Lime-Cakes

Results 9 issues of Lime-Cakes

Is GANformer2 code currently included in this repro? I can't seem to find anything on the two stage method anywhere in the code. If not included at the moment, is...

Allows saving mapping_dict when loading pytorch model through from_pretrained(from_pt=True), which can be used to save flax model in pytorch bin format. Add new option `return_mapping_dict` to from_pretrained. When set to...

I got the following out of the blue while testing some code on free pod4 machines. After restarting machine, the same loading code seemed to run fine, being stable and...

Is there an explanation of this error? ``` /usr/local/lib/python3.8/dist-packages/poptorch/experimental.py in __exit__(self, exc_type, value, traceback) 253 if self._compile_using == enums.Compiler.PopART: 254 # Compile the captured graph using PopART. --> 255 self._executable...

Is it possible to use dataloader with a custom sample/batch_sampler? At the moment, I cannot find any useful information on using poptorch's dataloader with custom sampler. Are there plans to...

### Description When updating to new Jax version (0.4.1/0.4.2) on Kaggle TPU VM, jax.local_devices() take a few minutes to run. Full error as below: ``` WARNING: Logging before InitGoogle() is...

bug

I have a question about the Discussion section of the paper. It mentions two issue, the issue of varying image size and the issue of training instabilities. I'd like to...

### Description Pallas doc claims automatic differentiation of pallas kernel work but is slower. However, when tested on tpu, AssertionError is raised. Tested with simple add kernel. ( Reference: https://jax.readthedocs.io/en/latest/pallas/design.html#grad-of-pallas-call...

bug
pallas

The main use for pallas is for tpu, but right now, a lot of options related to TPU compiler is left undocumented and require guesswork. Documentation for `PrefetchScalarGridSpec` is incomplete,...

enhancement
pallas