Hello,
Thank you for your work!
In our project, we trained an AttentionXML model on 4 GPUs but are now trying to load it in an environment where only one GPU is available.
After modifying the code according to this issue #34, we get the following error:
RuntimeError: Error(s) in loading state_dict for ModuleDict:
Missing key(s) in state_dict: "Network.attention.attention.weight".
Unexpected key(s) in state_dict: "AttentionWeights.emb.0.weight", "AttentionWeights.emb.1.weight", "AttentionWeights.emb.2.weight".
This error occurs only when prediction on Level-1 is performed. No error occurs in the 4 GPU environment.
We have already tried to concatenate "AttentionWeights.emb.0.weight", "AttentionWeights.emb.1.weight", "AttentionWeights.emb.2.weight" but they seem to have a different dimension than required.
Do you have any idea how we can get this to work?
Best wishes, Katja
Hello,
Thank you for your work!
In our project, we trained an AttentionXML model on 4 GPUs but are now trying to load it in an environment where only one GPU is available.
After modifying the code according to this issue #34, we get the following error:
This error occurs only when prediction on Level-1 is performed. No error occurs in the 4 GPU environment.
We have already tried to concatenate
"AttentionWeights.emb.0.weight", "AttentionWeights.emb.1.weight", "AttentionWeights.emb.2.weight"but they seem to have a different dimension than required.Do you have any idea how we can get this to work?
Best wishes, Katja