M1 Pro: 'NoneType' has no attribute 'name'
mateuspestana opened this issue ยท 6 comments
Hi,
I'm trying to train my model on my MacBook Pro M1 Pro, and I'm getting this error either I use SAEHD or Quick96. Am I doing something wrong?
Metal device set to: Apple M1 Pro
systemMemory: 16.00 GB
maxCacheSize: 5.33 GB
Running trainer.
[new] No saved models found. Enter a name of a new model : bolsobean
bolsobean
Model first run.
Error: 'NoneType' object has no attribute 'name'
Traceback (most recent call last):
File "/Users/mateuspestana/Downloads/DeepFaceLab_MacOS/.dfl/DeepFaceLab/mainscripts/Trainer.py", line 46, in trainerThread
model = models.import_model(model_class_name)(
File "/Users/mateuspestana/Downloads/DeepFaceLab_MacOS/.dfl/DeepFaceLab/models/ModelBase.py", line 161, in __init__
self.device_config = nn.DeviceConfig.GPUIndexes( force_gpu_idxs or nn.ask_choose_device_idxs(suggest_best_multi_gpu=True)) \
File "/Users/mateuspestana/Downloads/DeepFaceLab_MacOS/.dfl/DeepFaceLab/core/leras/nn.py", line 215, in ask_choose_device_idxs
best_device_indexes = [device.index for device in devices.get_equal_devices(devices.get_best_device()) ]
File "/Users/mateuspestana/Downloads/DeepFaceLab_MacOS/.dfl/DeepFaceLab/core/leras/device.py", line 78, in get_equal_devices
device_name = device.name
AttributeError: 'NoneType' object has no attribute 'name'
Did you work this out? I'm having trouble installing tensorflow on my M1.
I expect to get an M1 within next couple weeks and will look into it then.
To install tensorflow, change on the requirements from tensorflow to tensorflow-macos and tensorflow-metal and it works, but still unable to train the model, because of GPU indexing.
To install tensorflow, change on the requirements from tensorflow to tensorflow-macos and tensorflow-metal and it works, but still unable to train the model, because of GPU indexing.
.dfl/DeepFaceLab/core/leras/device.py
line27 replace to all_devices = []
To install tensorflow, change on the requirements from tensorflow to tensorflow-macos and tensorflow-metal and it works, but still unable to train the model, because of GPU indexing.
i can train model now, but it looks like run with cpu
The issue is tagged as "stale issue" due to no activity in 30 days. If there is no activity for 15 more days, the issue will be closed.