423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-12-16 20:00:36 +08:00
Code Issues Packages Projects Releases Wiki Activity

support batch infer in vllm

Browse Source
This commit is contained in:
hiyouga
2024-12-04 13:50:00 +00:00
parent dc78355002
commit 1324d158f9
29 changed files with 148 additions and 407 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
src/llamafactory/train/sft/trainer.py
View File

@@ -161,12 +161,9 @@ class CustomSeq2SeqTrainer(Seq2SeqTrainer):
preds[i] = np.concatenate((preds[i][pad_len[0] :], preds[i][: pad_len[0]]), axis=-1)
decoded_inputs = self.tokenizer.batch_decode(dataset["input_ids"], skip_special_tokens=True)
decoded_labels = self.tokenizer.batch_decode(labels, skip_special_tokens=True)
decoded_preds = self.tokenizer.batch_decode(preds, skip_special_tokens=True)
decoded_labels = self.tokenizer.batch_decode(labels, skip_special_tokens=True)
with open(output_prediction_file, "w", encoding="utf-8") as writer:
res: List[str] = []
for text, label, pred in zip(decoded_inputs, decoded_labels, decoded_preds):
res.append(json.dumps({"prompt": text, "label": label, "predict": pred}, ensure_ascii=False))
writer.write("\n".join(res))
with open(output_prediction_file, "w", encoding="utf-8") as f:
for text, pred, label in zip(decoded_inputs, decoded_preds, decoded_labels):
f.write(json.dumps({"prompt": text, "predict": pred, "label": label}, ensure_ascii=False) + "\n")