Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Quantized KV cache: update quanto #31052

Merged
merged 6 commits into from
May 29, 2024
Merged

Quantized KV cache: update quanto #31052

Changes from 1 commit
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
23aaf95
quanto latest version was refactored
zucchini-nlp May 25, 2024
f829acf
add error msg
zucchini-nlp May 27, 2024
f851298
incorrect compare sign
zucchini-nlp May 27, 2024
92aa648
Merge branch 'huggingface:main' into quant
zucchini-nlp May 28, 2024
3d6492d
Update src/transformers/cache_utils.py
zucchini-nlp May 28, 2024
d58b991
Merge branch 'huggingface:main' into quant
zucchini-nlp May 29, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 4 additions & 2 deletions src/transformers/cache_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11,7 +11,7 @@


if is_quanto_available():
from quanto import QBitsTensor, qint2, qint4
from quanto import AffineQuantizer, MaxOptimizer, qint2, qint4

if is_hqq_available():
from hqq.core.quantize import Quantizer as HQQQuantizer
Expand Down Expand Up @@ -500,9 +500,11 @@ def __init__(self, cache_config: CacheConfig) -> None:
)

self.qtype = qint4 if self.nbits == 4 else qint2
self.optimizer = MaxOptimizer() # hardcode as it's the only one for per-channel quantization

def _quantize(self, tensor, axis):
qtensor = QBitsTensor.quantize(tensor, axis=axis, qtype=self.qtype, group_size=self.q_group_size)
scale, zeropoint = self.optimizer(tensor, self.qtype.bits, axis, self.q_group_size)
qtensor = AffineQuantizer.apply(tensor, self.qtype, axis, self.q_group_size, scale, zeropoint)
return qtensor

def _dequantize(self, qtensor):
Expand Down
Loading