File size: 3,942 Bytes
cb2e2ab | 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 | # CLIP: zero-shot image classification with Transformers
This page is the Transformers contract of the CLIP MIC and also a usage example. [doc-builder](https://github.com/huggingface/doc-builder/blob/main/docs/runnable-code-blocks.md) renders the `runnable` blocks below as documentation, and its pytest plugin runs them as written against `openai/clip-vit-base-patch32`. The MIC runner runs the same blocks against each slim fixture and each checkpoint in [`integration.json`](../../integration.json). Lines marked `# doc-builder: hide` only matter there: they swap in the pinned checkpoint or variant and print the result record.
## Match images to captions
```py runnable:clip
import torch
from PIL import Image
from transformers import CLIPModel, CLIPProcessor, set_seed
import json # doc-builder: hide
import os # doc-builder: hide
model_id = "openai/clip-vit-base-patch32"
model_id = os.environ.get("MIC_MODEL", model_id) # doc-builder: hide
variant = json.loads(os.environ.get("MIC_MODEL_KWARGS", "{}")) # doc-builder: hide
device = "cuda" if torch.cuda.is_available() else "cpu"
set_seed(0) # seeds torch, NumPy, and Python: loading, dropout, and data augmentation draw random numbers
processor = CLIPProcessor.from_pretrained(model_id)
model = CLIPModel.from_pretrained(model_id).to(device)
if variant: # doc-builder: hide
model = CLIPModel.from_pretrained(model_id, **variant).to(device)
colors = {"red": (255, 0, 0), "green": (0, 255, 0), "blue": (0, 0, 255), "yellow": (255, 255, 0)}
images = [Image.new("RGB", (256, 256), rgb) for rgb in colors.values()]
captions = [f"a photo of a {name} square" for name in colors]
inputs = processor(text=captions, images=images, return_tensors="pt", padding=True).to(device, model.dtype) # floating-point inputs follow the model's dtype
with torch.no_grad():
logits_per_image = model(**inputs).logits_per_image
best = logits_per_image.softmax(dim=1).argmax(dim=1)
for name, index in zip(colors, best.tolist()):
print(f"{name} image -> {captions[index]!r}")
```
## Pass options to one modality
A processor call can route options to one of its inputs, here padding to the text only. Integrations also read the preprocessing configuration without building a processor.
```py runnable:clip:3
from transformers.models.auto.image_processing_auto import get_image_processor_config
text_only = processor(text=captions, images=images, return_tensors="pt", text_kwargs={"padding": True})
print(text_only["input_ids"].shape)
image_config = get_image_processor_config(model_id)
print(sorted(image_config))
```
## One training step
```py runnable:clip:3
model.train()
optimizer = torch.optim.SGD(model.parameters(), lr=0.01)
loss = model(**inputs, return_loss=True).loss
loss.backward()
before = {name: p.detach().clone() for name, p in model.named_parameters()} # doc-builder: hide
optimizer.step()
print(f"loss: {loss.item():.4f}")
if os.environ.get("MIC_RUNNER"): # doc-builder: hide
named = dict(model.named_parameters())
checked = ("visual_projection.weight", "text_projection.weight")
result = {
"runtime": {
"device": model.device.type,
"dtype": str(model.dtype).removeprefix("torch."),
"attn_implementation": model.config._attn_implementation,
},
"image_to_text": best.tolist(),
"text_to_image": logits_per_image.argmax(dim=0).tolist(),
"logits_per_image": logits_per_image.float().tolist(),
"text_kwargs_input_ids": text_only["input_ids"].tolist(),
"image_processor_config_keys": sorted(image_config),
"training": {
"loss": loss.item(),
"gradients": {n: bool(named[n].grad is not None and torch.count_nonzero(named[n].grad) > 0 and torch.isfinite(named[n].grad).all()) for n in checked},
"updated": any(not torch.equal(p, before[n]) for n, p in named.items()),
},
}
print(json.dumps(result))
```
|