trouble with batchsize

[algo-scope]

When I want to use a bigger batchsize of 8, I got an error at the last iteration of Epoch 0.

Traceback (most recent call last): File "/data/StereoSpike/train.py", line 201, in warmup_chunks_left = warmup_chunks_left.view(batchsize, N_warmup * nfpdm, 2, 260, 346) RuntimeError: shape '[8, 1, 2, 260, 346]' is invalid for input of size 1079520

So could you explain the meaning of the general parameters? Thanks a lot.

[urancon]

Hello, thank you for pointing out this bug ! It is because the drop_last argument is set to False in the train dataloader. Set it to True and that should fix the problem ! I've corrected this in a new commit.

https://github.com/urancon/StereoSpike/blob/b59f65121d618e1ed3039e41d37318917762ba22/train.py#L99

Concerning the meaning of the general parameters, I've also put some comments to help in train.py. Which one did you want to know more about ? Basically, the dataset is a continuous stream of events, with depth labels every 50 ms. I call a "chunk" the duration between 2 subsequent labels (i.e., 50 ms). Chunks are divided into nfpdm frames of smaller size (say, if nfpdm=2, 25 ms). nfpdm stands for "number of frames per depth map". You can choose with N_inference the number of training/testing chunks you want to grab at each dataloader iteration. Similarly, you can grab N_warmup warmup chunks. Warmup chunks are the chunks preceding the train/test chunks.

Did I answer well your questions ? Don't hesitate to ask if you need !

[algo-scope]

Thanks for your explaination! I've got a new problem. When I ran test.py, tqdm(test_data_loader) raise an IndexError:

Traceback (most recent call last): File "/data/StereoSpike/test.py", line 109, in test_data_loader): File "/home/dell/anaconda3/envs/stereoSNN/lib/python3.6/site-packages/tqdm/std.py", line 1180, in iter for obj in iterable: File "/home/dell/anaconda3/envs/stereoSNN/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 521, in next data = self._next_data() File "/home/dell/anaconda3/envs/stereoSNN/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 561, in _next_data data = self._dataset_fetcher.fetch(index) # may raise StopIteration File "/home/dell/anaconda3/envs/stereoSNN/lib/python3.6/site-packages/torch/utils/data/_utils/fetch.py", line 49, in fetch data = [self.dataset[idx] for idx in possibly_batched_index] File "/home/dell/anaconda3/envs/stereoSNN/lib/python3.6/site-packages/torch/utils/data/_utils/fetch.py", line 49, in data = [self.dataset[idx] for idx in possibly_batched_index] File "/home/dell/anaconda3/envs/stereoSNN/lib/python3.6/site-packages/torch/utils/data/dataset.py", line 363, in getitem return self.dataset[self.indices[idx]] File "/data/StereoSpike/datasets/MVSEC/mvsec_dataset.py", line 221, in getitem groundtruth = self.labels[index] # 13 IndexError: index 1060 is out of bounds for axis 0 with size 1060

I think it's because that the last index of SPLIT1_TEST_INDICES is 1060, which is out of bounds

[DongZeLiu]

Thanks for your explaination! I've got a new problem. When I ran test.py, tqdm(test_data_loader) raise an IndexError:

Traceback (most recent call last): File "/data/StereoSpike/test.py", line 109, in test_data_loader): File "/home/dell/anaconda3/envs/stereoSNN/lib/python3.6/site-packages/tqdm/std.py", line 1180, in iter for obj in iterable: File "/home/dell/anaconda3/envs/stereoSNN/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 521, in next data = self._next_data() File "/home/dell/anaconda3/envs/stereoSNN/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 561, in _next_data data = self._dataset_fetcher.fetch(index) # may raise StopIteration File "/home/dell/anaconda3/envs/stereoSNN/lib/python3.6/site-packages/torch/utils/data/_utils/fetch.py", line 49, in fetch data = [self.dataset[idx] for idx in possibly_batched_index] File "/home/dell/anaconda3/envs/stereoSNN/lib/python3.6/site-packages/torch/utils/data/_utils/fetch.py", line 49, in data = [self.dataset[idx] for idx in possibly_batched_index] File "/home/dell/anaconda3/envs/stereoSNN/lib/python3.6/site-packages/torch/utils/data/dataset.py", line 363, in getitem return self.dataset[self.indices[idx]] File "/data/StereoSpike/datasets/MVSEC/mvsec_dataset.py", line 221, in getitem groundtruth = self.labels[index] # 13 IndexError: index 1060 is out of bounds for axis 0 with size 1060

I think it's because that the last index of SPLIT1_TEST_INDICES is 1060, which is out of bounds

Hi, I have met the same problem, have you fix it ?