Goal: instantiate unet_learner() using weights.
weights is a str that I bring in from a user-defined .yaml file; hence eval().
file_path and training are classes that hold parameters.
Code:
import numpy as np from fastai.vision.all import * def train(dls, file_path, training): labels = np.loadtxt(file_path.labels, dtype=str) weights = torch.tensor(eval(training.weights)) print('#################') print(weights) print(type(weights)) print('#################') learner = unet_learner(dls, training.architecture,loss_func=CrossEntropyLossFlat( axis=1, weight=weights) ) return learner.load(file_path.weights) Placing torch.tensor() around weights again in the parameter line doesn't help. Same error.
Traceback:
(venv) me@ubuntu-pcs:~/PycharmProjects/project$ python pdl1_lung_train/main.py /home/me/miniconda3/envs/venv/lib/python3.7/site-packages/torch/cuda/__init__.py:52: UserWarning: CUDA initialization: Found no NVIDIA driver on your system. Please check that you have an NVIDIA GPU and installed a driver from http://www.nvidia.com/Download/index.aspx (Triggered internally at /opt/conda/conda-bld/pytorch_1607370156314/work/c10/cuda/CUDAFunctions.cpp:100.) return torch._C._cuda_getDeviceCount() > 0 ################# tensor([0.4000, 0.9000]) <class 'torch.Tensor'> ################# Traceback (most recent call last): File "pdl1_lung_train/main.py", line 27, in <module> main(ROOT) File "pdl1_lung_train/main.py", line 19, in main learner = train(dls, file_path, training) File "/home/me/PycharmProjects/project/pdl1_lung_train/train.py", line 16, in train weight=weights)) File "/home/me/miniconda3/envs/venv/lib/python3.7/site-packages/fastai/vision/learner.py", line 267, in unet_learner model = create_unet_model(arch, n_out, img_size, pretrained=pretrained, **kwargs) File "/home/me/miniconda3/envs/venv/lib/python3.7/site-packages/fastai/vision/learner.py", line 243, in create_unet_model model = arch(pretrained) TypeError: 'str' object is not callable Please let me know if I need to add other info. to post.
