vpermilp commited on
Commit
987fbcc
·
verified ·
1 Parent(s): 155a8ac

Add experimental FP8 Whisper Medium weights

Browse files

OCP E4M3 linear weights with per-output-channel FP16 scales; embeddings and convolutions retained at FP16.

Files changed (1) hide show
  1. model-fp8.safetensors +3 -0
model-fp8.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fea31295a43f7ddbdc429dd5b7b59e5b7bebef9ee8aee7f69bfd243897c59638
3
+ size 828788352