Open tonyageall opened 6 months ago
Hi everyone, first thanks for your work!
I have the same problem as @tonyageall. This error indicates a problem with the structure of the neural network. There is no correspondence between the keys expected, and the keys found by the downloaded model. I try with different models but, I have the same error. How can we fix that? Can anyone help us?
Thanks for your answers.
same question
I also have the same problems ,by use cuda 11.8 torch 2.0.0 openfold 2.0 import esm import torch model = esm.pretrained.esmfold_v1() ################################### RuntimeError: Keys 'trunk.structure_module.ipa.linear_q_points.linear.weight, trunk.structure_module.ipa.linear_kv_points.linear.bias, trunk.structure_module.ipa.linear_q_points.linear.bias, trunk.structure_module.ipa.linear_kv_points.linear.weight' are missing.
Thanks for your answers.
Same issue:
Traceback (most recent call last):
File "./esm-fold.py", line 5, in <module>
model = esm.pretrained.esmfold_v1()
File "/home/ark/miniconda3/envs/esmfold3/lib/python3.7/site-packages/esm/pretrained.py", line 420, in esmfold_v1
return esm.esmfold.v1.pretrained.esmfold_v1()
File "/home/ark/miniconda3/envs/esmfold3/lib/python3.7/site-packages/esm/esmfold/v1/pretrained.py", line 54, in esmfold_v1
return _load_model("esmfold_3B_v1")
File "/home/ark/miniconda3/envs/esmfold3/lib/python3.7/site-packages/esm/esmfold/v1/pretrained.py", line 29, in _load_model
raise RuntimeError(f"Keys '{', '.join(missing_essential_keys)}' are missing.")
RuntimeError: Keys 'trunk.structure_module.ipa.linear_q_points.linear.bias, trunk.structure_module.ipa.linear_kv_points.linear.bias, trunk.structure_module.ipa.linear_kv_points.linear.weight, trunk.structure_module.ipa.linear_q_points.linear.weight' are missing.
You may refer to issue: 435 for any specific solutions.
I'm using Pytorch CPU-only, on a ubuntu 18 OS. I'm trying to run ESM-fold. I'm getting the following error:
RuntimeError: Keys 'trunk.structure_module.ipa.linear_kv_points.linear.bias, tru nk.structure_module.ipa.linear_q_points.linear.weight, trunk.structure_module.ip a.linear_q_points.linear.bias, trunk.structure_module.ipa.linear_kv_points.linea r.weight' are missing.
Is this an issue with Pytorch? Or some other dependency?