search

gzBiomedical / EpiScan

Antibody-Specific Epitope Mapping
10 stars 2 forks source link

Seq_final.pth not saved correctly? #1

Open QUEST2179 opened 2 months ago

QUEST2179 commented 2 months ago

python ./EpiScan/commands/epimapping.py --test ./dataProcess/public/public_sep_testAg.tsv --embedding ./dataProcess/public/fasta/DB1.h5 --outfile ./dataProcess/public/public_sep_testAg.out --device 0 [2024-09-13-09:50:08] Using CUDA device 0 - NVIDIA GeForce GTX 1660 Ti [2024-09-13-09:50:08] Using CUDA device 0 - NVIDIA GeForce GTX 1660 Ti Traceback (most recent call last): File "C:\Users\antibody\EpiScan-master\EpiScan\commands\epimapping.py", line 124, in main(parser.parse_args()) File "C:\Users\antibody\EpiScan-master\EpiScan\commands\epimapping.py", line 34, in main train_model(args, output) File "C:\Users\antibody\EpiScan-master\EpiScan\commands\epimapping.py", line 77, in train_model modelCon = torch.load(modelCon_path).cuda() AttributeError: 'collections.OrderedDict' object has no attribute 'cuda'

gzBiomedical commented 2 months ago

Hi, I saved the model parameters on the CPU. If you need to perform inference on CUDA, you can first instantiate the model and then load the parameters:

from EpiScan.models.deep_ppi import DeepPPI modelCon_path = './Seq_final.pth' modelCon = DeepPPI(50, 5) modelCon.load_state_dict(torch.load(modelCon_path)) if use_cuda: modelCon.cuda()

cindyyeh commented 2 months ago

After instantiating the model, I ran into two additional issues: TypeError: unhashable type: 'Series' when defining p0Con. I found a workaround but the fixes were very extensive so I'm wondering if you have a more straightforward fix @gzBiomedical

Additionally, I ran into a class error when trying to run the map_predict() function on DeepPPI. Can you suggest an approach for calling this function?

gzBiomedical commented 2 months ago

After instantiating the model, I ran into two additional issues: TypeError: unhashable type: 'Series' when defining p0Con. I found a workaround but the fixes were very extensive so I'm wondering if you have a more straightforward fix @gzBiomedical

Additionally, I ran into a class error when trying to run the map_predict() function on DeepPPI. Can you suggest an approach for calling this function?

I wrote a demo in the commands directory: [demo_test.py], it should be worked now

cindyyeh commented 2 months ago

I tested out the demo with the updated repo and I'm still running into an error with the map_predict() function. setStorage: sizes [1, 46, 27], strides [46, 1, 0], storage offset 0, and itemsize 4 requiring a storage size of 184 are out of bounds for storage of size 0 Any suggestions?

Here's the full error:

---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
/tmp/ipykernel_179412/3899989061.py in <module>
----> 1 cmCon,_ = modelCon.map_predict(p0Con, p1Con,test_dfCon[3][indedx],index_cdrlist)

~/epitope_predictions/EpiScan_2/EpiScan/EpiScan/models/interaction_sep.py in map_predict(self, z0, z1, catsite, cdrindex)
    110     def map_predict(self, z0, z1,catsite,cdrindex):
    111 
--> 112         C = self.cpred(z0, z1,catsite,cdrindex)
    113 
    114         if self.do_w:

~/epitope_predictions/EpiScan_2/EpiScan/EpiScan/models/interaction_sep.py in cpred(self, z0, z1, catsite, cdrindex)
    104 
    105 
--> 106         B = self.contact.cmap(e0, e1,catsite,cdrindex)
    107         C = self.contact.predict(B)
    108         return C

~/epitope_predictions/EpiScan_2/EpiScan/EpiScan/models/contact_sep.py in cmap(self, z0, z1, catsite, cdrindex)
    165     def cmap(self, z0, z1, catsite,cdrindex):
    166 
--> 167         C = self.hidden(z0, z1, catsite,cdrindex)
    168         return C
    169 

~/anaconda3/envs/episcan/lib/python3.7/site-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
   1108         if not (self._backward_hooks or self._forward_hooks or self._forward_pre_hooks or _global_backward_hooks
   1109                 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1110             return forward_call(*input, **kwargs)
   1111         # Do not call functions when jit is used
   1112         full_backward_hooks, non_full_backward_hooks = [], []

~/epitope_predictions/EpiScan_2/EpiScan/EpiScan/models/contact_sep.py in forward(self, z0, z1, catsite, cdrindex)
     94             Z1Hcdr = torch.gather(z1H, 2, cdrHind)
     95             Z1Hnotcdr = torch.gather(z1H, 2, notcdrHind)
---> 96             Z1Lcdr = torch.gather(z1L, 2, cdrLind)
     97             Z1Lnotcdr = torch.gather(z1L, 2, notcdrLind)
     98 

RuntimeError: setStorage: sizes [1, 46, 27], strides [46, 1, 0], storage offset 0, and itemsize 4 requiring a storage size of 184 are out of bounds for storage of size 0
gzBiomedical commented 2 months ago

I tested out the demo with the updated repo and I'm still running into an error with the map_predict() function. setStorage: sizes [1, 46, 27], strides [46, 1, 0], storage offset 0, and itemsize 4 requiring a storage size of 184 are out of bounds for storage of size 0 Any suggestions?

Here's the full error:

---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
/tmp/ipykernel_179412/3899989061.py in <module>
----> 1 cmCon,_ = modelCon.map_predict(p0Con, p1Con,test_dfCon[3][indedx],index_cdrlist)

~/epitope_predictions/EpiScan_2/EpiScan/EpiScan/models/interaction_sep.py in map_predict(self, z0, z1, catsite, cdrindex)
    110     def map_predict(self, z0, z1,catsite,cdrindex):
    111 
--> 112         C = self.cpred(z0, z1,catsite,cdrindex)
    113 
    114         if self.do_w:

~/epitope_predictions/EpiScan_2/EpiScan/EpiScan/models/interaction_sep.py in cpred(self, z0, z1, catsite, cdrindex)
    104 
    105 
--> 106         B = self.contact.cmap(e0, e1,catsite,cdrindex)
    107         C = self.contact.predict(B)
    108         return C

~/epitope_predictions/EpiScan_2/EpiScan/EpiScan/models/contact_sep.py in cmap(self, z0, z1, catsite, cdrindex)
    165     def cmap(self, z0, z1, catsite,cdrindex):
    166 
--> 167         C = self.hidden(z0, z1, catsite,cdrindex)
    168         return C
    169 

~/anaconda3/envs/episcan/lib/python3.7/site-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
   1108         if not (self._backward_hooks or self._forward_hooks or self._forward_pre_hooks or _global_backward_hooks
   1109                 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1110             return forward_call(*input, **kwargs)
   1111         # Do not call functions when jit is used
   1112         full_backward_hooks, non_full_backward_hooks = [], []

~/epitope_predictions/EpiScan_2/EpiScan/EpiScan/models/contact_sep.py in forward(self, z0, z1, catsite, cdrindex)
     94             Z1Hcdr = torch.gather(z1H, 2, cdrHind)
     95             Z1Hnotcdr = torch.gather(z1H, 2, notcdrHind)
---> 96             Z1Lcdr = torch.gather(z1L, 2, cdrLind)
     97             Z1Lnotcdr = torch.gather(z1L, 2, notcdrLind)
     98 

RuntimeError: setStorage: sizes [1, 46, 27], strides [46, 1, 0], storage offset 0, and itemsize 4 requiring a storage size of 184 are out of bounds for storage of size 0

This is due to a mismatch in sequence length. Re-generating DB1.h5 with DB1.fasta(updated) should fix the problem