EngineX-Cambricon/enginex-mlu370-any2any
10
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

init

Browse Source
This commit is contained in:
luopingyi
2025-10-09 16:47:16 +08:00
parent c8feb4deb5
commit e27e3f16bb
5248 changed files with 1778505 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

212
transformers/examples/pytorch/image-classification/README.md Normal file
View File

@@ -0,0 +1,212 @@
<!---
Copyright 2021 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
-->
# Image classification examples
This directory contains 2 scripts that showcase how to fine-tune any model supported by the [`AutoModelForImageClassification` API](https://huggingface.co/docs/transformers/main/en/model_doc/auto#transformers.AutoModelForImageClassification) (such as [ViT](https://huggingface.co/docs/transformers/main/en/model_doc/vit), [ConvNeXT](https://huggingface.co/docs/transformers/main/en/model_doc/convnext), [ResNet](https://huggingface.co/docs/transformers/main/en/model_doc/resnet), [Swin Transformer](https://huggingface.co/docs/transformers/main/en/model_doc/swin)...) using PyTorch. They can be used to fine-tune models on both [datasets from the hub](#using-datasets-from-hub) as well as on [your own custom data](#using-your-own-data).
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/image_classification_inference_widget.png" height="400" />
Try out the inference widget here: https://huggingface.co/google/vit-base-patch16-224
Content:
- [PyTorch version, Trainer](#pytorch-version-trainer)
- [PyTorch version, no Trainer](#pytorch-version-no-trainer)
## PyTorch version, Trainer
Based on the script [`run_image_classification.py`](https://github.com/huggingface/transformers/blob/main/examples/pytorch/image-classification/run_image_classification.py).
The script leverages the 🤗 [Trainer API](https://huggingface.co/docs/transformers/main_classes/trainer) to automatically take care of the training for you, running on distributed environments right away.
### Using datasets from Hub
Here we show how to fine-tune a Vision Transformer (`ViT`) on the [beans](https://huggingface.co/datasets/beans) dataset, to classify the disease type of bean leaves.
```bash
python run_image_classification.py \
--dataset_name beans \
--output_dir ./beans_outputs/ \
--remove_unused_columns False \
--label_column_name labels \
--do_train \
--do_eval \
--push_to_hub \
--push_to_hub_model_id vit-base-beans \
--learning_rate 2e-5 \
--num_train_epochs 5 \
--per_device_train_batch_size 8 \
--per_device_eval_batch_size 8 \
--logging_strategy steps \
--logging_steps 10 \
--eval_strategy epoch \
--save_strategy epoch \
--load_best_model_at_end True \
--save_total_limit 3 \
--seed 1337
```
👀 See the results here: [nateraw/vit-base-beans](https://huggingface.co/nateraw/vit-base-beans).
Note that you can replace the model and dataset by simply setting the `model_name_or_path` and `dataset_name` arguments respectively, with any model or dataset from the [hub](https://huggingface.co/). For an overview of all possible arguments, we refer to the [docs](https://huggingface.co/docs/transformers/main_classes/trainer#transformers.TrainingArguments) of the `TrainingArguments`, which can be passed as flags.
> If your model classification head dimensions do not fit the number of labels in the dataset, you can specify `--ignore_mismatched_sizes` to adapt it.
### Using your own data
To use your own dataset, there are 2 ways:
- you can either provide your own folders as `--train_dir` and/or `--validation_dir` arguments
- you can upload your dataset to the hub (possibly as a private repo, if you prefer so), and simply pass the `--dataset_name` argument.
Below, we explain both in more detail.
#### Provide them as folders
If you provide your own folders with images, the script expects the following directory structure:
```bash
root/dog/xxx.png
root/dog/xxy.png
root/dog/[...]/xxz.png
root/cat/123.png
root/cat/nsdf3.png
root/cat/[...]/asd932_.png
```
In other words, you need to organize your images in subfolders, based on their class. You can then run the script like this:
```bash
python run_image_classification.py \
--train_dir <path-to-train-root> \
--output_dir ./outputs/ \
--remove_unused_columns False \
--do_train \
--do_eval
```
Internally, the script will use the [`ImageFolder`](https://huggingface.co/docs/datasets/v2.0.0/en/image_process#imagefolder) feature which will automatically turn the folders into 🤗 Dataset objects.
##### 💡 The above will split the train dir into training and evaluation sets
- To control the split amount, use the `--train_val_split` flag.
- To provide your own validation split in its own directory, you can pass the `--validation_dir <path-to-val-root>` flag.
#### Upload your data to the hub, as a (possibly private) repo
It's very easy (and convenient) to upload your image dataset to the hub using the [`ImageFolder`](https://huggingface.co/docs/datasets/v2.0.0/en/image_process#imagefolder) feature available in 🤗 Datasets. Simply do the following:
```python
from datasets import load_dataset
# example 1: local folder
dataset = load_dataset("imagefolder", data_dir="path_to_your_folder")
# example 2: local files (supported formats are tar, gzip, zip, xz, rar, zstd)
dataset = load_dataset("imagefolder", data_files="path_to_zip_file")
# example 3: remote files (supported formats are tar, gzip, zip, xz, rar, zstd)
dataset = load_dataset("imagefolder", data_files="https://download.microsoft.com/download/3/E/1/3E1C3F21-ECDB-4869-8368-6DEBA77B919F/kagglecatsanddogs_3367a.zip")
# example 4: providing several splits
dataset = load_dataset("imagefolder", data_files={"train": ["path/to/file1", "path/to/file2"], "test": ["path/to/file3", "path/to/file4"]})
```
`ImageFolder` will create a `label` column, and the label name is based on the directory name.
Next, push it to the hub!
```python
# assuming you have ran the hf auth login command in a terminal
dataset.push_to_hub("name_of_your_dataset")
# if you want to push to a private repo, simply pass private=True:
dataset.push_to_hub("name_of_your_dataset", private=True)
```
and that's it! You can now train your model by simply setting the `--dataset_name` argument to the name of your dataset on the hub (as explained in [Using datasets from the 🤗 hub](#using-datasets-from-hub)).
More on this can also be found in [this blog post](https://huggingface.co/blog/image-search-datasets).
### Sharing your model on 🤗 Hub
0. If you haven't already, [sign up](https://huggingface.co/join) for a 🤗 account
1. Make sure you have `git-lfs` installed and git set up.
```bash
$ apt install git-lfs
$ git config --global user.email "you@example.com"
$ git config --global user.name "Your Name"
```
2. Log in with your HuggingFace account credentials using `hf`:
```bash
$ hf auth login
# ...follow the prompts
```
3. When running the script, pass the following arguments:
```bash
python run_image_classification.py \
--push_to_hub \
--push_to_hub_model_id <name-your-model> \
...
```
## PyTorch version, no Trainer
Based on the script [`run_image_classification_no_trainer.py`](https://github.com/huggingface/transformers/blob/main/examples/pytorch/image-classification/run_image_classification_no_trainer.py).
Like `run_image_classification.py`, this script allows you to fine-tune any of the models on the [hub](https://huggingface.co/models) on an image classification task. The main difference is that this script exposes the bare training loop, to allow you to quickly experiment and add any customization you would like.
It offers less options than the script with `Trainer` (for instance you can easily change the options for the optimizer
or the dataloaders directly in the script) but still run in a distributed setup, and supports mixed precision by
the means of the [🤗 `Accelerate`](https://github.com/huggingface/accelerate) library. You can use the script normally
after installing it:
```bash
pip install git+https://github.com/huggingface/accelerate
```
You can then use your usual launchers to run in it in a distributed environment, but the easiest way is to run
```bash
accelerate config
```
and reply to the questions asked. Then
```bash
accelerate test
```
that will check everything is ready for training. Finally, you can launch training with
```bash
accelerate launch run_image_classification_no_trainer.py --image_column_name img
```
This command is the same and will work for:
- single/multiple CPUs
- single/multiple GPUs
- TPUs
Note that this library is in alpha release so your feedback is more than welcome if you encounter any problem using it.
Regarding using custom data with this script, we refer to [using your own data](#using-your-own-data).

6
transformers/examples/pytorch/image-classification/requirements.txt Normal file
View File

@@ -0,0 +1,6 @@
accelerate>=0.12.0
torch>=1.5.0
torchvision>=0.6.0
datasets>=2.14.0
evaluate
scikit-learn

453
transformers/examples/pytorch/image-classification/run_image_classification.py Executable file
View File

@@ -0,0 +1,453 @@
#!/usr/bin/env python
# Copyright 2021 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# /// script
# dependencies = [
# "transformers @ git+https://github.com/huggingface/transformers.git",
# "accelerate>=0.12.0",
# "torch>=1.5.0",
# "torchvision>=0.6.0",
# "datasets>=2.14.0",
# "evaluate",
# "scikit-learn",
# ]
# ///
import logging
import os
import sys
from dataclasses import dataclass, field
from typing import Optional
import evaluate
import numpy as np
import torch
from datasets import load_dataset
from PIL import Image
from torchvision.transforms import (
CenterCrop,
Compose,
Lambda,
Normalize,
RandomHorizontalFlip,
RandomResizedCrop,
Resize,
ToTensor,
)
import transformers
from transformers import (
MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING,
AutoConfig,
AutoImageProcessor,
AutoModelForImageClassification,
HfArgumentParser,
TimmWrapperImageProcessor,
Trainer,
TrainingArguments,
set_seed,
)
from transformers.trainer_utils import get_last_checkpoint
from transformers.utils import check_min_version, send_example_telemetry
from transformers.utils.versions import require_version
""" Fine-tuning a 🤗 Transformers model for image classification"""
logger = logging.getLogger(__name__)
# Will error if the minimal version of Transformers is not installed. Remove at your own risks.
check_min_version("4.57.0.dev0")
require_version("datasets>=2.14.0", "To fix: pip install -r examples/pytorch/image-classification/requirements.txt")
MODEL_CONFIG_CLASSES = list(MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING.keys())
MODEL_TYPES = tuple(conf.model_type for conf in MODEL_CONFIG_CLASSES)
def pil_loader(path: str):
with open(path, "rb") as f:
im = Image.open(f)
return im.convert("RGB")
@dataclass
class DataTrainingArguments:
"""
Arguments pertaining to what data we are going to input our model for training and eval.
Using `HfArgumentParser` we can turn this class into argparse arguments to be able to specify
them on the command line.
"""
dataset_name: Optional[str] = field(
default=None,
metadata={
"help": "Name of a dataset from the hub (could be your own, possibly private dataset hosted on the hub)."
},
)
dataset_config_name: Optional[str] = field(
default=None, metadata={"help": "The configuration name of the dataset to use (via the datasets library)."}
)
train_dir: Optional[str] = field(default=None, metadata={"help": "A folder containing the training data."})
validation_dir: Optional[str] = field(default=None, metadata={"help": "A folder containing the validation data."})
train_val_split: Optional[float] = field(
default=0.15, metadata={"help": "Percent to split off of train for validation."}
)
max_train_samples: Optional[int] = field(
default=None,
metadata={
"help": (
"For debugging purposes or quicker training, truncate the number of training examples to this "
"value if set."
)
},
)
max_eval_samples: Optional[int] = field(
default=None,
metadata={
"help": (
"For debugging purposes or quicker training, truncate the number of evaluation examples to this "
"value if set."
)
},
)
image_column_name: str = field(
default="image",
metadata={"help": "The name of the dataset column containing the image data. Defaults to 'image'."},
)
label_column_name: str = field(
default="label",
metadata={"help": "The name of the dataset column containing the labels. Defaults to 'label'."},
)
def __post_init__(self):
if self.dataset_name is None and (self.train_dir is None and self.validation_dir is None):
raise ValueError(
"You must specify either a dataset name from the hub or a train and/or validation directory."
)
@dataclass
class ModelArguments:
"""
Arguments pertaining to which model/config/tokenizer we are going to fine-tune from.
"""
model_name_or_path: str = field(
default="google/vit-base-patch16-224-in21k",
metadata={"help": "Path to pretrained model or model identifier from huggingface.co/models"},
)
model_type: Optional[str] = field(
default=None,
metadata={"help": "If training from scratch, pass a model type from the list: " + ", ".join(MODEL_TYPES)},
)
config_name: Optional[str] = field(
default=None, metadata={"help": "Pretrained config name or path if not the same as model_name"}
)
cache_dir: Optional[str] = field(
default=None, metadata={"help": "Where do you want to store the pretrained models downloaded from s3"}
)
model_revision: str = field(
default="main",
metadata={"help": "The specific model version to use (can be a branch name, tag name or commit id)."},
)
image_processor_name: str = field(default=None, metadata={"help": "Name or path of preprocessor config."})
token: str = field(
default=None,
metadata={
"help": (
"The token to use as HTTP bearer authorization for remote files. If not specified, will use the token "
"generated when running `hf auth login` (stored in `~/.huggingface`)."
)
},
)
trust_remote_code: bool = field(
default=False,
metadata={
"help": (
"Whether to trust the execution of code from datasets/models defined on the Hub."
" This option should only be set to `True` for repositories you trust and in which you have read the"
" code, as it will execute code present on the Hub on your local machine."
)
},
)
ignore_mismatched_sizes: bool = field(
default=False,
metadata={"help": "Will enable to load a pretrained model whose head dimensions are different."},
)
def main():
# See all possible arguments in src/transformers/training_args.py
# or by passing the --help flag to this script.
# We now keep distinct sets of args, for a cleaner separation of concerns.
parser = HfArgumentParser((ModelArguments, DataTrainingArguments, TrainingArguments))
if len(sys.argv) == 2 and sys.argv[1].endswith(".json"):
# If we pass only one argument to the script and it's the path to a json file,
# let's parse it to get our arguments.
model_args, data_args, training_args = parser.parse_json_file(json_file=os.path.abspath(sys.argv[1]))
else:
model_args, data_args, training_args = parser.parse_args_into_dataclasses()
# Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
# information sent is the one passed as arguments along with your Python/PyTorch versions.
send_example_telemetry("run_image_classification", model_args, data_args)
# Setup logging
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
datefmt="%m/%d/%Y %H:%M:%S",
handlers=[logging.StreamHandler(sys.stdout)],
)
if training_args.should_log:
# The default of training_args.log_level is passive, so we set log level at info here to have that default.
transformers.utils.logging.set_verbosity_info()
log_level = training_args.get_process_log_level()
logger.setLevel(log_level)
transformers.utils.logging.set_verbosity(log_level)
transformers.utils.logging.enable_default_handler()
transformers.utils.logging.enable_explicit_format()
# Log on each process the small summary:
logger.warning(
f"Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu}, "
+ f"distributed training: {training_args.parallel_mode.value == 'distributed'}, 16-bits training: {training_args.fp16}"
)
logger.info(f"Training/evaluation parameters {training_args}")
# Detecting last checkpoint.
last_checkpoint = None
if os.path.isdir(training_args.output_dir) and training_args.do_train and not training_args.overwrite_output_dir:
last_checkpoint = get_last_checkpoint(training_args.output_dir)
if last_checkpoint is None and len(os.listdir(training_args.output_dir)) > 0:
raise ValueError(
f"Output directory ({training_args.output_dir}) already exists and is not empty. "
"Use --overwrite_output_dir to overcome."
)
elif last_checkpoint is not None and training_args.resume_from_checkpoint is None:
logger.info(
f"Checkpoint detected, resuming training at {last_checkpoint}. To avoid this behavior, change "
"the `--output_dir` or add `--overwrite_output_dir` to train from scratch."
)
# Set seed before initializing model.
set_seed(training_args.seed)
# Initialize our dataset and prepare it for the 'image-classification' task.
if data_args.dataset_name is not None:
dataset = load_dataset(
data_args.dataset_name,
data_args.dataset_config_name,
cache_dir=model_args.cache_dir,
token=model_args.token,
trust_remote_code=model_args.trust_remote_code,
)
else:
data_files = {}
if data_args.train_dir is not None:
data_files["train"] = os.path.join(data_args.train_dir, "**")
if data_args.validation_dir is not None:
data_files["validation"] = os.path.join(data_args.validation_dir, "**")
dataset = load_dataset(
"imagefolder",
data_files=data_files,
cache_dir=model_args.cache_dir,
)
dataset_column_names = dataset["train"].column_names if "train" in dataset else dataset["validation"].column_names
if data_args.image_column_name not in dataset_column_names:
raise ValueError(
f"--image_column_name {data_args.image_column_name} not found in dataset '{data_args.dataset_name}'. "
"Make sure to set `--image_column_name` to the correct audio column - one of "
f"{', '.join(dataset_column_names)}."
)
if data_args.label_column_name not in dataset_column_names:
raise ValueError(
f"--label_column_name {data_args.label_column_name} not found in dataset '{data_args.dataset_name}'. "
"Make sure to set `--label_column_name` to the correct text column - one of "
f"{', '.join(dataset_column_names)}."
)
def collate_fn(examples):
pixel_values = torch.stack([example["pixel_values"] for example in examples])
labels = torch.tensor([example[data_args.label_column_name] for example in examples])
return {"pixel_values": pixel_values, "labels": labels}
# If we don't have a validation split, split off a percentage of train as validation.
data_args.train_val_split = None if "validation" in dataset else data_args.train_val_split
if isinstance(data_args.train_val_split, float) and data_args.train_val_split > 0.0:
split = dataset["train"].train_test_split(data_args.train_val_split)
dataset["train"] = split["train"]
dataset["validation"] = split["test"]
# Prepare label mappings.
# We'll include these in the model's config to get human readable labels in the Inference API.
labels = dataset["train"].features[data_args.label_column_name].names
label2id, id2label = {}, {}
for i, label in enumerate(labels):
label2id[label] = str(i)
id2label[str(i)] = label
# Load the accuracy metric from the datasets package
metric = evaluate.load("accuracy", cache_dir=model_args.cache_dir)
# Define our compute_metrics function. It takes an `EvalPrediction` object (a namedtuple with a
# predictions and label_ids field) and has to return a dictionary string to float.
def compute_metrics(p):
"""Computes accuracy on a batch of predictions"""
return metric.compute(predictions=np.argmax(p.predictions, axis=1), references=p.label_ids)
config = AutoConfig.from_pretrained(
model_args.config_name or model_args.model_name_or_path,
num_labels=len(labels),
label2id=label2id,
id2label=id2label,
finetuning_task="image-classification",
cache_dir=model_args.cache_dir,
revision=model_args.model_revision,
token=model_args.token,
trust_remote_code=model_args.trust_remote_code,
)
model = AutoModelForImageClassification.from_pretrained(
model_args.model_name_or_path,
from_tf=bool(".ckpt" in model_args.model_name_or_path),
config=config,
cache_dir=model_args.cache_dir,
revision=model_args.model_revision,
token=model_args.token,
trust_remote_code=model_args.trust_remote_code,
ignore_mismatched_sizes=model_args.ignore_mismatched_sizes,
)
image_processor = AutoImageProcessor.from_pretrained(
model_args.image_processor_name or model_args.model_name_or_path,
cache_dir=model_args.cache_dir,
revision=model_args.model_revision,
token=model_args.token,
trust_remote_code=model_args.trust_remote_code,
)
# Define torchvision transforms to be applied to each image.
if isinstance(image_processor, TimmWrapperImageProcessor):
_train_transforms = image_processor.train_transforms
_val_transforms = image_processor.val_transforms
else:
if "shortest_edge" in image_processor.size:
size = image_processor.size["shortest_edge"]
else:
size = (image_processor.size["height"], image_processor.size["width"])
# Create normalization transform
if hasattr(image_processor, "image_mean") and hasattr(image_processor, "image_std"):
normalize = Normalize(mean=image_processor.image_mean, std=image_processor.image_std)
else:
normalize = Lambda(lambda x: x)
_train_transforms = Compose(
[
RandomResizedCrop(size),
RandomHorizontalFlip(),
ToTensor(),
normalize,
]
)
_val_transforms = Compose(
[
Resize(size),
CenterCrop(size),
ToTensor(),
normalize,
]
)
def train_transforms(example_batch):
"""Apply _train_transforms across a batch."""
example_batch["pixel_values"] = [
_train_transforms(pil_img.convert("RGB")) for pil_img in example_batch[data_args.image_column_name]
]
return example_batch
def val_transforms(example_batch):
"""Apply _val_transforms across a batch."""
example_batch["pixel_values"] = [
_val_transforms(pil_img.convert("RGB")) for pil_img in example_batch[data_args.image_column_name]
]
return example_batch
if training_args.do_train:
if "train" not in dataset:
raise ValueError("--do_train requires a train dataset")
if data_args.max_train_samples is not None:
dataset["train"] = (
dataset["train"].shuffle(seed=training_args.seed).select(range(data_args.max_train_samples))
)
# Set the training transforms
dataset["train"].set_transform(train_transforms)
if training_args.do_eval:
if "validation" not in dataset:
raise ValueError("--do_eval requires a validation dataset")
if data_args.max_eval_samples is not None:
dataset["validation"] = (
dataset["validation"].shuffle(seed=training_args.seed).select(range(data_args.max_eval_samples))
)
# Set the validation transforms
dataset["validation"].set_transform(val_transforms)
# Initialize our trainer
trainer = Trainer(
model=model,
args=training_args,
train_dataset=dataset["train"] if training_args.do_train else None,
eval_dataset=dataset["validation"] if training_args.do_eval else None,
compute_metrics=compute_metrics,
processing_class=image_processor,
data_collator=collate_fn,
)
# Training
if training_args.do_train:
checkpoint = None
if training_args.resume_from_checkpoint is not None:
checkpoint = training_args.resume_from_checkpoint
elif last_checkpoint is not None:
checkpoint = last_checkpoint
train_result = trainer.train(resume_from_checkpoint=checkpoint)
trainer.save_model()
trainer.log_metrics("train", train_result.metrics)
trainer.save_metrics("train", train_result.metrics)
trainer.save_state()
# Evaluation
if training_args.do_eval:
metrics = trainer.evaluate()
trainer.log_metrics("eval", metrics)
trainer.save_metrics("eval", metrics)
# Write model card and (optionally) push to hub
kwargs = {
"finetuned_from": model_args.model_name_or_path,
"tasks": "image-classification",
"dataset": data_args.dataset_name,
"tags": ["image-classification", "vision"],
}
if training_args.push_to_hub:
trainer.push_to_hub(**kwargs)
else:
trainer.create_model_card(**kwargs)
if __name__ == "__main__":
main()

658
transformers/examples/pytorch/image-classification/run_image_classification_no_trainer.py Normal file
View File

@@ -0,0 +1,658 @@
# Copyright 2022 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# /// script
# dependencies = [
# "transformers @ git+https://github.com/huggingface/transformers.git",
# "accelerate>=0.12.0",
# "torch>=1.5.0",
# "torchvision>=0.6.0",
# "datasets>=2.14.0",
# "evaluate",
# "scikit-learn",
# ]
# ///
"""Finetuning any 🤗 Transformers model for image classification leveraging 🤗 Accelerate."""
import argparse
import json
import logging
import math
import os
from pathlib import Path
import datasets
import evaluate
import torch
from accelerate import Accelerator
from accelerate.logging import get_logger
from accelerate.utils import set_seed
from datasets import load_dataset
from huggingface_hub import HfApi
from torch.utils.data import DataLoader
from torchvision.transforms import (
CenterCrop,
Compose,
Lambda,
Normalize,
RandomHorizontalFlip,
RandomResizedCrop,
Resize,
ToTensor,
)
from tqdm.auto import tqdm
import transformers
from transformers import AutoConfig, AutoImageProcessor, AutoModelForImageClassification, SchedulerType, get_scheduler
from transformers.utils import check_min_version, send_example_telemetry
from transformers.utils.versions import require_version
# Will error if the minimal version of Transformers is not installed. Remove at your own risks.
check_min_version("4.57.0.dev0")
logger = get_logger(__name__)
require_version("datasets>=2.0.0", "To fix: pip install -r examples/pytorch/image-classification/requirements.txt")
def parse_args():
parser = argparse.ArgumentParser(description="Fine-tune a Transformers model on an image classification dataset")
parser.add_argument(
"--dataset_name",
type=str,
default="cifar10",
help=(
"The name of the Dataset (from the HuggingFace hub) to train on (could be your own, possibly private,"
" dataset)."
),
)
parser.add_argument("--train_dir", type=str, default=None, help="A folder containing the training data.")
parser.add_argument("--validation_dir", type=str, default=None, help="A folder containing the validation data.")
parser.add_argument(
"--max_train_samples",
type=int,
default=None,
help=(
"For debugging purposes or quicker training, truncate the number of training examples to this "
"value if set."
),
)
parser.add_argument(
"--max_eval_samples",
type=int,
default=None,
help=(
"For debugging purposes or quicker training, truncate the number of evaluation examples to this "
"value if set."
),
)
parser.add_argument(
"--train_val_split",
type=float,
default=0.15,
help="Percent to split off of train for validation",
)
parser.add_argument(
"--model_name_or_path",
type=str,
help="Path to pretrained model or model identifier from huggingface.co/models.",
default="google/vit-base-patch16-224-in21k",
)
parser.add_argument(
"--per_device_train_batch_size",
type=int,
default=8,
help="Batch size (per device) for the training dataloader.",
)
parser.add_argument(
"--per_device_eval_batch_size",
type=int,
default=8,
help="Batch size (per device) for the evaluation dataloader.",
)
parser.add_argument(
"--learning_rate",
type=float,
default=5e-5,
help="Initial learning rate (after the potential warmup period) to use.",
)
parser.add_argument("--weight_decay", type=float, default=0.0, help="Weight decay to use.")
parser.add_argument("--num_train_epochs", type=int, default=3, help="Total number of training epochs to perform.")
parser.add_argument(
"--max_train_steps",
type=int,
default=None,
help="Total number of training steps to perform. If provided, overrides num_train_epochs.",
)
parser.add_argument(
"--gradient_accumulation_steps",
type=int,
default=1,
help="Number of updates steps to accumulate before performing a backward/update pass.",
)
parser.add_argument(
"--lr_scheduler_type",
type=SchedulerType,
default="linear",
help="The scheduler type to use.",
choices=["linear", "cosine", "cosine_with_restarts", "polynomial", "constant", "constant_with_warmup"],
)
parser.add_argument(
"--num_warmup_steps", type=int, default=0, help="Number of steps for the warmup in the lr scheduler."
)
parser.add_argument("--output_dir", type=str, default=None, help="Where to store the final model.")
parser.add_argument("--seed", type=int, default=None, help="A seed for reproducible training.")
parser.add_argument("--push_to_hub", action="store_true", help="Whether or not to push the model to the Hub.")
parser.add_argument(
"--hub_model_id", type=str, help="The name of the repository to keep in sync with the local `output_dir`."
)
parser.add_argument("--hub_token", type=str, help="The token to use to push to the Model Hub.")
parser.add_argument(
"--trust_remote_code",
action="store_true",
help=(
"Whether to trust the execution of code from datasets/models defined on the Hub."
" This option should only be set to `True` for repositories you trust and in which you have read the"
" code, as it will execute code present on the Hub on your local machine."
),
)
parser.add_argument(
"--checkpointing_steps",
type=str,
default=None,
help="Whether the various states should be saved at the end of every n steps, or 'epoch' for each epoch.",
)
parser.add_argument(
"--resume_from_checkpoint",
type=str,
default=None,
help="If the training should continue from a checkpoint folder.",
)
parser.add_argument(
"--with_tracking",
action="store_true",
help="Whether to enable experiment trackers for logging.",
)
parser.add_argument(
"--report_to",
type=str,
default="all",
help=(
'The integration to report the results and logs to. Supported platforms are `"tensorboard"`,'
' `"wandb"`, `"comet_ml"` and `"clearml"`. Use `"all"` (default) to report to all integrations. '
"Only applicable when `--with_tracking` is passed."
),
)
parser.add_argument(
"--ignore_mismatched_sizes",
action="store_true",
help="Whether or not to enable to load a pretrained model whose head dimensions are different.",
)
parser.add_argument(
"--image_column_name",
type=str,
default="image",
help="The name of the dataset column containing the image data. Defaults to 'image'.",
)
parser.add_argument(
"--label_column_name",
type=str,
default="label",
help="The name of the dataset column containing the labels. Defaults to 'label'.",
)
args = parser.parse_args()
# Sanity checks
if args.dataset_name is None and args.train_dir is None and args.validation_dir is None:
raise ValueError("Need either a dataset name or a training/validation folder.")
if args.push_to_hub or args.with_tracking:
if args.output_dir is None:
raise ValueError(
"Need an `output_dir` to create a repo when `--push_to_hub` or `with_tracking` is specified."
)
if args.output_dir is not None:
os.makedirs(args.output_dir, exist_ok=True)
return args
def main():
args = parse_args()
# Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
# information sent is the one passed as arguments along with your Python/PyTorch versions.
send_example_telemetry("run_image_classification_no_trainer", args)
# Initialize the accelerator. We will let the accelerator handle device placement for us in this example.
# If we're using tracking, we also need to initialize it here and it will by default pick up all supported trackers
# in the environment
accelerator_log_kwargs = {}
if args.with_tracking:
accelerator_log_kwargs["log_with"] = args.report_to
accelerator_log_kwargs["project_dir"] = args.output_dir
accelerator = Accelerator(gradient_accumulation_steps=args.gradient_accumulation_steps, **accelerator_log_kwargs)
logger.info(accelerator.state)
# Make one log on every process with the configuration for debugging.
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
datefmt="%m/%d/%Y %H:%M:%S",
level=logging.INFO,
)
logger.info(accelerator.state, main_process_only=False)
if accelerator.is_local_main_process:
datasets.utils.logging.set_verbosity_warning()
transformers.utils.logging.set_verbosity_info()
else:
datasets.utils.logging.set_verbosity_error()
transformers.utils.logging.set_verbosity_error()
# If passed along, set the training seed now.
if args.seed is not None:
set_seed(args.seed)
# Handle the repository creation
if accelerator.is_main_process:
if args.push_to_hub:
# Retrieve of infer repo_name
repo_name = args.hub_model_id
if repo_name is None:
repo_name = Path(args.output_dir).absolute().name
# Create repo and retrieve repo_id
api = HfApi()
repo_id = api.create_repo(repo_name, exist_ok=True, token=args.hub_token).repo_id
with open(os.path.join(args.output_dir, ".gitignore"), "w+") as gitignore:
if "step_*" not in gitignore:
gitignore.write("step_*\n")
if "epoch_*" not in gitignore:
gitignore.write("epoch_*\n")
elif args.output_dir is not None:
os.makedirs(args.output_dir, exist_ok=True)
accelerator.wait_for_everyone()
# Get the datasets: you can either provide your own training and evaluation files (see below)
# or specify a Dataset from the hub (the dataset will be downloaded automatically from the datasets Hub).
# In distributed training, the load_dataset function guarantees that only one local process can concurrently
# download the dataset.
if args.dataset_name is not None:
# Downloading and loading a dataset from the hub.
dataset = load_dataset(args.dataset_name, trust_remote_code=args.trust_remote_code)
else:
data_files = {}
if args.train_dir is not None:
data_files["train"] = os.path.join(args.train_dir, "**")
if args.validation_dir is not None:
data_files["validation"] = os.path.join(args.validation_dir, "**")
dataset = load_dataset(
"imagefolder",
data_files=data_files,
)
# See more about loading custom images at
# https://huggingface.co/docs/datasets/v2.0.0/en/image_process#imagefolder.
dataset_column_names = dataset["train"].column_names if "train" in dataset else dataset["validation"].column_names
if args.image_column_name not in dataset_column_names:
raise ValueError(
f"--image_column_name {args.image_column_name} not found in dataset '{args.dataset_name}'. "
"Make sure to set `--image_column_name` to the correct audio column - one of "
f"{', '.join(dataset_column_names)}."
)
if args.label_column_name not in dataset_column_names:
raise ValueError(
f"--label_column_name {args.label_column_name} not found in dataset '{args.dataset_name}'. "
"Make sure to set `--label_column_name` to the correct text column - one of "
f"{', '.join(dataset_column_names)}."
)
# If we don't have a validation split, split off a percentage of train as validation.
args.train_val_split = None if "validation" in dataset else args.train_val_split
if isinstance(args.train_val_split, float) and args.train_val_split > 0.0:
split = dataset["train"].train_test_split(args.train_val_split)
dataset["train"] = split["train"]
dataset["validation"] = split["test"]
# Prepare label mappings.
# We'll include these in the model's config to get human readable labels in the Inference API.
labels = dataset["train"].features[args.label_column_name].names
label2id = {label: str(i) for i, label in enumerate(labels)}
id2label = {str(i): label for i, label in enumerate(labels)}
# Load pretrained model and image processor
#
# In distributed training, the .from_pretrained methods guarantee that only one local process can concurrently
# download model & vocab.
config = AutoConfig.from_pretrained(
args.model_name_or_path,
num_labels=len(labels),
id2label=id2label,
label2id=label2id,
finetuning_task="image-classification",
trust_remote_code=args.trust_remote_code,
)
image_processor = AutoImageProcessor.from_pretrained(
args.model_name_or_path,
trust_remote_code=args.trust_remote_code,
)
model = AutoModelForImageClassification.from_pretrained(
args.model_name_or_path,
from_tf=bool(".ckpt" in args.model_name_or_path),
config=config,
ignore_mismatched_sizes=args.ignore_mismatched_sizes,
trust_remote_code=args.trust_remote_code,
)
# Preprocessing the datasets
# Define torchvision transforms to be applied to each image.
if "shortest_edge" in image_processor.size:
size = image_processor.size["shortest_edge"]
else:
size = (image_processor.size["height"], image_processor.size["width"])
normalize = (
Normalize(mean=image_processor.image_mean, std=image_processor.image_std)
if hasattr(image_processor, "image_mean") and hasattr(image_processor, "image_std")
else Lambda(lambda x: x)
)
train_transforms = Compose(
[
RandomResizedCrop(size),
RandomHorizontalFlip(),
ToTensor(),
normalize,
]
)
val_transforms = Compose(
[
Resize(size),
CenterCrop(size),
ToTensor(),
normalize,
]
)
def preprocess_train(example_batch):
"""Apply _train_transforms across a batch."""
example_batch["pixel_values"] = [
train_transforms(image.convert("RGB")) for image in example_batch[args.image_column_name]
]
return example_batch
def preprocess_val(example_batch):
"""Apply _val_transforms across a batch."""
example_batch["pixel_values"] = [
val_transforms(image.convert("RGB")) for image in example_batch[args.image_column_name]
]
return example_batch
with accelerator.main_process_first():
if args.max_train_samples is not None:
dataset["train"] = dataset["train"].shuffle(seed=args.seed).select(range(args.max_train_samples))
# Set the training transforms
train_dataset = dataset["train"].with_transform(preprocess_train)
if args.max_eval_samples is not None:
dataset["validation"] = dataset["validation"].shuffle(seed=args.seed).select(range(args.max_eval_samples))
# Set the validation transforms
eval_dataset = dataset["validation"].with_transform(preprocess_val)
# DataLoaders creation:
def collate_fn(examples):
pixel_values = torch.stack([example["pixel_values"] for example in examples])
labels = torch.tensor([example[args.label_column_name] for example in examples])
return {"pixel_values": pixel_values, "labels": labels}
train_dataloader = DataLoader(
train_dataset, shuffle=True, collate_fn=collate_fn, batch_size=args.per_device_train_batch_size
)
eval_dataloader = DataLoader(eval_dataset, collate_fn=collate_fn, batch_size=args.per_device_eval_batch_size)
# Optimizer
# Split weights in two groups, one with weight decay and the other not.
no_decay = ["bias", "LayerNorm.weight"]
optimizer_grouped_parameters = [
{
"params": [p for n, p in model.named_parameters() if not any(nd in n for nd in no_decay)],
"weight_decay": args.weight_decay,
},
{
"params": [p for n, p in model.named_parameters() if any(nd in n for nd in no_decay)],
"weight_decay": 0.0,
},
]
optimizer = torch.optim.AdamW(optimizer_grouped_parameters, lr=args.learning_rate)
# Scheduler and math around the number of training steps.
overrode_max_train_steps = False
num_update_steps_per_epoch = math.ceil(len(train_dataloader) / args.gradient_accumulation_steps)
if args.max_train_steps is None:
args.max_train_steps = args.num_train_epochs * num_update_steps_per_epoch
overrode_max_train_steps = True
lr_scheduler = get_scheduler(
name=args.lr_scheduler_type,
optimizer=optimizer,
num_warmup_steps=args.num_warmup_steps * accelerator.num_processes,
num_training_steps=args.max_train_steps
if overrode_max_train_steps
else args.max_train_steps * accelerator.num_processes,
)
# Prepare everything with our `accelerator`.
model, optimizer, train_dataloader, eval_dataloader, lr_scheduler = accelerator.prepare(
model, optimizer, train_dataloader, eval_dataloader, lr_scheduler
)
# We need to recalculate our total training steps as the size of the training dataloader may have changed.
num_update_steps_per_epoch = math.ceil(len(train_dataloader) / args.gradient_accumulation_steps)
if overrode_max_train_steps:
args.max_train_steps = args.num_train_epochs * num_update_steps_per_epoch
# Afterwards we recalculate our number of training epochs
args.num_train_epochs = math.ceil(args.max_train_steps / num_update_steps_per_epoch)
# Figure out how many steps we should save the Accelerator states
checkpointing_steps = args.checkpointing_steps
if checkpointing_steps is not None and checkpointing_steps.isdigit():
checkpointing_steps = int(checkpointing_steps)
# We need to initialize the trackers we use, and also store our configuration.
# The trackers initializes automatically on the main process.
if args.with_tracking:
experiment_config = vars(args)
# TensorBoard cannot log Enums, need the raw value
experiment_config["lr_scheduler_type"] = experiment_config["lr_scheduler_type"].value
accelerator.init_trackers("image_classification_no_trainer", experiment_config)
# Get the metric function
metric = evaluate.load("accuracy")
# Train!
total_batch_size = args.per_device_train_batch_size * accelerator.num_processes * args.gradient_accumulation_steps
logger.info("***** Running training *****")
logger.info(f" Num examples = {len(train_dataset)}")
logger.info(f" Num Epochs = {args.num_train_epochs}")
logger.info(f" Instantaneous batch size per device = {args.per_device_train_batch_size}")
logger.info(f" Total train batch size (w. parallel, distributed & accumulation) = {total_batch_size}")
logger.info(f" Gradient Accumulation steps = {args.gradient_accumulation_steps}")
logger.info(f" Total optimization steps = {args.max_train_steps}")
# Only show the progress bar once on each machine.
progress_bar = tqdm(range(args.max_train_steps), disable=not accelerator.is_local_main_process)
completed_steps = 0
starting_epoch = 0
# Potentially load in the weights and states from a previous save
if args.resume_from_checkpoint:
if args.resume_from_checkpoint is not None or args.resume_from_checkpoint != "":
checkpoint_path = args.resume_from_checkpoint
path = os.path.basename(args.resume_from_checkpoint)
else:
# Get the most recent checkpoint
dirs = [f.name for f in os.scandir(os.getcwd()) if f.is_dir()]
dirs.sort(key=os.path.getctime)
path = dirs[-1] # Sorts folders by date modified, most recent checkpoint is the last
checkpoint_path = path
path = os.path.basename(checkpoint_path)
accelerator.print(f"Resumed from checkpoint: {checkpoint_path}")
accelerator.load_state(checkpoint_path)
# Extract `epoch_{i}` or `step_{i}`
training_difference = os.path.splitext(path)[0]
if "epoch" in training_difference:
starting_epoch = int(training_difference.replace("epoch_", "")) + 1
resume_step = None
completed_steps = starting_epoch * num_update_steps_per_epoch
else:
# need to multiply `gradient_accumulation_steps` to reflect real steps
resume_step = int(training_difference.replace("step_", "")) * args.gradient_accumulation_steps
starting_epoch = resume_step // len(train_dataloader)
completed_steps = resume_step // args.gradient_accumulation_steps
resume_step -= starting_epoch * len(train_dataloader)
# update the progress_bar if load from checkpoint
progress_bar.update(completed_steps)
for epoch in range(starting_epoch, args.num_train_epochs):
model.train()
if args.with_tracking:
total_loss = 0
if args.resume_from_checkpoint and epoch == starting_epoch and resume_step is not None:
# We skip the first `n` batches in the dataloader when resuming from a checkpoint
active_dataloader = accelerator.skip_first_batches(train_dataloader, resume_step)
else:
active_dataloader = train_dataloader
for step, batch in enumerate(active_dataloader):
with accelerator.accumulate(model):
outputs = model(**batch)
loss = outputs.loss
# We keep track of the loss at each epoch
if args.with_tracking:
total_loss += loss.detach().float()
accelerator.backward(loss)
optimizer.step()
lr_scheduler.step()
optimizer.zero_grad()
# Checks if the accelerator has performed an optimization step behind the scenes
if accelerator.sync_gradients:
progress_bar.update(1)
completed_steps += 1
if isinstance(checkpointing_steps, int):
if completed_steps % checkpointing_steps == 0 and accelerator.sync_gradients:
output_dir = f"step_{completed_steps}"
if args.output_dir is not None:
output_dir = os.path.join(args.output_dir, output_dir)
accelerator.save_state(output_dir)
if args.push_to_hub and epoch < args.num_train_epochs - 1:
accelerator.wait_for_everyone()
unwrapped_model = accelerator.unwrap_model(model)
unwrapped_model.save_pretrained(
args.output_dir,
is_main_process=accelerator.is_main_process,
save_function=accelerator.save,
)
if accelerator.is_main_process:
image_processor.save_pretrained(args.output_dir)
api.upload_folder(
commit_message=f"Training in progress epoch {epoch}",
folder_path=args.output_dir,
repo_id=repo_id,
repo_type="model",
token=args.hub_token,
)
if completed_steps >= args.max_train_steps:
break
model.eval()
for step, batch in enumerate(eval_dataloader):
with torch.no_grad():
outputs = model(**batch)
predictions = outputs.logits.argmax(dim=-1)
predictions, references = accelerator.gather_for_metrics((predictions, batch["labels"]))
metric.add_batch(
predictions=predictions,
references=references,
)
eval_metric = metric.compute()
logger.info(f"epoch {epoch}: {eval_metric}")
if args.with_tracking:
accelerator.log(
{
"accuracy": eval_metric,
"train_loss": total_loss.item() / len(train_dataloader),
"epoch": epoch,
"step": completed_steps,
},
step=completed_steps,
)
if args.push_to_hub and epoch < args.num_train_epochs - 1:
accelerator.wait_for_everyone()
unwrapped_model = accelerator.unwrap_model(model)
unwrapped_model.save_pretrained(
args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
)
if accelerator.is_main_process:
image_processor.save_pretrained(args.output_dir)
api.upload_folder(
commit_message=f"Training in progress epoch {epoch}",
folder_path=args.output_dir,
repo_id=repo_id,
repo_type="model",
token=args.hub_token,
)
if args.checkpointing_steps == "epoch":
output_dir = f"epoch_{epoch}"
if args.output_dir is not None:
output_dir = os.path.join(args.output_dir, output_dir)
accelerator.save_state(output_dir)
if args.output_dir is not None:
accelerator.wait_for_everyone()
unwrapped_model = accelerator.unwrap_model(model)
unwrapped_model.save_pretrained(
args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
)
if accelerator.is_main_process:
image_processor.save_pretrained(args.output_dir)
if args.push_to_hub:
api.upload_folder(
commit_message="End of training",
folder_path=args.output_dir,
repo_id=repo_id,
repo_type="model",
token=args.hub_token,
)
all_results = {f"eval_{k}": v for k, v in eval_metric.items()}
with open(os.path.join(args.output_dir, "all_results.json"), "w") as f:
json.dump(all_results, f)
accelerator.wait_for_everyone()
accelerator.end_training()
if __name__ == "__main__":
main()