requirements

[blurLake]

Hi,

I have been trying to reproduce the results in this repo. But it seems I am having some version issue for some packages, like pytorch_lightning. Is it possible to add a requirement file about all the packages that work together?

Thank you very much!

[ruizcrp]

We had the same issue. We cannot replicate it as we are running against several issues of pytorch and do not know which versions were used.

[ruizcrp]

We made it. pytorch_lightning==0.10.0 and torch==1.7.1 (in our case because of CUDA 10.1 restraints) (also install sentencepiece)

In train.py you have to delete the line parser.add_argument("--early_stop_callback", default=False) because this argument has been put twice.

Then in module.py change line 96 so that the parameter is called labels instead of lm_labels thus: labels=lm_labels,

[blurLake]

In line 79 of train.py, the author mentioned pytorch_lightning 0.8.4. With torch==1.7.1, we have another working combination.

I have also rename lm_labels to labels in module.py.

But now, I have two problems. 1. After training for 25 epochs, the success rate on dev dataset is still below 10%. 2. When I load the trained ckpt (from https://onebigdatabag.blob.core.windows.net/shared/base_gated_e09_0.02626.ckpt), the following error pops out which seems like a pytorch_lightning issue. `--------------------------------------------------------------------------- RuntimeError Traceback (most recent call last) in 49 50 print("Loading T5FinalTuner pretrained model...", end="") ---> 51 model = SeqGenSQL.load_from_checkpoint(args.ckpt_path) 52 print("Done!") 53

~/.local/lib/python3.8/site-packages/pytorch_lightning/core/saving.py in load_from_checkpoint(cls, checkpoint_path, map_location, hparams_file, tags_csv, *args, **kwargs) 167 checkpoint[cls.CHECKPOINT_HYPER_PARAMS_KEY].update(kwargs) 168 --> 169 model = cls._load_model_state(checkpoint, *args, **kwargs) 170 return model 171

~/.local/lib/python3.8/site-packages/pytorch_lightning/core/saving.py in _load_model_state(cls, checkpoint, *cls_args, **cls_kwargs) 205 model = cls(*cls_args, **cls_kwargs) 206 # load the state_dict on the model automatically --> 207 model.load_state_dict(checkpoint['state_dict']) 208 209 # give model a chance to load something

/usr/local/lib/python3.8/site-packages/torch/nn/modules/module.py in load_state_dict(self, state_dict, strict) 1049 1050 if len(error_msgs) > 0: -> 1051 raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format( 1052 self.class.name, "\n\t".join(error_msgs))) 1053 return _IncompatibleKeys(missing_keys, unexpected_keys)

RuntimeError: Error(s) in loading state_dict for SeqGenSQL: Unexpected key(s) in state_dict: "model.decoder.block.0.layer.1.EncDecAttention.relative_attention_bias.weight". `