tarekziade HF Staff
Publish MIC usage, slim fixture and reviewed Transformers expectations
e19ae54 verified |
Download integrations/transformers/usage.md from huggingface-mic/bigbird-roberta-base-mic: direct link, hf CLI and curl.
- Browser
- Download file 3.81 kB
-
https://huggingface.co/huggingface-mic/bigbird-roberta-base-mic/resolve/main/integrations/transformers/usage.md
- Command line
-
hf download hf://huggingface-mic/bigbird-roberta-base-mic/integrations/transformers/usage.md
-
curl -L -o usage.md https://huggingface.co/huggingface-mic/bigbird-roberta-base-mic/resolve/main/integrations/transformers/usage.md
3.81 kB
| # BigBird: masked language modeling and next-sentence prediction | |
| This page is the Transformers contract of the BigBird 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 `google/bigbird-roberta-base`. 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. | |
| ## Masked language modeling and next-sentence prediction | |
| ```py runnable:pretraining | |
| import torch | |
| from transformers import AutoModelForPreTraining, AutoTokenizer, set_seed | |
| import json # doc-builder: hide | |
| import os # doc-builder: hide | |
| model_id = "google/bigbird-roberta-base" | |
| 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 = AutoModelForPreTraining.from_pretrained(model_id).to(device) | |
| if variant: # doc-builder: hide | |
| del model # doc-builder: hide | |
| model = AutoModelForPreTraining.from_pretrained(model_id, **variant).to(device) # doc-builder: hide | |
| sentences = [ | |
| "The quick brown fox jumps over the lazy dog.", | |
| "Transformers are powerful models for natural language processing.", | |
| ] | |
| inputs = tokenizer(sentences, return_tensors="pt", padding=True).to(device) | |
| with torch.no_grad(): | |
| outputs = model(**inputs) | |
| prediction_logits = outputs.prediction_logits | |
| seq_relationship_logits = outputs.seq_relationship_logits | |
| # Probe a few logits: first token of each sequence and the NSP logits | |
| probe = prediction_logits[:, 0, :5].float().tolist() | |
| nsp_probe = seq_relationship_logits.float().tolist() | |
| print("prediction logits (first token, first 5):", probe) | |
| print("seq relationship logits:", nsp_probe) | |
| ``` | |
| ## One training step | |
| ```py runnable:pretraining:2 | |
| # Create labels for MLM: copy input_ids and mask padded positions | |
| labels = inputs["input_ids"].clone() | |
| labels[inputs["attention_mask"] == 0] = -100 | |
| # NSP labels: 0 for continuation, 1 for random (here we use 0 for all) | |
| next_sentence_label = torch.zeros(inputs["input_ids"].shape[0], dtype=torch.long, device=device) | |
| model.train() | |
| optimizer = torch.optim.SGD(model.parameters(), lr=0.01) | |
| loss = model(**inputs, labels=labels, next_sentence_label=next_sentence_label).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 = ("bert.encoder.layer.0.attention.self.query.weight", "bert.encoder.layer.0.intermediate.dense.weight") | |
| result = { | |
| "runtime": { | |
| "device": model.device.type, | |
| "dtype": str(model.dtype).removeprefix("torch."), | |
| "attn_implementation": model.config._attn_implementation, | |
| }, | |
| "prediction_logits": probe, | |
| "seq_relationship_logits": nsp_probe, | |
| "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)) | |
| ``` | |