lmdeploy: 'language_model.model.layers.0.feed_forward.w1.weight'

#1
by BenjaminTT - opened

File ~/anaconda3/envs/deepfake/lib/python3.8/site-packages/lmdeploy/turbomind/deploy/source_model/internlm2.py:62, in InternLM2Reader._ffn(self, i, kind)
60 result = []
61 for key in ['w1', 'w2', 'w3']:
---> 62 tensor = self.params[
63 f'{self.attn_layer_prefix}.{i}.feed_forward.{key}.{kind}']
64 result.append(tensor)
65 return (*result, )

KeyError: 'language_model.model.layers.0.feed_forward.w1.weight'

OpenGVLab org

Hi, this bug has been fixed, take a look at the code in the new readme.

czczup changed discussion status to closed

Sign up or log in to comment