Shubham Panchal

icon indicating a website link https://shubham0204.github.io

icon location Pune, Maharashtra, India icon location Building On-Device ML for Android, ML/CS enthusiast

Results 54 comments of Shubham Panchal

Nail tracker

@manhlt1990 There's no such pretrained model for nail detection. If in your use-case, a constant hand orientation is observed, you can locate the relative position of each nail ( for...

Nothing output

Please confirm if the issue is solved in the [latest commit](https://github.com/shubham0204/Hand_Detection_TFLite_Android/commit/8752a41e2706bb1719f51cdd8227475efbd99483).

Scoring accuracy is not at par with original CLIP model

@ua741 Are you using the `int8` quantized model from [clip.cpp](https://github.com/monatis/clip.cpp)? This repo is using `int8` quantized ONNX models.

Scoring accuracy is not at par with original CLIP model

@ua741 I have created an Android app that uses JNI bindings built over [clip.cpp](https://github.com/monatis/clip.cpp). Check the project here: https://github.com/shubham0204/CLIP-Android Also, here's the PR I made to clip.cpp for adding the...

Gemma-1.1-2B-it-cpu-int4.bin generating junk response in phone.

After switching to the `gemma-1.1-2b-it-gpu-int4` model to get a cleaner reply, I got the following error on initializing the inference session: `com.google.mediapipe.framework.MediaPipeException: internal: Failed to initialize session: %sCan not open...

library "libllm_inference_engine_jni.so" not found - LLM inference for Android Example.

@sachinsshetty The issue seems to arise because of the incompatibility of Mediapipe LLM Inference API for ARM-32 (armeabi-v7a) devices. See [this issue](https://github.com/google-ai-edge/mediapipe-samples/issues/423#issue-2433531578) I created a few days ago.

Adding a DOCX file to be extracted into chunks and added to the vector database, like the PDF's fails

@dcrowleyMunster I am not able to reproduce the issue with the steps you mentioned. It would be great if you can provide the crash logs from the 'Logcat' of Android...

Adding a DOCX file to be extracted into chunks and added to the vector database, like the PDF's fails

Closing this issue as additional details were not provided by the user.

Can we build a RAG Android application using Local LLM or SLM (HF tflite models)

@vimal-yuvabe that's a great idea! I have been working with `llama.cpp` for a while and will soon integrate it with this project.

Can we build a RAG Android application using Local LLM or SLM (HF tflite models)

I have an app which uses local GGUF LLMs to generate response: [SmolChat](https://github.com/shubham0204/SmolChat-Android). It has an issue open for adding RAG support: [SmolChat #3](https://github.com/shubham0204/SmolChat-Android/issues/3) We can track this feature in...