Lightning-AI/lightning-thunder

quantization: process tensors on meta device directly, maybe implement CPU quantization (if it is easy)

Open

#1,111 opened on Sep 6, 2024

View on GitHub
 (4 comments) (0 reactions) (0 assignees)Python (114 forks)github user discovery
good first issuetransforms

Repository metrics

Stars
 (1,460 stars)
PR merge metrics
 (PR metrics pending)

Description

Currently the BitsAndBytesLinearQuant4bit for submodule always calls bitsandbytes.functional.quantize_4bit. This is somewhat touchy for CPU tensors because quantize_4bit only works on GPU tensors but it is outright not so nice for meta tensors, where we only would need to get the right shapes.

https://github.com/Lightning-AI/lightning-thunder/blob/e64d347def39bb47101efafe4177adf9f77a63ec/thunder/transforms/quantization.py#L93-L103

Contributor guide

Research direction
Implement or modify quantization logic to handle meta tensors without calling bitsandbytes' quantize 4bit, which only works on GPU. For CPU, consider a simple quantization implementation if feasible.
Tech stack
pythonpytorch
Domain
backend
Issue type
Bug
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
Half day
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
Active
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
Mostly clear
Prerequisites
PythonPyTorch
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
60