Chloe-mxxxxc

Hello, I am also reproducing this work. Have you ever encountered this situation, that is, the model is not loaded with kv during eval, and the inference effect is the...

> train synthetic, eval synthetic, Meta-Llama-3-8B, 2w iterations: > > vary kb_size, kb generation: ○ 10：{'rouge1': 0.0659, 'rouge2': 0.0, 'rougeL': 0.0597, 'rougeLsum': 0.0597, 'bert_score_precision': 0.3326, 'bert_score_recall': 0.3724, 'bert_score_f1': 0.3505, 'mem_cost':...

> > Hello, have you run eval.py generation kb mode? The output effect is very poor. Have you fixed the bug in eval.py? > > I have found the reason....

>  Hi, has this problem been solved? I have the same problem.