mirrors/ollama-for-amd
1
0
Fork 0
mirror of https://github.com/likelovewant/ollama-for-amd.git synced 2025-12-25 07:58:01 +00:00
Code Issues Packages Projects Releases Wiki Activity

chore: update models to use slice/chunk/chunksections (#12934)

Browse Source 
* use slice/chunks

* bert

* llama4

* gemma3n

* gptoss

* mistral3

* qwen3vl

* qwen25vl

* deepseek2

* remove unused ops
This commit is contained in:
Michael Yang
2025-11-13 15:20:12 -08:00
committed by GitHub
parent c114987523
commit 333203d871
13 changed files with 59 additions and 135 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
model/models/qwen25vl/model_vision.go
View File

@@ -13,9 +13,9 @@ import (
var batchSize int = 1
func rotateHalf(ctx ml.Context, t ml.Tensor) ml.Tensor {
x1 := t.View(ctx, 0, t.Dim(0)/2, t.Stride(1), t.Dim(1), t.Stride(2), t.Dim(2), t.Stride(3), t.Dim(3))
x2 := t.View(ctx, t.Stride(0)*t.Dim(0)/2, t.Dim(0)/2, t.Stride(1), t.Dim(1), t.Stride(2), t.Dim(2), t.Stride(3), t.Dim(3)).Contiguous(ctx)
return x2.Neg(ctx).Concat(ctx, x1, 0)
x1 := t.Slice(ctx, 0, 0, t.Dim(0)/2, 1)
x2 := t.Slice(ctx, 0, t.Dim(0)/2, t.Dim(0), 1).Contiguous(ctx)
return x2.Scale(ctx, -1).Concat(ctx, x1, 0)
}
func applyRotaryPositionalEmbedding(ctx ml.Context, t, cos, sin ml.Tensor) ml.Tensor {