Julian Büchel

I don't understand two points in your code: `if l == L: #Convert z and apply softmax for the last layer. (TODO: Only for prediction or if we pass through...

Possible memory leak in uploading to blob (CA-13)

12

Hi, I am using the stable version (for IDF 3) of this repository. I found that I am constantly loosing between 30 and 40 bytes, each time the callback is...

torch.round() and tl.math.round() don't produce same results

4

To break the tie at numbers that end with .5, PyTorch rounds to the next even number. Triton seems to round up, i.e. 2.5 -> 3.0 instead of 2.0 Expected...

Hi. This is not really an issue. did somebody train the example all the way and can provide a checkpoint?

Fixed by checking if tile needs to be re-created

3

## Related issues Issue #609 ## Description Tile is re-created and reference between optimizer and tile is cut. ## Details For the torch tile, we don't re-create the tile unless...

Fix the support of different dtypes for the torch model

1

## Related issues Issue #623 ## Description Changed the .to function mostly. ## Details - Verification on GPU needed - A tests needs to be added that verifies the output...

Continuing training based on checkpoint using torch tile

7

## Description Saving and loading the model and optimizer state before resuming training changes the behavior. ## How to reproduce MWE: ``` import torch from aihwkit.optim.analog_optimizer import AnalogSGD from aihwkit.simulator.configs...

Hi when I am creating a new user, save and then try to login, I cant't. Do I have to chnage the base url or smth.? I am using VS17....

Hi, I see many criterions for accuracy, for example the accuracy must be within 99% of the FP-32 accuracy. Is there a similar idea for perplexity scores in language modelling?

[Q] How to turn off only model synching in huggingface integration

7

Hi. I am training large models and they are logged to wandb. This happens through the artifacts. How do I only turn off this feature? I tried googling but couldn’t...