SafeAILab/EAGLE

Issues

• The disk space need for generating data for train.

  #151 opened a month ago by bulaikexiansheng
  1

• Question regarding sampling technique in EAGLE papers

  #148 opened a month ago by llsj14
  2

• Inquiry about accept length results for EAGLE-Qwen2-7B-Instruct

  #143 opened 2 months ago by zhangtia16
  2

• DeepSeekV2

  #142 opened 2 months ago by fengyang95
  1

• Issue about the comparison with Lookahead

  #145 opened a month ago by lethean1
  1

• About the acceptance rate

  #147 opened a month ago by YunSeoHwan
  1

• The acceptance length is not being reproduced

  #146 opened a month ago by YunSeoHwan
  8

• Regarding the settings for depth, top_k, and total_token in gen_ea_answer_llama2chat.py.

  #144 opened a month ago by YunSeoHwan
  0

• VLLM does not support EAGLE Spec Decode when deploying EAGLE-Qwen2-7B-Instruct model

  #135 opened 2 months ago by crownz248
  1

• Deploy EAGLE-Qwen2 in vllm

  #130 opened 2 months ago by IvanDeng0
  4

• Why the bias of config in train scripts doesn't aligning at infererence?

  #141 opened 2 months ago by 870572761
  1

• How could I train llama3-8B?

  #139 opened 2 months ago by 870572761
  1

• What is "model.safetensors.index.json"?

  #138 opened 2 months ago by 870572761
  0

• LLama 3.1 8b training issue.

  #129 opened 3 months ago by dhananjaybhandiwad
  2

• Is the forward related to gen_ea_alpha_llama2chat.py fully implemented?

  #136 opened 2 months ago by YunSeoHwan
  3

• A question about Eagle.

  #137 opened 2 months ago by kechunFIVE
  0

• Could not reproduce Acceptance Length for Llama 2 7B Chat Model with chain drafts

  #132 opened 2 months ago by luyuzhe111
  5

• Qwen2 inference

  #117 opened 2 months ago by kechunFIVE
  10

• Data generation

  #133 opened 2 months ago by adamxyang
  0

• How can I inference a 70B model

  #134 opened 2 months ago by Siegfried-qgf
  0

• Memory Overflow Issue when Training Custom Dataset

  #131 opened 2 months ago by chenht2022
  1

• Regarding training data of EAGLE

  #128 opened 3 months ago by KanghoonYoon
  1

• Evaluating Eagle on HumanEval, GSM8K, Alpaca, CNN/DM, and Natural Questions Datasets

  #124 opened 3 months ago by garipovroma
  4

• Question on quantized LLAMA3 versions for use with EAGLE

  #127 opened 3 months ago by jin-eld
  0

• Batched speculation benchmarks? (incl. with compilation)

  #126 opened 3 months ago by poedator
  2

• How to use multi-GPU when gen data?

  #105 opened 3 months ago by jzzzf
  2

• python -m eagle.ge_data.allocation --outdir [path of data]

  #104 opened 3 months ago by zpz915
  2

• Training of Qwen2

  #125 opened 3 months ago by jzzzf
  1

• Can not reproduce speedup results reported in the paper.

  #121 opened 3 months ago by shaonan1993
  3

• what's the difference between ea_generate and eagenerate

  #123 opened 3 months ago by LSC527
  1

• Any hint to improve average accept length for fine tuned LLAMA 70b?

  #106 opened 3 months ago by yanjunplay
  4

• When Train the Auto-regression Head, RuntimeError: CUDA error: device-side assert triggered

  #122 opened 3 months ago by liuy-2019
  0

• Actual parameter setting for different sizes of the model?

  #118 opened 3 months ago by Lucas-TY
  1

• Some questions that arise when studying your code

  #120 opened 3 months ago by bulaikexiansheng
  2

• Can Ealge2 use multi-gpus to inference? I mean data parallelism

  #112 opened 3 months ago by Siegfried-qgf
  1

• Reproduction Bias

  #119 opened 3 months ago by SpectrezZ
  4

• Is this expected behavior?

  #113 opened 3 months ago by qwedaq
  2

• Will smaller models be supported?

  #116 opened 3 months ago by Puvoka
  1

• Question about EAGLE-2

  #110 opened 4 months ago by Yipeng1994
  5

• Support Qwen2 in Eagle's inference framwork

  #114 opened 4 months ago by xiaoyong-z
  3

• Confusion in Figure 5 (b) of EAGLE2 paper

  #115 opened 4 months ago by SimonSongg
  2

• error 'accept_length' in Eagle1 or 2?

  #95 opened 4 months ago by haiduo
  13

• Any plan to support Llama 405B (Llama 3.1)

  #103 opened 4 months ago by yanjunplay
  3

• Can Eagle2 use multi-gpus to inference？

  #111 opened 4 months ago by Siegfried-qgf
  1

• ge_data_all_llama2chat

  #108 opened 4 months ago by fousdfrf
  0

• What's the ideal loss and accuracy for eagle training

  #101 opened 4 months ago by yanjunplay
  3

• Best performance for fine tuned llama3 70b model

  #98 opened 4 months ago by yanjunplay
  2

• How to use the eagle model trained by myself

  #100 opened 4 months ago by yanjunplay
  2

• Training config for llama3

  #97 opened 4 months ago by yanjunplay
  2

• What's the relationship between total token and depth

  #96 opened 4 months ago by Lucas-TY
  1