Buckets:
| """Load a trained adapter checkpoint and classify one message/policy pair (either backbone).""" | |
| import argparse | |
| import json | |
| from training.common import LABELS | |
| from training.engine import load_trained_classifier, predict_logits, softmax_rows | |
| def main(): | |
| parser = argparse.ArgumentParser(description=__doc__) | |
| parser.add_argument("--checkpoint", required=True, help="final/ or checkpoint-*of4-* directory") | |
| parser.add_argument("--message", required=True) | |
| parser.add_argument("--policy", required=True) | |
| args = parser.parse_args() | |
| import torch | |
| device = "cuda" if torch.cuda.is_available() else "cpu" | |
| model, tokenizer, recipe = load_trained_classifier(args.checkpoint, device) | |
| row = {"message": args.message, "policy": args.policy} | |
| probabilities = softmax_rows(predict_logits(model, tokenizer, [row], recipe, 1, device))[0] | |
| print(json.dumps({"label": LABELS[int(probabilities.argmax())], | |
| "probabilities": dict(zip(LABELS, map(float, probabilities)))}, indent=2)) | |
| if __name__ == "__main__": | |
| main() | |
Xet Storage Details
- Size:
- 1.1 kB
- Xet hash:
- 0268e0e96fe03e0556f9d4de13b86e052f02b7388b6597ab39332fbd1d5de941
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.