highperfocused/ollama
1
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2025-08-25 09:51:25 +02:00
Code Issues Packages Projects Releases Wiki Activity

feat: qwen3 dense and sparse models (#10708)

Browse Source 
* feat: qwen3 dense
* feat: qwen3moe
* fix llama4 moe
This commit is contained in:
Michael Yang
2025-05-21 10:21:07 -07:00
committed by GitHub
parent 139f84cf21
commit e0ed984cde
5 changed files with 258 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
ml/backend.go
View File

@@ -128,6 +128,8 @@ type Tensor interface {
Neg(ctx Context) Tensor
Add(ctx Context, t2 Tensor) Tensor
Mul(ctx Context, t2 Tensor) Tensor
Div(ctx Context, t2 Tensor) Tensor
Mulmat(ctx Context, t2 Tensor) Tensor
MulmatFullPrec(ctx Context, t2 Tensor) Tensor
MulmatID(ctx Context, t2, ids Tensor) Tensor
@@ -136,6 +138,7 @@ type Tensor interface {
LayerNorm(ctx Context, weight, bias Tensor, eps float32) Tensor
RMSNorm(ctx Context, weight Tensor, eps float32) Tensor
Scale(ctx Context, s float64) Tensor
SumRows(ctx Context) Tensor
AvgPool2D(ctx Context, k, s int, p float32) Tensor
Conv2D(ctx Context, weight Tensor, s0, s1, p0, p1, d0, d1 int) Tensor

14
ml/backend/ggml/ggml.go
View File

@@ -887,6 +887,13 @@ func (t *Tensor) Mul(ctx ml.Context, t2 ml.Tensor) ml.Tensor {
}
}
func (t *Tensor) Div(ctx ml.Context, t2 ml.Tensor) ml.Tensor {
return &Tensor{
b: t.b,
t: C.ggml_div(ctx.(*Context).ctx, t.t, t2.(*Tensor).t),
}
}
func (t *Tensor) Mulmat(ctx ml.Context, t2 ml.Tensor) ml.Tensor {
return &Tensor{
b: t.b,
@@ -1004,6 +1011,13 @@ func (t *Tensor) Scale(ctx ml.Context, s float64) ml.Tensor {
}
}
func (t *Tensor) SumRows(ctx ml.Context) ml.Tensor {
return &Tensor{
b: t.b,
t: C.ggml_sum_rows(ctx.(*Context).ctx, t.t),
}
}
func (t *Tensor) Softmax(ctx ml.Context) ml.Tensor {
return &Tensor{
b: t.b,