tarekziade HF Staff
Publish MIC usage, slim fixture and reviewed Transformers expectations
7aa363d verified |
Download integrations/transformers/usage.md from huggingface-mic/layoutlm-base-uncased-mic: direct link, hf CLI and curl.
- Browser
- Download file 3.84 kB
-
https://huggingface.co/huggingface-mic/layoutlm-base-uncased-mic/resolve/main/integrations/transformers/usage.md
- Command line
-
hf download hf://huggingface-mic/layoutlm-base-uncased-mic/integrations/transformers/usage.md
-
curl -L -o usage.md https://huggingface.co/huggingface-mic/layoutlm-base-uncased-mic/resolve/main/integrations/transformers/usage.md
3.84 kB
LayoutLM: masked language modeling with Transformers
This page is the Transformers contract of the LayoutLM MIC and also a usage example. doc-builder renders the runnable blocks below as documentation, and its pytest plugin runs them as written against microsoft/layoutlm-base-uncased. The MIC runner runs the same blocks against each slim fixture and each checkpoint in integration.json. Lines marked # doc-builder: hide only matter there: they swap in the pinned checkpoint or variant and print the result record.
Predict the masked word
import torch
from transformers import AutoModelForMaskedLM, AutoTokenizer, set_seed
import json # doc-builder: hide
import os # doc-builder: hide
model_id = "microsoft/layoutlm-base-uncased"
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
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForMaskedLM.from_pretrained(model_id).to(device)
if variant: # doc-builder: hide
del model # doc-builder: hide
model = AutoModelForMaskedLM.from_pretrained(model_id, **variant).to(device) # doc-builder: hide
# LayoutLM reads a word's bounding box (x0, y0, x1, y1) on a 0-1000 grid alongside its token id.
texts = [f"The capital of France is {tokenizer.mask_token}.", f"The capital of Germany is {tokenizer.mask_token}."]
inputs = tokenizer(texts, return_tensors="pt", padding=True)
# One box per token: the tokens laid out left to right on one line of the page.
length = inputs["input_ids"].shape[1]
x0 = torch.arange(length) * 50
boxes = torch.stack([x0, torch.zeros_like(x0), x0 + 50, torch.full_like(x0, 20)], dim=-1)
inputs["bbox"] = boxes.unsqueeze(0).expand(len(texts), -1, -1).clone()
inputs = inputs.to(device)
with torch.no_grad():
logits = model(**inputs).logits
masked = inputs["input_ids"] == tokenizer.mask_token_id
predicted = logits[masked].argmax(dim=-1)
print(tokenizer.convert_ids_to_tokens(predicted))
One training step
answers = ["paris", "berlin"]
answer_ids = tokenizer.convert_tokens_to_ids(answers)
labels = torch.full_like(inputs["input_ids"], -100)
labels[masked] = torch.tensor(answer_ids, device=device)
model.train()
optimizer = torch.optim.SGD(model.parameters(), lr=0.01)
loss = model(**inputs, labels=labels).loss
loss.backward()
before = {n: p.detach().cpu().clone() for n, p in model.named_parameters() if p.requires_grad} # 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 = ("layoutlm.encoder.layer.0.attention.self.query.weight", "layoutlm.encoder.layer.0.intermediate.dense.weight")
result = {
"runtime": {
"device": model.device.type,
"dtype": str(model.dtype).removeprefix("torch."),
"attn_implementation": model.config._attn_implementation,
},
"predicted_tokens": tokenizer.convert_ids_to_tokens(predicted),
"masked_logits": logits[masked][:, answer_ids].float().tolist(),
"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.detach().cpu(), before[n]) for n, p in model.named_parameters() if n in before),
},
}
print(json.dumps(result))