Outputting WandB performance metrics

[canihaveabravo]

Hello,

I'm having ongoing difficulty outputting wandb logs into useful performance metrics. Can you provide any tips for this? The wandb_session argument from the CNN training tutorial doesnt seem to run with my code.

Thanks very much,
Lorenzo

[sammlapp]

Hi Lorenzo, please provide some details of your code and error messages so that we can assist you.

An example of using wandb with CNN.train is given here: https://opensoundscape.org/en/latest/tutorials/train_cnn.html#Set-up-WandB-model-logging

Make sure you have created a wandb account and run wandb.login() with your API key, or wandb login in terminal, using the API key from https://wandb.ai/authorize

It will also be worth checking if wandb logging works when you use the tutorial notebook from that link

[canihaveabravo]

Hi Samm,

Yes I am able to log a run in WandB using the code below and I can see basic hyperparameter plots, but none of the interesting graphs of performance metrics, or samples described in the tutorial. Any ideas?

import` wandb
#Import WandB Utility
try:
wandb.login()
wandb_session = wandb.init(
entity='seiche',
project='multispecies_CNN',
name='multispecies_CNN',
)

except: #if wandb.init fails, don't use wandb logging
print('failed to create wandb session. wandb session will be None')
wandb_session = `None'

However when inserting the wandb_session argument into my training loop:

model.train(
balanced_train_df,
valid_df,
epochs= 10,
batch_size= 128,
log_interval = 5,
num_workers= 0, # if multiple processors are available utilise them to speed up training
wandb_session = wandb_session,
save_interval= 10,
save_path= checkpoint_folder #location to save checkpoints
)

I get the error:

c:\Users\lscala\AppData\Local\anaconda3\envs\opensoundscape\Lib\site-packages\opensoundscape\audio.py:340: UserWarning: Audio object is shorter than requested duration: 10.000020833333334 sec instead of 30 sec
warnings.warn(error_msg)
c:\Users\lscala\AppData\Local\anaconda3\envs\opensoundscape\Lib\site-packages\opensoundscape\audio.py:340: UserWarning: Audio object is shorter than requested duration: 12.0 sec instead of 30 sec
warnings.warn(error_msg)
c:\Users\lscala\AppData\Local\anaconda3\envs\opensoundscape\Lib\site-packages\opensoundscape\audio.py:340: UserWarning: Audio object is shorter than requested duration: 10.000020833333334 sec instead of 30 sec
warnings.warn(error_msg)
:1047: ImportWarning: ImportHookFinder.find_spec() not found; falling back to find_module()

FileNotFoundError Traceback (most recent call last)
c:\Users\lscala\Python_learning\multispecies_trial_chris.ipynb Cell 20 line 1
12 model.optimizer_params['weight_decay']=0.0005
14 # Train the model on the full training set and validate on the validation set. Save the model to directory.
---> 15 model.train(
16 balanced_train_df,
17 valid_df,
18 epochs= 2,
19 batch_size= 128,
20 log_interval = 5,
21 num_workers= 0, # if multiple processors are available utilise them to speed up training
22 wandb_session = wandb_session,
23 save_interval= 10,
24 save_path= checkpoint_folder #location to save checkpoints
25 )
27 #Let Wandb know that we finished training successfully
28 wandb.unwatch(model.network)

File c:\Users\lscala\AppData\Local\anaconda3\envs\opensoundscape\Lib\site-packages\opensoundscape\ml\cnn.py:841, in CNN.train(self, train_df, validation_df, epochs, batch_size, num_workers, save_path, save_interval, log_interval, validation_interval, invalid_samples_log, raise_errors, wandb_session, progress_bar)
833 wandb_session.watch(
834 self.network,
835 log="all",
836 log_freq=log_freq,
837 log_graph=(self.wandb_logging["log_graph"]),
...
-> 1617 with open(path, mode, encoding=encoding) as f:
1618 yield f
1620 f.flush()

FileNotFoundError: [Errno 2] No such file or directory: 'C:\Users\lscala\AppData\Local\Temp\tmplk94vkyk\Samples / training samples.table.json'

[sammlapp]

interesting, this seems like a wandb error. Your error message is not showing fully (the relevant part is within the ...) so it's hard to debug, but it seems like a caching error from wandb. It's possible that deleting the temporary folder C:\Users\lscala\AppData\Local\Temp\tmplk94vkyk could help.

[canihaveabravo]

Deleting the temp folder didnt help, but here is the error message in full. Thank you very much for your help.

---

FileNotFoundError Traceback (most recent call last)
c:\Users\lscala\Python_learning\multispecies_trial_chris.ipynb Cell 20 line 1
12 model.optimizer_params['weight_decay']=0.0005
14 # Train the model on the full training set and validate on the validation set. Save the model to directory.
---> 15 model.train(
16 balanced_train_df,
17 valid_df,
18 epochs= 2,
19 batch_size= 128,
20 log_interval = 5,
21 num_workers= 0, # if multiple processors are available utilise them to speed up training
22 wandb_session = wandb_session,
23 save_interval= 10,
24 save_path= checkpoint_folder #location to save checkpoints
25 )
27 #Let Wandb know that we finished training successfully
28 wandb.unwatch(model.network)

File c:\Users\lscala\AppData\Local\anaconda3\envs\opensoundscape\Lib\site-packages\opensoundscape\ml\cnn.py:841, in CNN.train(self, train_df, validation_df, epochs, batch_size, num_workers, save_path, save_interval, log_interval, validation_interval, invalid_samples_log, raise_errors, wandb_session, progress_bar)
833 wandb_session.watch(
834 self.network,
835 log="all",
836 log_freq=log_freq,
837 log_graph=(self.wandb_logging["log_graph"]),
838 )
840 # log tables of preprocessed samples
--> 841 wandb_session.log(
842 {
843 "Samples / training samples": wandb_table(
844 AudioFileDataset(
845 train_df, self.preprocessor, bypass_augmentations=False
846 ),
847 self.wandb_logging["n_preview_samples"],
848 ),
849 "Samples / training samples no aug": wandb_table(
850 AudioFileDataset(
851 train_df, self.preprocessor, bypass_augmentations=True
852 ),
853 self.wandb_logging["n_preview_samples"],
854 ),
855 "Samples / validation samples": wandb_table(
856 AudioFileDataset(
857 validation_df, self.preprocessor, bypass_augmentations=True
858 ),
859 self.wandb_logging["n_preview_samples"],
860 ),
861 }
862 )
864 # Move network to device
865 self.network.to(self.device)

File c:\Users\lscala\AppData\Local\anaconda3\envs\opensoundscape\Lib\site-packages\wandb\sdk\wandb_run.py:368, in _run_decorator._noop..wrapper(self, *args, **kwargs)
365 wandb.termwarn(message, repeat=False)
366 return cls.Dummy()
--> 368 return func(self, *args, **kwargs)

File c:\Users\lscala\AppData\Local\anaconda3\envs\opensoundscape\Lib\site-packages\wandb\sdk\wandb_run.py:331, in _run_decorator._attach..wrapper(self, *args, **kwargs)
329 raise e
330 cls._is_attaching = ""
--> 331 return func(self, *args, **kwargs)

File c:\Users\lscala\AppData\Local\anaconda3\envs\opensoundscape\Lib\site-packages\wandb\sdk\wandb_run.py:1712, in Run.log(self, data, step, commit, sync)
1705 if sync is not None:
1706 deprecate.deprecate(
1707 field_name=deprecate.Deprecated.run__log_sync,
1708 warning_message=(
1709 "sync argument is deprecated and does not affect the behaviour of wandb.log"
1710 ),
1711 )
-> 1712 self._log(data=data, step=step, commit=commit)

File c:\Users\lscala\AppData\Local\anaconda3\envs\opensoundscape\Lib\site-packages\wandb\sdk\wandb_run.py:1494, in Run._log(self, data, step, commit)
1491 if any(not isinstance(key, str) for key in data.keys()):
1492 raise ValueError("Key values passed to wandb.log must be strings.")
-> 1494 self._partial_history_callback(data, step, commit)
1496 if step is not None:
1497 if os.getpid() != self._init_pid or self._is_attached:

File c:\Users\lscala\AppData\Local\anaconda3\envs\opensoundscape\Lib\site-packages\wandb\sdk\wandb_run.py:1364, in Run._partial_history_callback(self, row, step, commit)
1361 if self._backend and self._backend.interface:
1362 not_using_tensorboard = len(wandb.patched["tensorboard"]) == 0
-> 1364 self._backend.interface.publish_partial_history(
1365 row,
1366 user_step=self._step,
1367 step=step,
1368 flush=commit,
1369 publish_step=not_using_tensorboard,
1370 )

File c:\Users\lscala\AppData\Local\anaconda3\envs\opensoundscape\Lib\site-packages\wandb\sdk\interface\interface.py:567, in InterfaceBase.publish_partial_history(self, data, user_step, step, flush, publish_step, run)
556 def publish_partial_history(
557 self,
558 data: dict,
(...)
563 run: Optional["Run"] = None,
564 ) -> None:
565 run = run or self._run
--> 567 data = history_dict_to_json(run, data, step=user_step, ignore_copy_err=True)
568 data.pop("_step", None)
570 # add timestamp to the history request, if not already present
571 # the timestamp might come from the tensorboard log logic

File c:\Users\lscala\AppData\Local\anaconda3\envs\opensoundscape\Lib\site-packages\wandb\sdk\data_types\utils.py:52, in history_dict_to_json(run, payload, step, ignore_copy_err)
48 payload[key] = history_dict_to_json(
49 run, val, step=step, ignore_copy_err=ignore_copy_err
50 )
51 else:
---> 52 payload[key] = val_to_json(
53 run, key, val, namespace=step, ignore_copy_err=ignore_copy_err
54 )
56 return payload

File c:\Users\lscala\AppData\Local\anaconda3\envs\opensoundscape\Lib\site-packages\wandb\sdk\data_types\utils.py:152, in val_to_json(run, key, val, namespace, ignore_copy_err)
148 sanitized_key = re.sub(r"[^a-zA-Z0-9_]+", "", key)
149 art = wandb.wandb_sdk.wandb_artifacts.Artifact(
150 f"run-{run.id}-{sanitized_key}", "run_table"
151 )
--> 152 art.add(val, key)
153 run.log_artifact(art)
155 # Partitioned tables and joined tables do not support being bound to runs.

File c:\Users\lscala\AppData\Local\anaconda3\envs\opensoundscape\Lib\site-packages\wandb\sdk\wandb_artifacts.py:557, in Artifact.add(self, obj, name)
555 do_write(tmp_f)
556 else:
--> 557 with self.new_file(name) as f:
558 file_path = f.name
559 do_write(f)

File c:\Users\lscala\AppData\Local\anaconda3\envs\opensoundscape\Lib\contextlib.py:137, in _GeneratorContextManager.enter(self)
135 del self.args, self.kwds, self.func
136 try:
--> 137 return next(self.gen)
138 except StopIteration:
139 raise RuntimeError("generator didn't yield") from None

File c:\Users\lscala\AppData\Local\anaconda3\envs\opensoundscape\Lib\site-packages\wandb\sdk\wandb_artifacts.py:390, in Artifact.new_file(self, name, mode, encoding)
388 filesystem.mkdir_exists_ok(os.path.dirname(path))
389 try:
--> 390 with util.fsync_open(path, mode, encoding) as f:
391 yield f
392 except UnicodeEncodeError as e:

File c:\Users\lscala\AppData\Local\anaconda3\envs\opensoundscape\Lib\contextlib.py:137, in _GeneratorContextManager.enter(self)
135 del self.args, self.kwds, self.func
136 try:
--> 137 return next(self.gen)
138 except StopIteration:
139 raise RuntimeError("generator didn't yield") from None

File c:\Users\lscala\AppData\Local\anaconda3\envs\opensoundscape\Lib\site-packages\wandb\util.py:1617, in fsync_open(path, mode, encoding)
1612 @contextlib.contextmanager
1613 def fsync_open(
1614 path: Union[pathlib.Path, str], mode: str = "w", encoding: Optional[str] = None
1615 ) -> Generator[IO[Any], None, None]:
1616 """Open a path for I/O and guarante that the file is flushed and synced."""
-> 1617 with open(path, mode, encoding=encoding) as f:
1618 yield f
1620 f.flush()

FileNotFoundError: [Errno 2] No such file or directory: 'C:\Users\lscala\AppData\Local\Temp\tmpqt2c4dg7\Samples / training samples.table.json'

[sammlapp]

It seems like other wandb users have experiences a similar issue: wandb/wandb#1869 (even since that issue was "closed")

Are you perhaps able to run wandb.finish() then start a new wandb session with wandb.init() ? Does the error occur again?

[canihaveabravo]

Yeah unfortunately it occurs every time I start a new wandb_session. Lorenzo Scala Senior Marine Scientist Seiche Ltd T: +44(0)1409 404050 E: ***@***.*** W: https://www.seiche.com Bradworthy Industrial Estate, Langdon Road, Bradworthy, Holsworthy, Devon, EX22 7SF, United Kingdom Seiche Ltd is registered in England & Wales. Registered number: 1468514. Registered Office: Bradworthy Industrial Estate, Langdon Road, Bradworthy, Holsworthy, Devon, EX22 7SF, United Kingdom The information contained in this e-mail transmission, and any documents, files or previous e-mail messages attached to it, is privileged and confidential, and solely intended for the use of the individual(s) to whom it is addressed. If you are not the intended recipient or a person responsible for delivering it to the intended recipient you should not read, copy, distribute or otherwise use the information, and you are hereby notified that any disclosure, copying, distribution or use of any of the information contained in or attached to this message is strictly prohibited. If you have received this message in error, please notify sender immediately and delete this e-mail and attached documents. From: sammlapp ***@***.***> Sent: Wednesday, November 1, 2023 9:27 PM To: kitzeslab/opensoundscape ***@***.***> Cc: Lorenzo Scala ***@***.***>; Author ***@***.***> Subject: Re: [kitzeslab/opensoundscape] Outputting WandB performance metrics (Discussion #903) CAUTION: This email originated from outside of SWTG. It seems like other wandb users have experiences a similar issue: wandb/wandb#1869<wandb/wandb#1869> (even since that issue was "closed") Are you perhaps able to run wandb.finish() then start a new wandb session with wandb.init() ? Does the error occur again? — Reply to this email directly, view it on GitHub<#903 (reply in thread)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AC4F3MGRJKVCLXLX6PCLFTTYCK5D7AVCNFSM6AAAAAA6TEFXH6VHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM3TINJQGA2DS>. You are receiving this because you authored the thread.Message ID: ***@***.******@***.***>>

[sammlapp]

can you make sure you have the latest version of wandb installed in your environment?

[canihaveabravo]

I have updated to wandb version 0.16.0 but the issue persists. I am scratching my head a bit with it 😊 Lorenzo Scala Senior Marine Scientist Seiche Ltd T: +44(0)1409 404050 E: ***@***.*** W: https://www.seiche.com Bradworthy Industrial Estate, Langdon Road, Bradworthy, Holsworthy, Devon, EX22 7SF, United Kingdom Seiche Ltd is registered in England & Wales. Registered number: 1468514. Registered Office: Bradworthy Industrial Estate, Langdon Road, Bradworthy, Holsworthy, Devon, EX22 7SF, United Kingdom The information contained in this e-mail transmission, and any documents, files or previous e-mail messages attached to it, is privileged and confidential, and solely intended for the use of the individual(s) to whom it is addressed. If you are not the intended recipient or a person responsible for delivering it to the intended recipient you should not read, copy, distribute or otherwise use the information, and you are hereby notified that any disclosure, copying, distribution or use of any of the information contained in or attached to this message is strictly prohibited. If you have received this message in error, please notify sender immediately and delete this e-mail and attached documents. From: sammlapp ***@***.***> Sent: Wednesday, November 8, 2023 4:04 AM To: kitzeslab/opensoundscape ***@***.***> Cc: Lorenzo Scala ***@***.***>; Author ***@***.***> Subject: Re: [kitzeslab/opensoundscape] Outputting WandB performance metrics (Discussion #903) CAUTION: This email originated from outside of SWTG. can you make sure you have the latest version of wandb installed in your environment? — Reply to this email directly, view it on GitHub<#903 (reply in thread)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AC4F3MEQFFHULDZ4O3XRH73YDMAE3AVCNFSM6AAAAAA6TEFXH6VHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM3TKMBVHA2DC>. You are receiving this because you authored the thread.Message ID: ***@***.******@***.***>>

[sammlapp]

Sorry, I also don't know what is causing the issue and it seems to be caused by wandb rather than opensoundscape. I suspect some sort of "cache clearing" like deleting temporary or cached files will resolve the issue

[sammlapp]

@canihaveabravo is it possible your disk is nearly full? See this thread: wandb/wandb#1869 (comment)

[sammlapp]

@canihaveabravo I encountered a similar issue recently. I resolved it by deleting all folders in the temp folder that had wandb in the name and killing all wandb processes (can check for them using htop or a similar tool).

[canihaveabravo]

Hi Sam, Thanks for the tip. I’ve tried this but unfortunately no luck so far. Lorenzo Scala Senior Marine Scientist Seiche Ltd T: +44(0)1409 404050 E: ***@***.*** W: https://www.seiche.com Bradworthy Industrial Estate, Langdon Road, Bradworthy, Holsworthy, Devon, EX22 7SF, United Kingdom Seiche Ltd is registered in England & Wales. Registered number: 1468514. Registered Office: Bradworthy Industrial Estate, Langdon Road, Bradworthy, Holsworthy, Devon, EX22 7SF, United Kingdom The information contained in this e-mail transmission, and any documents, files or previous e-mail messages attached to it, is privileged and confidential, and solely intended for the use of the individual(s) to whom it is addressed. If you are not the intended recipient or a person responsible for delivering it to the intended recipient you should not read, copy, distribute or otherwise use the information, and you are hereby notified that any disclosure, copying, distribution or use of any of the information contained in or attached to this message is strictly prohibited. If you have received this message in error, please notify sender immediately and delete this e-mail and attached documents. From: sammlapp ***@***.***> Sent: Wednesday, November 15, 2023 2:45 PM To: kitzeslab/opensoundscape ***@***.***> Cc: Lorenzo Scala ***@***.***>; Mention ***@***.***> Subject: Re: [kitzeslab/opensoundscape] Outputting WandB performance metrics (Discussion #903) CAUTION: This email originated from outside of SWTG. @canihaveabravo<https://github.com/canihaveabravo> I encountered a similar issue recently. I resolved it by deleting all folders in the temp folder that had wandb in the name and killing all wandb processes (can check for them using htop or a similar tool). — Reply to this email directly, view it on GitHub<#903 (reply in thread)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AC4F3MDH4J5SCVUJVFB3KCTYETIN5AVCNFSM6AAAAAA6TEFXH6VHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM3TKNZXHA3TC>. You are receiving this because you were mentioned.Message ID: ***@***.******@***.***>>

[sammlapp]

If you still need help resolving this to move forward, send me an email and we can try to tackle it together

[canihaveabravo]

Hi Sam. Yes I have no yet found a solution to this and have therefore taken another direction. However, happy to return to it and tackle it together.

[fascimare]

If you found a solution, please let me know. I am having the same issue.

[sammlapp]

Hi @fascimare sorry for not getting back to you. Can you also share your full error text to assist with debugging?

[sammlapp]

Also, it would be helpful if you can update wandb to the newest version and check if the error still occurs.

conda activate environment_name
pip install --upgrade wandb

[sammlapp]

@canihaveabravo and @fascimare , wandb seems to have several issues that could be causing this, but its worth checking if updating your wandb package to the most recent version fixes it. I still haven't been able to reproduce the behavior myself, unfortunately, so its difficult to debug. If you do try with the newest wandb version, I would love to hear if it works or not (please also post the version you install)

[m1alksne]

Hi All,

I have previously trained a model with opensoundscape and logged the metrics in real time with wandb. However, when I did that, I was using a Linux machine. I am currently trying to reproduce this work using a Windows machine. Originally when I tried this, I got an error messge similar to @canihaveabravo. When I updated my wandb package and made sure to activate my conda environment, I no longer got the error message.

I simply typed:

conda activate environment_name
pip install --upgrade wandb
print(wandb.version)
0.16.2

Unfortunately, although my code is no longer throwing an error message caused by wandb not finding the correct temp file, I am still unable to visualize the metrics using the wandb interface. The run does not crash but no metrics get logged or plotted. So, I didn't solve the problem but maybe got a little bit closer. I hope this helps.

[sammlapp]

it would also be helpful to know what operating system (and version) everyone is on @m1alksne @canihaveabravo @fascimare

our current hypothesis is that this is a Windows-specific problem

[canihaveabravo]

my_cnn.wandb_logging['log_graph']=False

this appears to help a bit. I dont get the same error message anymore, however I do get the below:

AttributeError Traceback (most recent call last)
Cell In[23], line 15
2 model.train(
3 balanced_train_df,
4 valid_df,
(...)
11 save_path= checkpoint_folder #location to save checkpoints
12 )
14 #Let Wandb know that we finished training successfully
---> 15 wandb.unwatch(model.network)
16 wandb.finish()

File c:\Users\lscala\AppData\Local\anaconda3\envs\opensoundscape\Lib\site-packages\wandb\sdk\wandb_watch.py:123, in unwatch(models)
121 else:
122 for name in model._wandb_hook_names:
--> 123 wandb.run._torch.unhook(name)
124 delattr(model, "_wandb_hook_names")
125 # TODO: we should also remove recursively model._wandb_watch_called
126
127 else:

AttributeError: 'NoneType' object has no attribute '_torch'

[fascimare]

Apologies for the late reply. I'm on a Windows 11 computer. I upgraded wandb to version 0.16.3 and turned off the logging of gradients, but still get the following error message:
Traceback (most recent call last):

Cell In[14], line 1
model.train(

File ~\AppData\Local\anaconda3\Lib\site-packages\opensoundscape\ml\cnn.py:841 in train
wandb_session.log(

File ~\AppData\Local\anaconda3\Lib\site-packages\wandb\sdk\wandb_run.py:420 in wrapper
return func(self, *args, **kwargs)

File ~\AppData\Local\anaconda3\Lib\site-packages\wandb\sdk\wandb_run.py:371 in wrapper_fn
return func(self, *args, **kwargs)

File ~\AppData\Local\anaconda3\Lib\site-packages\wandb\sdk\wandb_run.py:361 in wrapper
return func(self, *args, **kwargs)

File ~\AppData\Local\anaconda3\Lib\site-packages\wandb\sdk\wandb_run.py:1828 in log
self._log(data=data, step=step, commit=commit)

File ~\AppData\Local\anaconda3\Lib\site-packages\wandb\sdk\wandb_run.py:1596 in _log
self._partial_history_callback(data, step, commit)

File ~\AppData\Local\anaconda3\Lib\site-packages\wandb\sdk\wandb_run.py:1468 in _partial_history_callback
self._backend.interface.publish_partial_history(

File ~\AppData\Local\anaconda3\Lib\site-packages\wandb\sdk\interface\interface.py:558 in publish_partial_history
data = history_dict_to_json(run, data, step=user_step, ignore_copy_err=True)

File ~\AppData\Local\anaconda3\Lib\site-packages\wandb\sdk\data_types\utils.py:52 in history_dict_to_json
payload[key] = val_to_json(

File ~\AppData\Local\anaconda3\Lib\site-packages\wandb\sdk\data_types\utils.py:156 in val_to_json
art.add(val, key)

File ~\AppData\Local\anaconda3\Lib\site-packages\wandb\sdk\artifacts\artifact.py:1375 in add
with self.new_file(name) as f:

File ~\AppData\Local\anaconda3\Lib\contextlib.py:137 in enter
return next(self.gen)

File ~\AppData\Local\anaconda3\Lib\site-packages\wandb\sdk\artifacts\artifact.py:1122 in new_file
with util.fsync_open(path, mode, encoding) as f:

File ~\AppData\Local\anaconda3\Lib\contextlib.py:137 in enter
return next(self.gen)

File ~\AppData\Local\anaconda3\Lib\site-packages\wandb\util.py:1513 in fsync_open
with open(path, mode, encoding=encoding) as f:

FileNotFoundError: [Errno 2] No such file or directory: 'C:\Users\P310512\AppData\Local\Temp\tmpyqt9fy2m\Samples / training samples.table.json'

[fascimare]

I also tried the debug of the first blogpost you referred to, but no luck.

[sammlapp]

Ok thanks for checking. I'm curious, do you have a folder "C:\Users\P310512\AppData\Local\Temp\tmpyqt9fy2m" and if so what files are in it?

[fascimare]

The folder is there, and in it a folder 'Samples', but no other files and the Samples folder is empty.

[sammlapp]

Following @louisfh 's suggestion, I created a branch of opensoundscape where some special characters are removed from wandb logging calls (especially the "/" character). The branch is called patch_wandb_windows. If any of you @canihaveabravo @fascimare @m1alksne have a chance to try it out, I'm curious if it will resolve the wandb logging issue that we haven't been able to reproduce on our end.

To try it, you'll need to clone the opensoundscape github, checkout the branch, and pip install it. Here is an example of command line operations:

cd
git clone git@github.com:kitzeslab/opensoundscape.git
cd opensoundscape
git checkout patch_wandb_windows
git pull

now create a new conda environment (assuming you use conda for environment management)

conda deactivate
conda create -n opso_env_with_patch python=3.10

you might have to enter 'y' here to confirm you want to create the new conda environment
conda activate opso_env_with_patch.

You can install the package of the directory you're currently in with pip install . after activating the new environment

conda activate opso_env_with_patch
pip install .

Then check if you get the same errors when using this environment.

Fingers crossed that this helps!

[fascimare]

I tried it out today, and it's working! The only thing I'm not getting are the preview samples, but the rest is there. Thank you so much @sammlapp !

[m1alksne]

Hey @fascimare what type of Windows machine are you running on? I have not been able to get the wandb plots to work. Although I followed these steps and the code no longer throws an error.