AlpinDale
AlpinDale
This PR updates the existing code to check GPU support for bfloat16 using `torch.cuda.is_bf16_supported()` function, which returns a boolean value. This should provide a more robust approach than using `torch.cuda.get_device_capability()`.
This pull request adds a model-agnostic script for merging PEFT models back into the base model.
Moving the issue from [here](https://github.com/asdf-vm/asdf/issues/1620#issue-1852810928). The current build instructions don't specify the system package requirements for building python with asdf on a fresh OS. Might be worth adding these, or...
This PR adds support for [Aphrodite Engine](https://github.com/PygmalionAI/aphrodite-engine). WIP as this is currently untested.
The current `urllib.request.urlopen()` implementation doesn't seem to support a resume functionality. This PR checks whether the host system has `aria2` installed, and will start the download using aria2. Otherwise, it'll...
It seems as if the model is being loaded in FP16. I've also noticed how `QuantizedWeight8bit` is imported in run.py, but not actually used. Is that for runtime quantization with...
`hf_transfer`, to my knowledge, has become very stable recently. I use it daily, and I find it a bit cumbersome that we have to manually install the package, then export...
This PR adds support for the [T5](https://huggingface.co/google/flan-t5-large) family of models, a series of encoder-decoder models. Currently a work in progress. TODO: - [x] Add the modeling code - [x] Add...
PR adds a new hyperparameter to typical_p sampling, which scales the maximum threshold for positive deviations in typ_p. Credits to Suikamelon (@BugReporterZ ). Untested yet.
This [WIP] PR adds support for the [grok-1](https://github.com/xai-org/grok-1) model.