Closed zluw1117 closed 1 year ago
Hi there. I have saw your great improvement on NL2SQL task vi STAR. Do you still have the final_generation.json file? Thanks.
Thank you for your interest, unfortunately we do not save this intermediate data. You can follow the instructions to regenerate them, hopefully this helps.
, unfortunately we do not save this intermediate data. You can follow the instructions to regenerate them, hopefully this helps.
When running the eval.py file, I encountered an error, it seems that the checkpoint-epoch-10.0 model mentioned in the README does not relative weights:
│ /data/sunzheng/project/DAMO-ConvAI/star/data_systhesis/snowball/eval.py:58 │
│ in <module> │
│ │
│ 55 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu") │
│ 56 generator = RelogicModel('facebook/bart-large').to(device) │
│ 57 generator_param = torch.load(generator_path) │
│ ❱ 58 generator.load_state_dict(generator_param) │
│ 59 logger.info("Successfully loaded generator") │
│ 60 │
│ 61 data_collator = DataCollatorForRelogic(tokenizer=tokenizer) │
│ │
│ /home/cv/anaconda3/lib/python3.8/site-packages/torch/nn/modules/module.py:16 │
│ 04 in load_state_dict │
│ │
│ 1601 │ │ │ │ │ │ ', '.join('"{}"'.format(k) for k in missing_k │
│ 1602 │ │ │
│ 1603 │ │ if len(error_msgs) > 0: │
│ ❱ 1604 │ │ │ raise RuntimeError('Error(s) in loading state_dict for {} │
│ 1605 │ │ │ │ │ │ │ self.__class__.__name__, "\n\t".join(e │
│ 1606 │ │ return _IncompatibleKeys(missing_keys, unexpected_keys) │
│ 1607 │
╰──────────────────────────────────────────────────────────────────────────────╯
RuntimeError: Error(s) in loading state_dict for RelogicModel:
Missing key(s) in state_dict: "bert.lm_head.weight".
, unfortunately we do not save this intermediate data. You can follow the instructions to regenerate them, hopefully this helps.
When running the eval.py file, I encountered an error, it seems that the checkpoint-epoch-10.0 model mentioned in the README does not relative weights:
│ /data/sunzheng/project/DAMO-ConvAI/star/data_systhesis/snowball/eval.py:58 │ │ in <module> │ │ │ │ 55 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu") │ │ 56 generator = RelogicModel('facebook/bart-large').to(device) │ │ 57 generator_param = torch.load(generator_path) │ │ ❱ 58 generator.load_state_dict(generator_param) │ │ 59 logger.info("Successfully loaded generator") │ │ 60 │ │ 61 data_collator = DataCollatorForRelogic(tokenizer=tokenizer) │ │ │ │ /home/cv/anaconda3/lib/python3.8/site-packages/torch/nn/modules/module.py:16 │ │ 04 in load_state_dict │ │ │ │ 1601 │ │ │ │ │ │ ', '.join('"{}"'.format(k) for k in missing_k │ │ 1602 │ │ │ │ 1603 │ │ if len(error_msgs) > 0: │ │ ❱ 1604 │ │ │ raise RuntimeError('Error(s) in loading state_dict for {} │ │ 1605 │ │ │ │ │ │ │ self.__class__.__name__, "\n\t".join(e │ │ 1606 │ │ return _IncompatibleKeys(missing_keys, unexpected_keys) │ │ 1607 │ ╰──────────────────────────────────────────────────────────────────────────────╯ RuntimeError: Error(s) in loading state_dict for RelogicModel: Missing key(s) in state_dict: "bert.lm_head.weight".
Hello, we cannot reproduce the situation you described, it works fine on our side. Perhaps you can check the runtime environment and whether the checkpoint file is complete. Or provide more information. Hope this helps.
Hi there. I have saw your great improvement on NL2SQL task vi STAR. Do you still have the final_generation.json file? Thanks.