Implement QLoRA-style NF4 4-bit quantization of a weight matrix plus a LoRA adapter, verif

Super-hard

SFT, Instruction Tuning & PEFT5 / 6

Chapter 115 / 6

SFT, Instruction Tuning & PEFT

Implement QLoRA-style NF4 4-bit quantization of a weight matrix plus a LoRA adapter, verif

Super-hard

Implement QLoRA-style NF4 4-bit quantization of a weight matrix plus a LoRA adapter, verifying the forward matches fp16 $+$ adapter within tolerance.

Target interface

quantize_nf4(W, block=64)

Implement the function/class skeleton in the editor. Any correct approach is accepted.

Hints0 / 2

Reference solution

Your own code stays in the editor.

solution.pyPython

Local draft

import torch

def quantize_nf4(W, block=64):
    raise NotImplementedError

def dequantize_nf4(idx, absmax, shape, pad, block=64):
    raise NotImplementedError

⌘/Ctrl↵to submit

AI review

No review yet

Submit your solution for a verdict, a score out of 100 and concrete feedback. Any correct approach passes.

Chapter 115 / 6

SFT, Instruction Tuning & PEFT

Implement QLoRA-style NF4 4-bit quantization of a weight matrix plus a LoRA adapter, verif

Super-hard

Implement QLoRA-style NF4 4-bit quantization of a weight matrix plus a LoRA adapter, verifying the forward matches fp16 $+$ adapter within tolerance.

Target interface

quantize_nf4(W, block=64)

Implement the function/class skeleton in the editor. Any correct approach is accepted.

Hints0 / 2

Reference solution

Your own code stays in the editor.