hydai

LFX Mentorship (Jun-Aug, 2024): Enabling LLM fine tuning in the WASI-NN ggml plugin

15

### Summary ## Motivation WasmEdge is a lightweight and cross-platform runtime for LLM applications. It allows developers to create LLM apps on a Mac or Windows dev machine, compile them...

### Summary When running a large MoE model, the large tensors should be split across into multiple GPUs. Especially when we have multiple different GPUs with various VRAM sizes, this...

### Summary When using the ggml plugin in a multiple-GPU machine, it's important to specify the main GPU. The default value is using GPU 0 now. ### Details * Support...

### Summary WasmEdge currently provides a CUDA 12.0 pre-built `ggml` plugin, which is nice. However, some users only have a legacy CUDA version `11.x`. Unfortunately, they cannot upgrade their CUDA...

feat: Enabling LLM fine tuning in the WASI-NN PyTorch plugin

9

### Summary ## Motivation WasmEdge is a lightweight and cross-platform runtime for AI / LLM applications. It allows developers to create LLM apps on a Mac or Windows dev machine,...

### Summary ## Motivation WasmEdge is a WebAssembly runtime that supports both interpreter and ahead-of-time modes. For the [proxy-wasm](https://github.com/proxy-wasm/proxy-wasm-cpp-host) support, WasmEdge only currently provides the interpreter mode. Like the other...

feat: Implementing Coredump for Post-Mortem Debugging in WebAssembly

9

### Summary This project, [wasm-coredump](https://github.com/WebAssembly/tool-conventions/blob/main/Coredump.md), aims to enhance the debugging capabilities of WebAssembly by implementing a coredump feature for post-mortem analysis. When WebAssembly encounters a runtime error, it will initiate...

LFX Mentorship (Mar-May, 2024): Integrate MLX as a new WASI-NN backend

5

### Summary LLM is a hot topic, there are more and more frameworks to make the execution of LLM faster. WasmEdge already integrated the [llama.cpp](https://github.com/ggerganov/llama.cpp) as one of the backend....

### Summary The extension architecture is provided from 0.10.0. WasmEdge implements two extensions for enabling the ML and Image processing solutions, including `wasmedge-tensorflow-extension` and `wasmedge-image-extension.` However, the original extension design...

LFX Mentorship (Mar-May, 2024): Integrate burn.rs as a new WASI-NN backend

6

### Summary WasmEdge supports PyTorch, TensorFlow Lite, llama.cpp, and more NN backend. [Burn.rs](https://github.com/tracel-ai/burn) is a new deep-learning framework built using Rust. The portability, flexibility, and compute efficiency are important to...