AttributeError: 'HuskyQFormerFlashAttention2' object has no attribute 'embed_size'

[upcly123]

when i run the demo/test.py，i got this error.
File "/data/h/anaconda3/envs/robohusky/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1511, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/data/h/anaconda3/envs/robohusky/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1520, in _call_impl
return forward_call(*args, **kwargs)
File "/data/h/workspace/EmbodiedGPT_Pytorch-main/robohusky/model/modeling_husky_embody2.py", line 888, in forward
context_layer = attn_output.reshape(bsz, tgt_len, self.embed_size).contiguous()
File "/data/h/anaconda3/envs/robohusky/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1688, in getattr
raise AttributeError(f"'{type(self).name}' object has no attribute '{name}'")
AttributeError: 'HuskyQFormerFlashAttention2' object has no attribute 'embed_size'

i checked the modeling_husky_embody2.py, and find that the HuskyQFormerFlashAttention2 class really didn't have attribute 'embed_size'

[FanZhang91]

+1

[LongXinKou]

+1

[LongXinKou]

+1
it works by replacing embed_size with embed_dim in modeling_husky_embody2.py line 838.
context_layer = attn_output.reshape(bsz, tgt_len, embed_dim).contiguous()

[Friedrichqi]

+1
it works by replacing embed_size with embed_dim in modeling_husky_embody2.py line 838.
context_layer = attn_output.reshape(bsz, tgt_len, embed_dim).contiguous()

It helps! Thank you!