breknddone
breknddone
Hello, @xingyuuchen , I added the two functions compute_sgt_loss and compute_affinity, but during training, the error is reported as follows: Traceback (most recent call last): File "train.py", line 74, in...
Here is the preprocessing code for **[RobotCar](https://github.com/w2kun/RNW/blob/651f8f425d115afc9fd4808383e95a0b2d928d76/datasets/robot_car.py)** and **[Nuscenes](https://github.com/w2kun/RNW/blob/651f8f425d115afc9fd4808383e95a0b2d928d76/datasets/nuscenes.py)** datasets.
> The KeyError is because you tried to load the key that did not exist. You can debug the input loading process step by step using e.g. pycharm, to check...
This situation is not caused by the damage of the 7z.002 file, but because the file is too large. **In this case, you only need to decompress the 7z.001 file.**...
> We used the pretrained ResNet models and the daytime pretraining is used to train patch-GAN. So can it be understood that the pre-trained weights during the day and night...
> Can you provide the results of adding `--sandbox_debug` and `--verbose_failures` to the build command in setup.py? Seems like from what is here there is a toolchain issue with g++....
> this seems relevant since you are using conda [pytorch/vision#8540](https://github.com/pytorch/vision/issues/8540) I follow this method, it seems worked. ``` conda install -c conda-forge gxx_linux-64 cd /home/ms/anaconda3/envs/NAME/bin ln -s /home/ms/anaconda3/envs/NAME/bin/aarch64-conda_cos7-linux-gnu-g++ g++ ```...
> Can you change the bazel build command in setup.py to add `--sandbox-debug` and `--verbose_failures` and share the output? This message does not convey that much information sure. ``` DEBUG:...
I use this to install bazel： ``` wget -v https://github.com/bazelbuild/bazelisk/releases/download/v1.20.0/bazelisk-linux-arm64 sudo mv bazelisk-linux-arm64 /usr/bin/bazel chmod +x /usr/bin/bazel ``` > Hmm, how did you install bazel and can you share bazel/bazelisk...
Because your magnification or reduction factor exceeds 4 or is less than 1,You can see the comment on line 26 to 27 of quick_test.py: > model is trained on scale...