DeepSpeed
Fix fp8 gemm
#7265
Merged
Overview
Commits
12
Changes
View On GitHub

Fix fp8 gemm #7265

loadams merged 12 commits into deepspeedai:master from RezaYazdaniAminabadi:fix-fp8-gemm
RezaYazdaniAminabadi
sfc-gh-reyazda Optimize the fp-dequantizer to get high memory-BW utilization
10bad7de
sfc-gh-reyazda fix formating
c5ba68e3
RezaYazdaniAminabadi Merge branch 'master' into master
9975f753
RezaYazdaniAminabadi Merge branch 'microsoft:master' into master
f950f722
RezaYazdaniAminabadi Merge branch 'microsoft:master' into master
6381aaec
RezaYazdaniAminabadi Merge branch 'deepspeedai:master' into master
8a0f1d5e
test
85e533fa
fix the fp8-gemm by removing prefetching from bf16 conversion (New Tr…
01a24d15
RezaYazdaniAminabadi RezaYazdaniAminabadi requested a review from tohtana tohtana 308 days ago
RezaYazdaniAminabadi RezaYazdaniAminabadi requested a review from tjruwase tjruwase 308 days ago
jeffra formatting
e31f87f8
jeffra
jeffra approved these changes on 2025-04-30
sfc-gh-mwyatt
sfc-gh-mwyatt commented on 2025-04-30
sfc-gh-mwyatt Update deepspeed/ops/fp_quantizer/quantize.py
c4a90677
sfc-gh-mwyatt Update fp_quantizer.py
f6da0b58
sfc-gh-mwyatt sfc-gh-mwyatt requested a review from loadams loadams 308 days ago
sfc-gh-mwyatt sfc-gh-mwyatt requested a review from jomayeri jomayeri 308 days ago
loadams Merge branch 'master' into fix-fp8-gemm
931c69c5
loadams
loadams approved these changes on 2025-05-08
loadams loadams merged 069ec31c into master 300 days ago

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
None yet
Milestone
No milestone
Terms Of Use Privacy Policy Refund Policy Impressum