mirrors/ollama-for-amd
1
0
Fork 0
mirror of https://github.com/likelovewant/ollama-for-amd.git synced 2025-12-21 22:33:56 +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

2
model/models/bert/embed.go
View File

@@ -29,7 +29,7 @@ type Model struct {
// Forward implements model.Model.
func (m *Model) Forward(ctx ml.Context, batch input.Batch) (ml.Tensor, error) {
hiddenStates := m.TokenEmbedding.Forward(ctx, batch.Inputs)
hiddenStates = hiddenStates.Add(ctx, m.TypeEmbedding.Weight.View(ctx, 0, m.hiddenSize))
hiddenStates = hiddenStates.Add(ctx, m.TypeEmbedding.Weight.Slice(ctx, 1, 0, 1, 1))
hiddenStates = hiddenStates.Add(ctx, m.PositionEmbedding.Forward(ctx, ctx.Input().FromInts(batch.Positions, len(batch.Positions))))
hiddenStates = m.TokenEmbeddingNorm.Forward(ctx, hiddenStates, m.eps)