File size: 5,350 Bytes
b1fe7e3 | 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 | # ๐ฝ๏ธ ์์ ์ด๋ฏธ์ง ๋ถ๋ฅ ํ๋ก์ ํธ ๊ฐ์ด๋
## ๐ ํ์ผ ๊ตฌ์ฑ
| ํ์ผ | ์ญํ | ํ์? |
|------|------|-------|
| `app.py` | Gradio ์น ๋ฐ๋ชจ | โ
Space ํ์ |
| `README.md` | Space ์ค์ (YAML) | โ
Space ํ์ |
| `requirements.txt` | ์์กด์ฑ | โ
Space ํ์ |
| `configuration_myresnet.py` | MyResNet Config | ์ต์
B์์๋ง ํ์ |
| `modeling_myresnet.py` | MyResNet ๋ชจ๋ธ | ์ต์
B์์๋ง ํ์ |
| `train_food.py` | Food-101 ํ์ต ์คํฌ๋ฆฝํธ | ๋ก์ปฌ ํ์ต์ฉ |
## ๐ฏ ๋ ๊ฐ์ง ์ฌ์ฉ ์๋๋ฆฌ์ค
### ์๋๋ฆฌ์ค 1: ํ์ต ์์ด ๋ฐ๋ก ๋ฐ๋ชจ ์คํ (์ถ์ฒ) โญ
`app.py`๋ ๊ธฐ๋ณธ์ ์ผ๋ก **ํ๊น
ํ์ด์ค ํ๋ธ์ ๊ณต๊ฐ ๋ชจ๋ธ `nateraw/food`**๋ฅผ ์ฌ์ฉํฉ๋๋ค.
์ด ๋ชจ๋ธ์ ์ด๋ฏธ Food-101์์ ํ์ต๋์ด ์ฝ 90% ์ ํ๋๋ฅผ ๋ณด์ฌ์.
**Space ์
๋ก๋ ํ์ผ (3๊ฐ๋ฉด ์ถฉ๋ถ):**
1. `app.py`
2. `README.md`
3. `requirements.txt`
์ฆ, ์ด ์๋๋ฆฌ์ค์์๋ `configuration_myresnet.py`, `modeling_myresnet.py`๋ฅผ
Space์ ์ฌ๋ฆด ํ์๊ฐ ์์ต๋๋ค!
### ์๋๋ฆฌ์ค 2: ์์ฒด ํ์ตํ MyResNet ์ฌ์ฉ
์ง์ ResNet์ ํ์ต์์ผ์ ๊ทธ ๋ชจ๋ธ๋ก ๋ฐ๋ชจ๋ฅผ ๋๋ฆฌ๊ณ ์ถ๋ค๋ฉด:
**Step 1: ๋ก์ปฌ์์ ํ์ต**
```bash
pip install torch torchvision transformers datasets accelerate
python train_food.py
```
โ `./my-resnet18-food101/` ํด๋์ ํ์ต๋ ๋ชจ๋ธ ์ ์ฅ๋จ
**Step 2: ๋ชจ๋ธ ํ๋ธ์ ์
๋ก๋**
```bash
huggingface-cli login
python -c "
from modeling_myresnet import MyResNetForImageClassification
from configuration_myresnet import MyResNetConfig
MyResNetConfig.register_for_auto_class()
MyResNetForImageClassification.register_for_auto_class('AutoModelForImageClassification')
model = MyResNetForImageClassification.from_pretrained('./my-resnet18-food101')
model.push_to_hub('your-username/my-resnet18-food101')
"
```
**Step 3: Space `app.py`์์ ์ต์
์ ํ**
```python
# ์ต์
A ๋ถ๋ถ์ ์ฃผ์ ์ฒ๋ฆฌ
# from transformers import AutoImageProcessor, AutoModelForImageClassification
# ...
# ์ต์
B ์ฃผ์ ํด์
from configuration_myresnet import MyResNetConfig
from modeling_myresnet import MyResNetForImageClassification
MODEL_ID = "your-username/my-resnet18-food101"
model = MyResNetForImageClassification.from_pretrained(MODEL_ID)
model.eval()
device = "cuda" if torch.cuda.is_available() else "cpu"
model = model.to(device)
id2label = model.config.id2label
```
**Step 4: ์ ์ฒ๋ฆฌ ํจ์๋ ์๋ ๋ฒ์ ์ผ๋ก ๋ณ๊ฒฝ**
```python
from torchvision.transforms import Compose, Resize, ToTensor, Normalize
_transform = Compose([
Resize((224, 224)),
ToTensor(),
Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
])
def preprocess(image):
return _transform(image.convert("RGB")).unsqueeze(0).to(device)
```
## ๐ Food-101์ 101๊ฐ ํด๋์ค
```
apple_pie, baby_back_ribs, baklava, beef_carpaccio, beef_tartare,
beet_salad, beignets, bibimbap, bread_pudding, breakfast_burrito,
bruschetta, caesar_salad, cannoli, caprese_salad, carrot_cake,
ceviche, cheesecake, cheese_plate, chicken_curry, chicken_quesadilla,
chicken_wings, chocolate_cake, chocolate_mousse, churros, clam_chowder,
club_sandwich, crab_cakes, creme_brulee, croque_madame, cup_cakes,
deviled_eggs, donuts, dumplings, edamame, eggs_benedict,
escargots, falafel, filet_mignon, fish_and_chips, foie_gras,
french_fries, french_onion_soup, french_toast, fried_calamari, fried_rice,
frozen_yogurt, garlic_bread, gnocchi, greek_salad, grilled_cheese_sandwich,
grilled_salmon, guacamole, gyoza, hamburger, hot_and_sour_soup,
hot_dog, huevos_rancheros, hummus, ice_cream, lasagna,
lobster_bisque, lobster_roll_sandwich, macaroni_and_cheese, macarons, miso_soup,
mussels, nachos, omelette, onion_rings, oysters,
pad_thai, paella, pancakes, panna_cotta, peking_duck,
pho, pizza, pork_chop, poutine, prime_rib,
pulled_pork_sandwich, ramen, ravioli, red_velvet_cake, risotto,
samosa, sashimi, scallops, seaweed_salad, shrimp_and_grits,
spaghetti_bolognese, spaghetti_carbonara, spring_rolls, steak, strawberry_shortcake,
sushi, tacos, takoyaki, tiramisu, tuna_tartare, waffles
```
**์ฌ๋ฏธ์๋ ์ฌ์ค**: ๐ฅ **bibimbap (๋น๋น๋ฐฅ)**, ๐ฃ **sushi**, ๐ฅ **gyoza (๋ง๋)**,
๐ **ramen**, ๐ฎ **tacos** ๋ฑ ํ์/์ผ์/์ค์/๋ฉ์์นธ ๋ฑ ๋ค์ํ ๊ตญ๊ฐ ์๋ฆฌ๊ฐ ํฌํจ๋์ด ์์ด์!
## ๐ก ํ์ต ์๊ฐ/์ ํ๋ ์ฐธ๊ณ
| ๋ฐฉ์ | ์๊ฐ (GPU 1์ฅ) | ์ ํ๋ |
|------|---------------|--------|
| Scratch ํ์ต (30 epochs) | ์ฝ 6-10์๊ฐ | ์ฝ 60-70% |
| Pretrained + Fine-tuning (10 epochs) | ์ฝ 2-4์๊ฐ | ์ฝ 80-85% |
| ๊ณต๊ฐ ๋ชจ๋ธ `nateraw/food` ์ฌ์ฉ | 0 | ~90% |
**์ด๋ณด์ ์ถ์ฒ**: ์๋๋ฆฌ์ค 1 (๊ณต๊ฐ ๋ชจ๋ธ ์ฌ์ฉ)๋ก ๋จผ์ Space๋ฅผ ๋์๋ณด๊ณ ,
๋์ค์ ๊ด์ฌ ์์ผ๋ฉด ์๋๋ฆฌ์ค 2๋ก ์ง์ ํ์ตํด๋ณด์ธ์!
## ๐ ์์ฃผ ๊ฒช๋ ๋ฌธ์
### 1. ํ์ต ์ OutOfMemory
```python
per_device_train_batch_size=64 # 32 ๋๋ 16์ผ๋ก ์ค์ด๊ธฐ
gradient_accumulation_steps=2 # ์ด๊ฑธ ๋์ ์ถ๊ฐ
```
### 2. Food-101 ๋ค์ด๋ก๋๊ฐ ๋๋ฆผ
- Food-101์ ์ฝ 5GB์
๋๋ค
- ์ฒซ ๋ค์ด๋ก๋ ํ์๋ `~/.cache/huggingface/datasets/`์ ์บ์๋จ
- ๋ค์ด๋ก๋๋ ํ ๋ฒ๋ง ํ๋ฉด ๋ฉ๋๋ค
### 3. "KeyError: '0'" ์๋ฌ
- CIFAR ๋ฒ์ ๊ณผ ๋์ผํ ๋ฌธ์ . `id2label[str(i)]` โ `id2label[i]`๋ก ์์
- ์ด ํ๋ก์ ํธ์ `app.py`๋ ์ด๋ฏธ ์ฌ๋ฐ๋ฅด๊ฒ ๋์ด ์์
|