Spaces:
Paused
Paused
File size: 9,187 Bytes
ee6e328 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 |
<!--Copyright 2023 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.
โ ๏ธ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# ์ด๋ฏธ์ง ์บก์
๋[[image-captioning]]
[[open-in-colab]]
์ด๋ฏธ์ง ์บก์
๋(Image captioning)์ ์ฃผ์ด์ง ์ด๋ฏธ์ง์ ๋ํ ์บก์
์ ์์ธกํ๋ ์์
์
๋๋ค.
์ด๋ฏธ์ง ์บก์
๋์ ์๊ฐ ์ฅ์ ์ธ์ด ๋ค์ํ ์ํฉ์ ํ์ํ๋ ๋ฐ ๋์์ ์ค ์ ์๋๋ก ์๊ฐ ์ฅ์ ์ธ์ ๋ณด์กฐํ๋ ๋ฑ ์ค์ํ์์ ํํ ํ์ฉ๋ฉ๋๋ค.
๋ฐ๋ผ์ ์ด๋ฏธ์ง ์บก์
๋์ ์ด๋ฏธ์ง๋ฅผ ์ค๋ช
ํจ์ผ๋ก์จ ์ฌ๋๋ค์ ์ฝํ
์ธ ์ ๊ทผ์ฑ์ ๊ฐ์ ํ๋ ๋ฐ ๋์์ด ๋ฉ๋๋ค.
์ด ๊ฐ์ด๋์์๋ ์๊ฐํ ๋ด์ฉ์ ์๋์ ๊ฐ์ต๋๋ค:
* ์ด๋ฏธ์ง ์บก์
๋ ๋ชจ๋ธ์ ํ์ธํ๋ํฉ๋๋ค.
* ํ์ธํ๋๋ ๋ชจ๋ธ์ ์ถ๋ก ์ ์ฌ์ฉํฉ๋๋ค.
์์ํ๊ธฐ ์ ์ ํ์ํ ๋ชจ๋ ๋ผ์ด๋ธ๋ฌ๋ฆฌ๊ฐ ์ค์น๋์ด ์๋์ง ํ์ธํ์ธ์:
```bash
pip install transformers datasets evaluate -q
pip install jiwer -q
```
Hugging Face ๊ณ์ ์ ๋ก๊ทธ์ธํ๋ฉด ๋ชจ๋ธ์ ์
๋ก๋ํ๊ณ ์ปค๋ฎค๋ํฐ์ ๊ณต์ ํ ์ ์์ต๋๋ค.
ํ ํฐ์ ์
๋ ฅํ์ฌ ๋ก๊ทธ์ธํ์ธ์.
```python
from huggingface_hub import notebook_login
notebook_login()
```
## ํฌ์ผ๋ชฌ BLIP ์บก์
๋ฐ์ดํฐ์ธํธ ๊ฐ์ ธ์ค๊ธฐ[[load-the-pokmon-blip-captions-dataset]]
{์ด๋ฏธ์ง-์บก์
} ์์ผ๋ก ๊ตฌ์ฑ๋ ๋ฐ์ดํฐ์ธํธ๋ฅผ ๊ฐ์ ธ์ค๋ ค๋ฉด ๐ค Dataset ๋ผ์ด๋ธ๋ฌ๋ฆฌ๋ฅผ ์ฌ์ฉํฉ๋๋ค.
PyTorch์์ ์์ ๋ง์ ์ด๋ฏธ์ง ์บก์
๋ฐ์ดํฐ์ธํธ๋ฅผ ๋ง๋ค๋ ค๋ฉด [์ด ๋
ธํธ๋ถ](https://github.com/NielsRogge/Transformers-Tutorials/blob/master/GIT/Fine_tune_GIT_on_an_image_captioning_dataset.ipynb)์ ์ฐธ์กฐํ์ธ์.
```python
from datasets import load_dataset
ds = load_dataset("lambdalabs/pokemon-blip-captions")
ds
```
```bash
DatasetDict({
train: Dataset({
features: ['image', 'text'],
num_rows: 833
})
})
```
์ด ๋ฐ์ดํฐ์ธํธ๋ `image`์ `text`๋ผ๋ ๋ ํน์ฑ์ ๊ฐ์ง๊ณ ์์ต๋๋ค.
<Tip>
๋ง์ ์ด๋ฏธ์ง ์บก์
๋ฐ์ดํฐ์ธํธ์๋ ์ด๋ฏธ์ง๋น ์ฌ๋ฌ ๊ฐ์ ์บก์
์ด ํฌํจ๋์ด ์์ต๋๋ค.
์ด๋ฌํ ๊ฒฝ์ฐ, ์ผ๋ฐ์ ์ผ๋ก ํ์ต ์ค์ ์ฌ์ฉ ๊ฐ๋ฅํ ์บก์
์ค์์ ๋ฌด์์๋ก ์ํ์ ์ถ์ถํฉ๋๋ค.
</Tip>
[~datasets.Dataset.train_test_split] ๋ฉ์๋๋ฅผ ์ฌ์ฉํ์ฌ ๋ฐ์ดํฐ์ธํธ์ ํ์ต ๋ถํ ์ ํ์ต ๋ฐ ํ
์คํธ ์ธํธ๋ก ๋๋๋๋ค:
```python
ds = ds["train"].train_test_split(test_size=0.1)
train_ds = ds["train"]
test_ds = ds["test"]
```
ํ์ต ์ธํธ์ ์ํ ๋ช ๊ฐ๋ฅผ ์๊ฐํํด ๋ด
์๋ค.
Let's visualize a couple of samples from the training set.
```python
from textwrap import wrap
import matplotlib.pyplot as plt
import numpy as np
def plot_images(images, captions):
plt.figure(figsize=(20, 20))
for i in range(len(images)):
ax = plt.subplot(1, len(images), i + 1)
caption = captions[i]
caption = "\n".join(wrap(caption, 12))
plt.title(caption)
plt.imshow(images[i])
plt.axis("off")
sample_images_to_visualize = [np.array(train_ds[i]["image"]) for i in range(5)]
sample_captions = [train_ds[i]["text"] for i in range(5)]
plot_images(sample_images_to_visualize, sample_captions)
```
<div class="flex justify-center">
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/sample_training_images_image_cap.png" alt="Sample training images"/>
</div>
## ๋ฐ์ดํฐ์ธํธ ์ ์ฒ๋ฆฌ[[preprocess-the-dataset]]
๋ฐ์ดํฐ์ธํธ์๋ ์ด๋ฏธ์ง์ ํ
์คํธ๋ผ๋ ๋ ๊ฐ์ง ์์์ด ์๊ธฐ ๋๋ฌธ์, ์ ์ฒ๋ฆฌ ํ์ดํ๋ผ์ธ์์ ์ด๋ฏธ์ง์ ์บก์
์ ๋ชจ๋ ์ ์ฒ๋ฆฌํฉ๋๋ค.
์ ์ฒ๋ฆฌ ์์
์ ์ํด, ํ์ธํ๋ํ๋ ค๋ ๋ชจ๋ธ์ ์ฐ๊ฒฐ๋ ํ๋ก์ธ์ ํด๋์ค๋ฅผ ๊ฐ์ ธ์ต๋๋ค.
```python
from transformers import AutoProcessor
checkpoint = "microsoft/git-base"
processor = AutoProcessor.from_pretrained(checkpoint)
```
ํ๋ก์ธ์๋ ๋ด๋ถ์ ์ผ๋ก ํฌ๊ธฐ ์กฐ์ ๋ฐ ํฝ์
ํฌ๊ธฐ ์กฐ์ ์ ํฌํจํ ์ด๋ฏธ์ง ์ ์ฒ๋ฆฌ๋ฅผ ์ํํ๊ณ ์บก์
์ ํ ํฐํํฉ๋๋ค.
```python
def transforms(example_batch):
images = [x for x in example_batch["image"]]
captions = [x for x in example_batch["text"]]
inputs = processor(images=images, text=captions, padding="max_length")
inputs.update({"labels": inputs["input_ids"]})
return inputs
train_ds.set_transform(transforms)
test_ds.set_transform(transforms)
```
๋ฐ์ดํฐ์ธํธ๊ฐ ์ค๋น๋์์ผ๋ ์ด์ ํ์ธํ๋์ ์ํด ๋ชจ๋ธ์ ์ค์ ํ ์ ์์ต๋๋ค.
## ๊ธฐ๋ณธ ๋ชจ๋ธ ๊ฐ์ ธ์ค๊ธฐ[[load-a-base-model]]
["microsoft/git-base"](https://huggingface.co/microsoft/git-base)๋ฅผ [`AutoModelForCausalLM`](https://huggingface.co/docs/transformers/model_doc/auto#transformers.AutoModelForCausalLM) ๊ฐ์ฒด๋ก ๊ฐ์ ธ์ต๋๋ค.
```python
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained(checkpoint)
```
## ํ๊ฐ[[evaluate]]
์ด๋ฏธ์ง ์บก์
๋ชจ๋ธ์ ์ผ๋ฐ์ ์ผ๋ก [Rouge ์ ์](https://huggingface.co/spaces/evaluate-metric/rouge) ๋๋ [๋จ์ด ์ค๋ฅ์จ(Word Error Rate)](https://huggingface.co/spaces/evaluate-metric/wer)๋ก ํ๊ฐํฉ๋๋ค.
์ด ๊ฐ์ด๋์์๋ ๋จ์ด ์ค๋ฅ์จ(WER)์ ์ฌ์ฉํฉ๋๋ค.
์ด๋ฅผ ์ํด ๐ค Evaluate ๋ผ์ด๋ธ๋ฌ๋ฆฌ๋ฅผ ์ฌ์ฉํฉ๋๋ค.
WER์ ์ ์ฌ์ ์ ํ ์ฌํญ ๋ฐ ๊ธฐํ ๋ฌธ์ ์ ์ [์ด ๊ฐ์ด๋](https://huggingface.co/spaces/evaluate-metric/wer)๋ฅผ ์ฐธ์กฐํ์ธ์.
```python
from evaluate import load
import torch
wer = load("wer")
def compute_metrics(eval_pred):
logits, labels = eval_pred
predicted = logits.argmax(-1)
decoded_labels = processor.batch_decode(labels, skip_special_tokens=True)
decoded_predictions = processor.batch_decode(predicted, skip_special_tokens=True)
wer_score = wer.compute(predictions=decoded_predictions, references=decoded_labels)
return {"wer_score": wer_score}
```
## ํ์ต![[train!]]
์ด์ ๋ชจ๋ธ ํ์ธํ๋์ ์์ํ ์ค๋น๊ฐ ๋์์ต๋๋ค. ์ด๋ฅผ ์ํด ๐ค [`Trainer`]๋ฅผ ์ฌ์ฉํฉ๋๋ค.
๋จผ์ , [`TrainingArguments`]๋ฅผ ์ฌ์ฉํ์ฌ ํ์ต ์ธ์๋ฅผ ์ ์ํฉ๋๋ค.
```python
from transformers import TrainingArguments, Trainer
model_name = checkpoint.split("/")[1]
training_args = TrainingArguments(
output_dir=f"{model_name}-pokemon",
learning_rate=5e-5,
num_train_epochs=50,
fp16=True,
per_device_train_batch_size=32,
per_device_eval_batch_size=32,
gradient_accumulation_steps=2,
save_total_limit=3,
evaluation_strategy="steps",
eval_steps=50,
save_strategy="steps",
save_steps=50,
logging_steps=50,
remove_unused_columns=False,
push_to_hub=True,
label_names=["labels"],
load_best_model_at_end=True,
)
```
ํ์ต ์ธ์๋ฅผ ๋ฐ์ดํฐ์ธํธ, ๋ชจ๋ธ๊ณผ ํจ๊ป ๐ค Trainer์ ์ ๋ฌํฉ๋๋ค.
```python
trainer = Trainer(
model=model,
args=training_args,
train_dataset=train_ds,
eval_dataset=test_ds,
compute_metrics=compute_metrics,
)
```
ํ์ต์ ์์ํ๋ ค๋ฉด [`Trainer`] ๊ฐ์ฒด์์ [`~Trainer.train`]์ ํธ์ถํ๊ธฐ๋ง ํ๋ฉด ๋ฉ๋๋ค.
```python
trainer.train()
```
ํ์ต์ด ์งํ๋๋ฉด์ ํ์ต ์์ค์ด ์ํํ๊ฒ ๊ฐ์ํ๋ ๊ฒ์ ๋ณผ ์ ์์ต๋๋ค.
ํ์ต์ด ์๋ฃ๋๋ฉด ๋ชจ๋ ์ฌ๋์ด ๋ชจ๋ธ์ ์ฌ์ฉํ ์ ์๋๋ก [`~Trainer.push_to_hub`] ๋ฉ์๋๋ฅผ ์ฌ์ฉํ์ฌ ๋ชจ๋ธ์ ํ๋ธ์ ๊ณต์ ํ์ธ์:
```python
trainer.push_to_hub()
```
## ์ถ๋ก [[inference]]
`test_ds`์์ ์ํ ์ด๋ฏธ์ง๋ฅผ ๊ฐ์ ธ์ ๋ชจ๋ธ์ ํ
์คํธํฉ๋๋ค.
```python
from PIL import Image
import requests
url = "https://huggingface.co/datasets/sayakpaul/sample-datasets/resolve/main/pokemon.png"
image = Image.open(requests.get(url, stream=True).raw)
image
```
<div class="flex justify-center">
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/test_image_image_cap.png" alt="Test image"/>
</div>
๋ชจ๋ธ์ ์ฌ์ฉํ ์ด๋ฏธ์ง๋ฅผ ์ค๋นํฉ๋๋ค.
```python
device = "cuda" if torch.cuda.is_available() else "cpu"
inputs = processor(images=image, return_tensors="pt").to(device)
pixel_values = inputs.pixel_values
```
[`generate`]๋ฅผ ํธ์ถํ๊ณ ์์ธก์ ๋์ฝ๋ฉํฉ๋๋ค.
```python
generated_ids = model.generate(pixel_values=pixel_values, max_length=50)
generated_caption = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
print(generated_caption)
```
```bash
a drawing of a pink and blue pokemon
```
ํ์ธํ๋๋ ๋ชจ๋ธ์ด ๊ฝค ๊ด์ฐฎ์ ์บก์
์ ์์ฑํ ๊ฒ ๊ฐ์ต๋๋ค!
|