Search results

The legacy Medusa Head structure is inconsistent with the new one.

Status: Open.

#140 In FasterDecoding/Medusa;

· Jianhua-Cui opened on Oct 11, 2025

loss value nan

Status: Open.

#139 In FasterDecoding/Medusa;

· wittycheng opened on Aug 22, 2025

why什么使用连续提前开辟好的KVcache? 这样本身就引入一个比huggingface实现快的因素？

Status: Open.

#137 In FasterDecoding/Medusa;

· wenhaoli-xmu opened on Jun 19, 2025

why my train medusa head result is {'medusa0_top1': nan, 'medusa0_loss': nan, 'medusa1_top1': nan, 'medusa1_loss': nan, 'medusa2_top1': nan, 'medusa2_loss': nan, 'epoch': 0}

Status: Open.

#134 In FasterDecoding/Medusa;

· Mewo518 opened on Mar 22, 2025

Support for other types of LLM

Status: Open.

#132 In FasterDecoding/Medusa;

· Shubin-vadim opened on Mar 4, 2025

Is Medusa(-2) compatible with vision language models (VLMs) ?

Status: Open.

#129 In FasterDecoding/Medusa;

· MoritzLaurer opened on Dec 3, 2024

Question about the Tree Attention Mechanism

Status: Open.

#127 In FasterDecoding/Medusa;

· chansonzhang opened on Nov 18, 2024

About Code compatability

Status: Open.

#126 In FasterDecoding/Medusa;

· kimjoohyungsd opened on Nov 1, 2024

Ask for data recipe to reproduce Medusa-2

Status: Open.

#125 In FasterDecoding/Medusa;

· Achazwl opened on Oct 24, 2024

[report bug] Encountered when inferencing with Mistral models

Status: Open.

#124 In FasterDecoding/Medusa;

· shrango opened on Oct 10, 2024

How to run inference of a baseline model without medusa support?

Status: Open.

#123 In FasterDecoding/Medusa;

· kailashg26 opened on Sep 16, 2024

About the Tree Sparsity

Status: Open.

#122 In FasterDecoding/Medusa;

· PineTreeWss opened on Sep 13, 2024