fastai 0.7.0 not working Google Colaboratory anymore?
I am using Google Colaboratory and fastai 0.7.0
It worked fine some time back (I would guess in January this year),
now I am having a look at the notebooks, and they fail to run (without any code changes)
I have been using the dogs vs cats example from the course.
!pip install -q torchtext==0.2.3
!pip install -q fastai==0.7.0
torch.version
gives 0.3.1
arch=resnet34
data = ImageClassifierData.from_paths(PATH, tfms=tfms_from_model(arch, sz))
learn = ConvLearner.pretrained(arch, data, precompute=True)
learn.fit(0.01, 2)
module ‘torch.nn.init’ has no attribute ‘kaiming_normal_’
No idea why it worked some months back and what changed.
I try to figure out what the issue is, and it seems to be due a need to upgrade to pytorch:
I kill the Colab seesions, and reset try to upgrade
!pip install -q http://download.pytorch.org/whl/{accelerator}/torch-0.4.1-{platform}-linux_x86_64.whl torchvision
torch.version
gives 0.4.1
Running the code again
arch=resnet34
data = ImageClassifierData.from_paths(PATH, tfms=tfms_from_model(arch, sz))
learn = ConvLearner.pretrained(arch, data, precompute=True)
learn.fit(0.01, 2)
/usr/local/lib/python3.6/dist-packages/torch/cuda/init.py:114: UserWarning:
Found GPU0 Tesla T4 which requires CUDA_VERSION >= 9000 for
optimal performance and fast startup time, but your PyTorch was compiled
with CUDA_VERSION 8000. Please install the correct PyTorch binary
using instructions from http://pytorch.org
warnings.warn(incorrect_binary_warn % (d, name, 9000, CUDA_VERSION))
so I think it is runs on CPU now instead, and it never finishes (it runs out of memory on colab after a while)
If I try to upgrade to pytorch 1.1 I get
ERROR: fastai 0.7.0 has requirement torch<0.4, but you’ll have torch 1.1.0 which is incompatible.
Anyone successfully running 0.7.0 in Colab now?