For the NeuralForecast release post tomorrow, it l...
# communications
t
For the NeuralForecast release post tomorrow, it looks like this PR is pending for one of the links. https://github.com/Nixtla/neuralforecast/pull/960 @Marco @Cristian Challu should we update what we have in the text?
c
@Marco there seems to be an error in the tests, can someone check it please?
For the image of the post I think maybe the results table would be better. This one:
t
That image is smaller, so I think it would work better for the LinkedIn format.
1
m
The error is in models.hint.ipynb. Don't know why it happens, the PR doesn't modify that file. I'll try to fix it
Updated PR. I increased the test tolerance for HINT. The difference between the expected and real values was 0.0003, but the tolerance was set to 0 and 1e-7. Fingers crossed, the test should pass and PR can be merged! 🙂
We have another fail on tests due to the MPS backend:
Copy code
RuntimeError: MPS backend out of memory (MPS allocated: 0 bytes, other allocations: 0 bytes, max allowed: 7.93 GB). Tried to allocate 256 bytes on shared pool. Use PYTORCH_MPS_HIGH_WATERMARK_RATIO=0.0 to disable upper limit for memory allocations (may cause system failure).
I'm not sure it's a good idea to set an env variable in a notebook. Otherwise, I'm not sure how to fix this