MannLabs / alphapeptdeep

Deep learning framework for proteomics
Apache License 2.0
101 stars 19 forks source link

peptdeep bug on linux - destination is read-only #177

Open timheymann opened 3 weeks ago

timheymann commented 3 weeks ago

Describe the bug Run alphaDIA 1.7.0 from the GUI with transfer learning enabled on the GUI:

2:28:49.711954  [32;20mPROGRESS: Fine tuning finished at epoch: 51 
train_loss: nan 
learning_rate: nan 
test_loss: 0.00532 
test_r_square: 0.99819 
test_r: 0.9991 
test_slope: 1.001 
test_intercept: -0.0 
abs_error_95th_percentile: 0.01528 
 [0m
2:28:49.778870  ERROR: Output failed with error assignment destination is read-only 
Traceback (most recent call last):
File "/home/tim/git_repositories/alphadia/alphadia/planning.py", line 372, in run
output.build(workflow_folder_list, base_spec_lib)
File "/home/tim/git_repositories/alphadia/alphadia/outputtransform.py", line 383, in build
_ = self.build_transfer_model()
File "/home/tim/git_repositories/alphadia/alphadia/outputtransform.py", line 407, in build_transfer_model
stats = tune_mgr.finetune_charge(transfer_lib.precursor_df)
File "/home/tim/git_repositories/alphadia/alphadia/transferlearning/train.py", line 736, in finetune_charge
self._test_charge(-1, 0, test_df, test_metric_manager)
File "/home/tim/git_repositories/alphadia/alphadia/transferlearning/train.py", line 648, in _test_charge
pred = self.charge_model.predict(test_df)
File "/home/tim/miniconda3/envs/alpha/lib/python3.9/site-packages/peptdeep/model/model_interface.py", line 552, in predict
self._set_batch_predict_data(
File "/home/tim/miniconda3/envs/alpha/lib/python3.9/site-packages/peptdeep/model/generic_property_prediction.py", line 496, in _set_batch_predict_data
self.predict_df.loc[:,self.target_column_to_predict].values[
ValueError: assignment destination is read-only
2:29:13.289104  WARNING: WARNING: Temp mmap arrays were written to /tmp/temp_mmap_v9hdonwh. Cleanup of this folder is OS dependant, and might need to be triggered manually! Current space: 354,485,784,576 

To Reproduce Steps to reproduce the behavior: Run AlphaDIA 1.7.0 on Linux with transfer learning enabled

Version (please complete the following information):