dilab-zju self-speculative-decoding issues - Githubissues

dilab-zju / self-speculative-decoding

Code associated with the paper **Draft & Verify: Lossless Large Language Model Acceleration via Self-Speculative Decoding**

Apache License 2.0

114 stars 8 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

Bayesian Optimization Search Method for CodeLlama and human-eval

#17 MichaelMtl closed 3 weeks ago
4
Question about self-speculative + greedy decoding

#16 EganGu closed 2 months ago
2
Unable to reimplement 1.4x speedup with llama-2-chat

#15 hemingkx closed 2 months ago
2
some problems occured when execuating file search.ipynb

#14 hunzhizi closed 3 months ago
1
questions about Llama2-70b

#13 xinlong-yang closed 3 months ago
2
Unable to get 1.5 speedup using 13B model?

#12 w32zhong closed 4 months ago
1
Code for Llama-7b and Mistral

#11 DRXD1000 closed 5 months ago
1
Questions about modeling_llama.py

#10 qiyuangong closed 5 months ago
3
question about the training data for bayesian and model size

#9 irasin closed 6 months ago
2
skipped layer and llama2 70b chat

#8 jaemin-han closed 5 months ago
13
Proposal: Evaluating Faster, Deterministic Alternatives to Bayesian Optimization for Layer Skipping in Large Models

#7 azurespace closed 7 months ago
1
could you release the test dataset in your experiment?

#6 pengfeiwu1999 closed 8 months ago
1
Can I get skipped layer index set of LLaMA-70B tested on your paper??

#5 je1lee closed 8 months ago
0
KV cache footprint

#4 JYYHH closed 8 months ago
1
Can you share your prompt of LLaMA-2-70B?

#3 jaemin-han closed 8 months ago
2
Data on optimal layers to skip?

#2 KerfuffleV2 closed 9 months ago
5
the decoding code

#1 Ma-Yongqiang closed 8 months ago
5