Search results

Does EAGLE3 support offloading?

Status: Open.

#322 In SafeAILab/EAGLE;

· zszdsze opened on Jan 12, 2026

[Question]Why do hidden states extracted from numerical data typically lead to suboptimal performance?

Status: Open.

#320 In SafeAILab/EAGLE;

· 330205812 opened on Jan 9, 2026

<code>loss_mask</code> not adjusted after target shift, causing loss on invalid padding

Status: Open.

#317 In SafeAILab/EAGLE;

· oluxembourg opened on Dec 22, 2025

Fix dimension mismatch error when reproducing Eagle3-Qwen3-4B (hidden_size != head_dim * num_heads)

Status: Open.

#315 In SafeAILab/EAGLE;

· gcy-hw95 opened on Nov 21, 2025

ImportError: transformers.utils.LossKwargs removed in transformers&gt;=4.54 breaks import in modeling_qwen3_kv.py

Status: Open.

#306 In SafeAILab/EAGLE;

· timxu826 opened on Oct 11, 2025

Some thoughts on tree decoding

Status: Open.

#301 In SafeAILab/EAGLE;

· ggg-s opened on Sep 18, 2025

Question regarding new_token in naivegenerate

Status: Open.

#296 In SafeAILab/EAGLE;

· aladinggit opened on Sep 11, 2025

Matrix shape not match with Qwen3-8B and Tengyunw/qwen3_8b_eagle3

Status: Open.

#294 In SafeAILab/EAGLE;

· wyattxuanyang opened on Sep 9, 2025

next token embedding was not used

Status: Open.

#293 In SafeAILab/EAGLE;

· jos-fisher-jejgkej opened on Sep 9, 2025

Questions about eagenerate

Status: Open.

#288 In SafeAILab/EAGLE;

· ebubekir-pulat opened on Sep 2, 2025

没找到Eagle2对应DeepSeek-R1-Distill-LLaMA-8B的权重文件下载链接

Status: Open.

#287 In SafeAILab/EAGLE;

· minaduki-sora opened on Aug 29, 2025

Training loss does not decrease

Status: Open.

#286 In SafeAILab/EAGLE;

· fan-niu opened on Aug 21, 2025