Summary

Motivation

WasmEdge is a lightweight inference runtime for AI and LLM applications. Build specialized and finetuned models for WasmEdge community. The model should be supported by WasmEdge and its applications should benefit the WasmEdge community.

In this project, we will build and compare two finetuned model for Rust coding assistance.

Model 1

Objective: Rust Q/A model for PR review bot

• Current Dataset: Dataset1
• Current Model:Model1

Model 2

• Current Dataset: Dataset2
• Current Model:Model2

Collab : Currently used for fine tuning of both models

Milestones

• M1 [1 week]
• Discuss This timeline with mentors and get suggestions on my tasks and my current methods of fine tuning
• M2 [2 Weeks]
• Get deeper understanding of gaianet
• Improving quality of dataset
• Researching on prompt templates and testing their performance
• Researching on training parameters
• M3 [1 Week]
• Comparison of currently tuned model with Codestral model
• Testing RAG implementation with current model and codestral using gaianet nodes
• M4 [2 Weeks]
• Working on API inference of Model
• Get deeper understanding of Github PR review bot
• Create test cases for PR reviews
• Testing various versions of base model for fine tuning of both models
• M5 [1.5 Week]
• Integrating the model 1 with PR review bot
• Hosting model 2
• Get real life feedback for further improvement in their functionality
• M6 [1.5 Weeks]
• Working on final optimisation and integration
• M7 [2 Weeks]
• Bug solving
• Working on documentation

Appendix

Reference links
#3371
https://docs.gaianet.ai/intro