not found models/model.zh.seed-1

[canghaiyunfan]

感谢作者开源代码，但运行pred_zh.sh 过程中遇到了问题

RuntimeError: File models/model.zh.seed-1 unavailable. Please try other sources.

想请教下脚本中用到的文件 去哪下载，例如
gec_path:=models/model.zh.seed-1
lm_path:=uer/gpt2-chinese-cluecorpussmall
ged_path:=models/ged_model.zh.seed-1
mucgec.dev.collapsed

[Jacob-Zhou]

您好，GEC 和 GED 文件可以参考 Download Trained Models 章节下载，其中 GEC 和 GED 文件的仓库名 HQZhou/bart-large-chinese-gec 和 HQZhou/bart-large-chinese-ged, 下载完后请参考 Run 设置路径。

数据集文件可以到 MuCGEC 中下载，然后参考 Setup 修改格式。

[canghaiyunfan]

@Jacob-Zhou 按照要求下载文件后仍报错

config: configs/bart.ini
path: /tmp/tmp.KdmvoLL3Sr
suffix: �
gec_path: models/gecdi/bart-large-gec/model
lm_path: uer/gpt2-chinese-cluecorpussmall
lm_alpha: 0.0
lm_beta: 0
ged_path: models/gecdi/bart-large-ged/model
ged_alpha: 0.0
ged_beta: 0
devices: 0
batch: 1000
beam: 12
dataset: mucgec.dev
data: data/cgec/mucgec.dev.collapsed
gold: data/cgec/mucgec.dev.m2
pred: models/gecdi/bart-large-gec/pred/mucgec.dev/baseline/mucgec.dev.beam-12.pred
Traceback (most recent call last):
File "gecdi/scripts/predict_utils/split_discourse.py", line 25, in
tokenize_func = Seq2SeqParser.load(args.path).SRC.tokenize
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "gecdi/supar/parser.py", line 495, in load
model.load_state_dict(state['state_dict'], False)
File "/miniconda3/envs/bert/lib/python3.11/site-packages/torch/nn/modules/module.py", line 2152, in load_state_dict
raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format(
RuntimeError: Error(s) in loading state_dict for Seq2SeqModel:
size mismatch for model.shared.weight: copying a param with shape torch.Size([50265, 1024]) from checkpoint, the shape in current model is torch.Size([50264, 1024]).
size mismatch for model.encoder.embed_tokens.weight: copying a param with shape torch.Size([50265, 1024]) from checkpoint, the shape in current model is torch.Size([50264, 1024]).
size mismatch for model.decoder.embed_tokens.weight: copying a param with shape torch.Size([50265, 1024]) from checkpoint, the shape in current model is torch.Size([50264, 1024]).
size mismatch for encoder.embed_tokens.weight: copying a param with shape torch.Size([50265, 1024]) from checkpoint, the shape in current model is torch.Size([50264, 1024]).
size mismatch for decoder.embed_tokens.weight: copying a param with shape torch.Size([50265, 1024]) from checkpoint, the shape in current model is torch.Size([50264, 1024]).
size mismatch for classifier.weight: copying a param with shape torch.Size([50265, 1024]) from checkpoint, the shape in current model is torch.Size([50264, 1024]).

[Jacob-Zhou]

您好我这边尝试复现一下这个问题，可能需要点时间。