matt/ollama37
1
0
Fork 0
mirror of https://github.com/dogkeeper886/ollama37.git synced 2025-12-10 07:46:59 +00:00
Code Issues Packages Projects Releases Wiki Activity

backend: API to support full precision matmul

Browse Source 
Most tensor backends try to optimize performance by using a lower
precision for matmuls. However, some operations (such as kq) on
some models are sensitive to this and require full precision.
This commit is contained in:
Jesse Gross
2025-02-13 10:01:14 -08:00
committed by Jesse Gross
parent 4d4463b2bd
commit d773b7d671
4 changed files with 12 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
ml/backend.go
View File

@@ -66,6 +66,7 @@ type Tensor interface {
Add(ctx Context, t2 Tensor) Tensor
Mul(ctx Context, t2 Tensor) Tensor
Mulmat(ctx Context, t2 Tensor) Tensor
MulmatFullPrec(ctx Context, t2 Tensor) Tensor
Softmax(ctx Context) Tensor
LayerNorm(ctx Context, weight, bias Tensor, eps float32) Tensor