Sebastian Schmidt comments

Results 217 comments of


                                            Sebastian Schmidt

Failed to get LLM params: INVALID_ARGUMENT: LLM model file is null

Which file are your downloading? https://www.kaggle.com/models/google/gemma/tfLite/gemma-2b-it-gpu-int4 should work on Android (among others).

Linker error when trying to build the C++ examples

This is also breaking MediaPipe's Python build for Windows. I am trying to track it down. The offending commit is https://github.com/tensorflow/tensorflow/commit/d6e25b069d9ddf4111d7def74c145d118c6f4f81, though there doesn't seem to be an obvious problem...

Merge LoRA weights to LLM at initialization time on-device (Gemma)

Thank you. We are logging this as a feature request.

Add support for Python 3.12

@ayushgdev started working on this.

Add support for Python 3.12

We have added 3.12 wheels.

Allow cancelling ongoing async generation in LlmInference

Logged this as an internal feature request: b/335859045

Executing ImageGenerator.createFromOptions crashes the program

You are getting an error from OpenCL here. I will route this issue to our GPU team.

How to set a system prompt for RAG implementation for Inference for Gemma 2b on IOS ?

We are working on making it easier to build more advanced use case on top of our LLM Inference. That being said, you can already tell the model to act...

Unable to build with Webpack after importing Mediapipe (Module parse failed: Unexpected token)

Are you able to force Webpack to use the .cjs files (instead of .mjs)?

ImageEmbedderOption quantize behavior

In our current pipeline, your model output has to match the output format of the models that our tasks are designed to handle. It seems pretty likely that we are...