Fine-Tuning Low-Bit Models with Gradient in Quantized Code Space

Name: Fine-Tuning Low-Bit Models with Gradient in Quantized Code Space
Uploaded: 2026-09-09T01:50:31.198648+00:00
Duration: 491 s
Description: The episode discusses 'Fine-Tuning Low-Bit Models with Gradient in Quantized Code Space.' Hosts explain how gradient descent can be applied to quantized code spaces, allowing for efficient fine-tuning. They further detail using low-rank decomposition to structurally optimize parameter updates, making large models more accessible and manageable.

The episode discusses 'Fine-Tuning Low-Bit Models with Gradient in Quantized Code Space.' Hosts explain how gradient descent can be applied to quantized code spaces, allowing for efficient fine-tuning. They further detail using low-rank decomposition to structurally optimize parameter updates, making large models more accessible and manageable.

Open the full episode