issues
search
SafeAILab
/
EAGLE
Official Implementation of EAGLE
https://arxiv.org/pdf/2406.16858
Apache License 2.0
612
stars
57
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
Are you going to support Gemma-2?
#85
jumasheff
opened
23 hours ago
0
have you guys tried other speculative draft model with eagle-2 method?
#84
hanlinxuy
opened
1 day ago
0
when eagle-2 will be opensourced?
#83
hanlinxuy
closed
1 day ago
1
Question about past_key_value modification
#82
baihuajun24
opened
1 week ago
1
I use Chinese and English mixed dataset and a another custom 7B model to train eagle,but the performance in Chinese is not good enough
#81
2024WY
opened
1 week ago
1
Question about alpha
#80
zkqq
opened
3 weeks ago
3
reproducing eagle on mistral-7b-v0.3-instruct
#79
alcholiclg
opened
3 weeks ago
4
Is there a guide to Train eagle heads on custom models?
#78
cryoco
opened
3 weeks ago
4
Why does instruction_len need to subtract 2?
#77
cdliang11
closed
3 weeks ago
1
Why does EAGLE remove the input_layernorm of llama?
#76
bisunny
opened
1 month ago
1
Quality loss in greedy mode.
#75
w32zhong
closed
1 month ago
4
Question about sampling
#74
Monoclinic
closed
1 day ago
6
Using EAGLE will slow down inference
#73
zkqq
opened
1 month ago
7
During training, only one decoder layer(draft model) is trained, right?
#72
CRH400AF-A
closed
1 month ago
1
Updates on training moe eagle model?
#71
cyLi-Tiger
opened
1 month ago
0
difference between the target input_ids and original LLMs in function getkacc()
#70
kafkayu
opened
1 month ago
0
bug fix: hidden_states nan
#69
dtlzhuangz
closed
1 month ago
1
The initiization of tree_buffers
#68
kafkayu
closed
1 month ago
0
Llama 3 support?
#67
joshmackwilliams
opened
1 month ago
3
new model
#66
shunxing12345
opened
2 months ago
6
About reproducing speedup ratio
#65
hyomin14
opened
2 months ago
1
training efficiency
#64
pengfeiwu1999
opened
2 months ago
1
Qs on self.act
#63
cyLi-Tiger
closed
2 months ago
1
Is there scripts to calculate the overall acceptance rate?
#62
sugar1946
opened
2 months ago
1
About KVCache of Eagle vs Origin LLM
#61
jcao-ai
closed
2 months ago
0
Questions about the EAGLE parameters, not match the pytorch_model.bin
#60
DeclK
closed
2 months ago
2
Questions about pre-allocated kv_cache.
#59
Hevans123
closed
2 months ago
1
Question on draft process
#58
cyLi-Tiger
closed
2 months ago
1
About project structure
#57
zsnoob
closed
2 months ago
1
run with vllm
#56
Hevans123
closed
2 months ago
1
How to support new model?
#55
wanghuihhh
closed
2 months ago
1
bsne1 branch "last_hidden = out_hidden[ab,last_nopadding][:,None]" gets wrong
#54
lihuahua123
closed
2 months ago
4
About tree_buffer
#53
reflectionie
closed
3 months ago
2
VLLM contribution
#52
arunpatala
closed
3 months ago
2
can you share ShareGPT dataset used to train the models?
#51
arunpatala
closed
3 months ago
3
Generation quality loss
#50
cyLi-Tiger
closed
3 months ago
7
Can Eagle improve the inference throughput in continuous batch mode?
#49
xiongqisong
closed
3 months ago
4
Inference on the CPU, the speed improvement is limited
#48
tigerliu10
closed
2 months ago
2
fix: update verify process
#47
cyLi-Tiger
closed
3 months ago
1
How many decoder layer needs to add in EAGLE?
#46
yjdy
closed
2 months ago
5
RuntimeError: probability tensor contains either `inf`, `nan` or element < 0
#45
cyLi-Tiger
closed
4 months ago
5
vicuna 7b oom
#44
yjdy
closed
4 months ago
0
Support vLLM
#43
ruidongtd
opened
4 months ago
0
Data processing script (ge_data/allocation.py) script does not work out of the box
#42
avnermay
opened
4 months ago
1
Can EAGLE be applied to any auto-regressive model?
#41
sam-sepiola
closed
2 months ago
3
fix typo in README.md
#40
sonald
closed
4 months ago
0
Is there scripts to calculate the acceptance rate at each position?
#39
agentup
closed
4 months ago
1
How to train LLama-70B eagle?
#38
YixinSong-e
closed
3 months ago
4
KV Cache initialization throwing an error
#37
haim-barad
closed
4 months ago
8
Can EAGLE actually improve throughput?
#36
chenzhengda
closed
4 months ago
1
Next