data(test_tissuePath,test_tissuelabelPath,index_organ_pd)

train_dict = get_data(train_tissuePath, train_tissuelabelPath, index_organ_pd) test_dict = get_data(test_tissuePath, test_tissuelabelPath, index_organ_pd)

import datasets

train_ds = datasets.Dataset.from_dict(train_dict) test_ds = datasets.Dataset.from_dict(test_dict)

from transformers import AutoImageProcessor

checkpoint = "microsoft/beit-large-finetuned-ade-640-640"

checkpoint = "nvidia/mit-b0"

image_processor = AutoImageProcessor.from_pretrained(checkpoint, do_reduce_labels=True)

from transformers import BeitImageProcessor

image_processor = BeitImageProcessor(do_reduce_labels=True)

from torchvision.transforms import ColorJitter

jitter = ColorJitter(brightness=0.25, contrast=0.25, saturation=0.25, hue=0.1)

def train_transforms(example_batch): images = [jitter(x) for x in example_batch["image"]] labels = [x for x in example_batch["annotation"]] inputs = image_processor(images, labels) return inputs

def val_transforms(example_batch): images = [x for x in example_batch["image"]] labels = [x for x in example_batch["annotation"]] inputs = image_processor(images, labels) return inputs

train_ds.set_transform(train_transforms) test_ds.set_transform(val_transforms)

import evaluate

metric = evaluate.load("mean_iou")

num_labels = len(id2label)

num_labels = 3

def compute_metrics(eval_pred): with torch.no_grad(): logits, labels = eval_pred logits_tensor = torch.from_numpy(logits) logits_tensor = nn.functional.interpolate( logits_tensor, size=labels.shape[-2:], mode="bilinear", align_corners=False, ).argmax(dim=1)

    pred_labels = logits_tensor.detach().cpu().numpy()
    metrics = metric.compute(
        predictions=pred_labels,
        references=labels,
        num_labels=num_labels,
        ignore_index=255,
        reduce_labels=False,
    )
    for key, value in metrics.items():
        if type(value) is np.ndarray:
            metrics[key] = value.tolist()
    return metrics

from transformers import AutoModelForSemanticSegmentation, TrainingArguments, Trainer

from transformers import BeitForSemanticSegmentation model = AutoModelForSemanticSegmentation.from_pretrained(checkpoint, id2label=id2label, label2id=label2id, ignore_mismatched_sizes=True)

model = BeitForSemanticSegmentation.from_pretrained(checkpoint, id2label=id2label, label2id=label2id)

from transformers import BertConfig, BertModel

Download model and configuration from huggingface.co and cache.

model = BertModel.from_pretrained("bert-base-uncased")

training_args = TrainingArguments( output_dir="beit-b0-oce", learning_rate=6e-5, num_train_epochs=50, per_device_train_batch_size=2, per_device_eval_batch_size=2, save_total_limit=3, evaluation_strategy="steps", save_strategy="steps", save_steps=20, eval_steps=20, logging_steps=1, eval_accumulation_steps=5, remove_unused_columns=False )

trainer = Trainer( model=model, args=training_args, train_dataset=train_ds, eval_dataset=test_ds, compute_metrics=compute_metrics

) trainer.train() " I want to train my model about beit on my dataset that is my code

Expected behavior

"0%| | 0/9100 [00:00<?, ?it/s]Traceback (most recent call last): File "D:\workplace\workplace_python\jupyter_lab\timm_model\beit_mode finetuned.py", line 188, in trainer.train() File "D:\anaconda\envs\timm_model\lib\site-packages\transformers\trainer.py", line 1664, in train return inner_training_loop( File "D:\anaconda\envs\timm_model\lib\site-packages\transformers\trainer.py", line 1940, in _inner_training_loop tr_loss_step = self.training_step(model, inputs) File "D:\anaconda\envs\timm_model\lib\site-packages\transformers\trainer.py", line 2735, in training_step loss = self.compute_loss(model, inputs) File "D:\anaconda\envs\timm_model\lib\site-packages\transformers\trainer.py", line 2767, in compute_loss outputs = model(inputs) File "D:\anaconda\envs\timm_model\lib\site-packages\torch\nn\modules\module.py", line 1501, in _call_impl return forward_call(*args, *kwargs) File "D:\anaconda\envs\timm_model\lib\site-packages\transformers\models\beit\modeling_beit.py", line 1276, in forward loss = self.compute_loss(logits, auxiliary_logits, labels) File "D:\anaconda\envs\timm_model\lib\site-packages\transformers\models\beit\modeling_beit.py", line 1194, in compute_loss main_loss = loss_fct(upsampled_logits, labels) File "D:\anaconda\envs\timm_model\lib\site-packages\torch\nn\modules\module.py", line 1501, in _call_impl return forward_call(args, kwargs) File "D:\anaconda\envs\timm_model\lib\site-packages\torch\nn\modules\loss.py", line 1174, in forward return F.cross_entropy(input, target, weight=self.weight, File "D:\anaconda\envs\timm_model\lib\site-packages\torch\nn\functional.py", line 3029, in cross_entropy return torch._C._nn.cross_entropy_loss(input, target, weight, _Reduction.get_enum(reduction), ignore_index, label_smoothing) IndexError: Target 62 is out of bounds. 0%| | 0/9100 [01:16<?, ?it/s]" this is my worry,I can not find where is the ‘62’ out of bounds.

sgugger commented 1 year ago

Please use the forums to debug your code as we keep the issues for bugs and feature requests only.

github-actions[bot] commented 1 year ago

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.