issues
search
facebookresearch
/
LLM-QAT
Code repo for the paper "LLM-QAT Data-Free Quantization Aware Training for Large Language Models"
Other
249
stars
24
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
Does LLM-QAT support group-wise quantization?
#32
mxjmtxrm
opened
1 month ago
0
运行报错elif self.deepspeed:
#31
LSB0798
opened
2 months ago
0
How to run inference
#30
wangkuiyi
opened
3 months ago
0
Looks like an Incorrect READ ME file
#29
gdsaikrishna
closed
3 months ago
1
Question about the training cost
#28
KimythAnly
closed
3 months ago
1
Does this method support chat models as well as Llama-2 models?
#27
Saoyu99
opened
6 months ago
1
How long will it take to train
#26
XA23i
opened
10 months ago
1
Accuracy
#25
yileijin
opened
11 months ago
0
Suggest change the README
#24
jingyao-zhang
closed
12 months ago
1
FileNotFoundError: [Errno 2] No such file or directory: 'wiki2.jsonl'
#23
StiphyJay
opened
1 year ago
1
Training is not working.
#22
XinnuoXu
opened
1 year ago
0
Hi, 可以开源你们生成的训练数据吗，感谢！
#21
Xingrun-Xing
opened
1 year ago
1
docs: blockquote cite article format README
#20
guspan-tanadi
closed
9 months ago
0
Inconsistent results with LLM.int8() and SmoothQuant papers
#19
fxmarty
closed
1 year ago
1
Questions about the valid_dataset format
#18
TravisL24
closed
1 year ago
1
Questions about the valid_dataset format
#17
TravisL24
closed
1 year ago
0
run run_train.sh, CUDA out of memory
#16
priscilla-pan
closed
1 year ago
0
no smoothquant in QuantizeLinear
#15
priscilla-pan
closed
1 year ago
5
Is there an efficient way to generate data?
#14
benyang0506
closed
1 year ago
3
The choice of kd_loss_scale
#13
zhanlaoban
closed
1 year ago
1
can you provide inference example for QuantizeLinear in 8 8 8
#12
jackzhou121
closed
1 year ago
2
How should I save the 8 bit model？
#11
liguodongiot
closed
1 year ago
2
If 4-8-8 is used to do QAT, how to process weight in inference?
#10
jackzhou121
closed
1 year ago
1
No randomization operation for the first token in data generation phrase.
#9
xingyueye
closed
1 year ago
1
Why use clip_tensor[-2.0, 2.0] in the backward?
#8
jackzhou121
closed
1 year ago
1
APEX and FSDP can not run
#7
jackzhou121
closed
1 year ago
1
why generated data was not used
#6
jackzhou121
closed
1 year ago
1
Does this method support Bloom？
#5
18140663659
closed
1 year ago
1
Expects full precision but got torch.bfloat16 error
#4
liguodongiot
opened
1 year ago
1
Harcoded train paths and configuration for table in readme
#3
aitorormazabal
closed
1 year ago
1
Why do smoothquant dynamically in the forward() function of QuantizeLinear layer
#2
Starmys
closed
1 year ago
2
Change teaser image to relative path to properlly display on Github
#1
Lyken17
closed
1 year ago
0