Matt Watson

icon company Google icon location San Francisco

Results 342 comments of Matt Watson

Add LLaMa v2 to KerasNLP

This is done!

Mixed precision doesn't work properly on NVIDIA A10G GPUs

@reedwm could you take a look?

Add Swin-UNETR to Keras-Hub

Do we have some indication of how many downloads/users this model has today? I would think that it is lower priority than, say, qwen, which we are working on adding...

Run build if formatting succeeds

> One additional comment - to abandon previous commit tests when a new commit is pushed, we can do something like this: https://github.com/TeamAmaze/AmazeFileManager/blob/release/3.7/.github/workflows/android-feature.yml#L10-L12 Nice! I am down to try this...

Run build if formatting succeeds

Looks like this still needs some of the changes discussed above, but ping whenever this is ready for review!

The LLaMA implementation by Keras Hub exhibits significant deviations in accuracy compared to the standard implementation (Hugging Face).

Thanks! Will take a look.

Example usage code of models available in the repo.

@ramkumarkoppu thanks for the issue! Whisper is actually currently under development by @abheesht17 and has not made it in to a stable release quite yet. We still would like to...

Rendering issue on keras.io for argument lists

@ADITYADAS1999 the fix will likely need to be located in https://github.com/keras-team/keras-io. We should not change the actual docstrings, rather fix the rendering code if possible. You are welcome to take...

add vicuna preset

Thank you! Trying it out now.

add vicuna preset

Thanks! This looks good. I was having trouble loading this, but then remembered that since this is under the llama community license, you will actually have to be logged in...