I'm developing a project based on yolov7, but I started facing this error where torch recognizes my GPU but torchvision throws an Not Implemented Error.
This is the error
NotImplementedError: Could not run 'torchvision::nms' with arguments from the 'CUDA' backend. This could be because the operator doesn't exist for this backend, or was omitted during the selective/custom build process (if using custom build). If you are a Facebook employee using PyTorch on mobile, please visit https://fburl.com/ptmfixes for possible resolutions. 'torchvision::nms' is only available for these backends: [CPU, QuantizedCPU, BackendSelect, Python, FuncTorchDynamicLayerBackMode, Functionalize, Named, Conjugate, Negative, ZeroTensor, ADInplaceOrView, AutogradOther, AutogradCPU, AutogradCUDA, AutogradXLA, AutogradMPS, AutogradXPU, AutogradHPU, AutogradLazy, Tracer, AutocastCPU, AutocastCUDA, FuncTorchBatched, FuncTorchVmapMode, Batched, VmapMode, FuncTorchGradWrapper, PythonTLSSnapshot, FuncTorchDynamicLayerFrontMode, PythonDispatcher].
I tried installing torchvision with cuda built-in but that gave me the same error, also tried reinstalling pytorch , that didn't work either
the version of torch vision installed in my env was not equipped with cuda as it was a common install with pip with pip install torchvision whereas for torchvision to function with cuda it has to be equipped with cuda in-order for it to function with an Nvidia GPU to do so install torch with the following command conda install pytorch torchvision torchaudio pytorch-cuda={CUDA version} -c pytorch -c nvidia
I was able to run PyTorch with Macbook Pro M1 Max GPU. However Caffe2 does not use the GPUs.
import torch
torch.device("mps")
from caffe2.python import core
WARNING:root:This caffe2 python run failed to load cuda module:No module named 'caffe2.python.caffe2_pybind11_state_gpu',and AMD hip module:No module named 'caffe2.python.caffe2_pybind11_state_hip'.Will run in CPU only mode.
I created the PyTorch and Caffe2 from the nightly code using
export CMAKE_PREFIX_PATH=${CONDA_PREFIX:-"$(dirname $(which conda))/../"}
BUILD_CAFFE2=1 MACOSX_DEPLOYMENT_TARGET=10.9 CC=clang CXX=clang++ python setup.py install
Any suggestions on how to solve this?
Here is the Google Colab notebook dropped on github
As explained in the notebook, my purpose is to to experiment Deeplab v2 with Caffe (not with Tensorflow -With Tensorflow lib, I've managed to run an implementation of Deeplab in Google Colab).
Python does not fetch the link to the lib. As you'll notice in notebook, instruction import caffe throws an error when executed.
I have also tried straight forward install of the package using !apt install caffe-cuda in a different notebook but this does not work and everywhere I read like here, it suggests to build Caffe from source.
OS: Ubuntu 18.04.5 LTS
Python: Python 3.7.12
Cuda: Cuda compilation tools, release 11.1, V11.1.105
openCV: 4.1.2
I have a project I'm trying to work on but it's based on code that's a few years old and for whatever reason this code tends to fail if Tensorflow or NumPy aren't the correct versions (which means everything I'm using has to be old). This has meant that I've needed to dual-install an older version of Python to then be able to install the correct versions of the dependencies.
I'm running:
Python 3.7.5
NumPy 1.17.4
Pandas 0.25.3
pyyaml 5.1.2
more_itertools 7.2.0
keras 2.3.1
tensorflow 2.0.1
CUDA 10.0
CuDNN 7.4.1
I'm particularly interested in the keras and tensorflow versions. From my research, it seems they should work with GPU (as is?) according to this:
https://www.tensorflow.org/install/source (towards the bottom under tested build configurations for GPU).
However, when I try to detect GPU devices on my build with
from tensorflow.python.client import device_lib
print(device_lib.list_local_devices())
I get
2022-03-22 19:34:53.410102: I tensorflow/core/platform/cpu_feature_guard.cc:142] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2
[name: "/device:CPU:0"
device_type: "CPU"
memory_limit: 268435456
locality {
}
incarnation: 1722158755601489749
]
Which seems like it isn't recognising my GPU.
Is there something I'm missing in the setup process that I need to enable GPU support? As far as I can tell my versions of Tensorflow and Keras are compatible with GPU processing and have compatible versions of CUDA and CuDNN installed.
The goal is to run python -m spacy train with FP16 mixed precision to enable the use of large transformers (roberta-large, albert-large, etc.) in limited VRAM (RTX 2080ti 11 GB).
The new Spacy3 project.yml approach to training directly uses Huggingface-transformers models loaded via Spacy-transformers v1.0. Huggingface models can be run with mixed precision just by adding the --fp16 flag (as described here).
The spacy config was generated using python -m spacy init config --lang en --pipeline ner --optimize efficiency --gpu -F default.cfg, and checked to be complete by python -m spacy init fill-config default.cfg config.cfg --diff. Yet no FP16 / mixed-precision is to be found.
To reproduce
Use the spaCy Project: Named Entity Recognition (WikiNER) with changed init-config in project.yml to use a GPU and a transformer (roberta-base by default):
commands:
-
name: init-config
help: "Generate a transformer English NER config"
script:
- "python -m spacy init config --lang en --pipeline ner --gpu -F --optimize efficiency -C configs/${vars.config}.cfg"
What was tested
Added --fp16 to python -m spacy project run
Added --fp16 to python -m spacy train
Added fp16 = true to default.cfg in various sections ([components.transformer], [components.transformer.model], [training], [initialize])
The logic was transformers are run in FP16 as such:
from transformers import TrainingArguments
TrainingArguments(..., fp16=True, ...)
SW stack specifics
- spacy 3.0.3
- spacy-transformers 1.0.1
- transformers 4.2.2
- torch 1.6.0+cu101