mirrors/ollama-for-amd
1
0
Fork 0
mirror of https://github.com/likelovewant/ollama-for-amd.git synced 2025-12-21 14:26:30 +00:00
Code Issues Packages Projects Releases Wiki Activity

s/From*Slice/From*s/ (#12255)

Browse Source
This commit is contained in:
Michael Yang
2025-10-28 12:08:49 -07:00
committed by GitHub
parent 15c7d30d9a
commit 1188f408dd
24 changed files with 95 additions and 76 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
runner/ollamarunner/multimodal.go
View File

@@ -102,7 +102,7 @@ func (m multimodalStore) getTensor(backend ml.Backend, ctx ml.Context, in ml.Ten
for i, t := range entry.mm {
if in == t.Tensor {
if !reserve {
return ctx.Input().FromFloatSlice(entry.data[i], t.Tensor.Shape()...), nil
return ctx.Input().FromFloats(entry.data[i], t.Tensor.Shape()...), nil
} else {
return ctx.Input().Empty(t.Tensor.DType(), t.Tensor.Shape()...), nil
}

6
runner/ollamarunner/runner.go
View File

@@ -599,7 +599,7 @@ func (s *Server) forwardBatch(pendingBatch batchState) (nextBatch batchState, er
// Actual batchInputs values will be injected into the batch.Inputs tensor before calling Compute
batch.Inputs = nextBatch.ctx.Input().Empty(ml.DTypeI32, len(batchInputs))
batch.Outputs = nextBatch.ctx.Input().FromIntSlice(batchOutputs, len(batchOutputs))
batch.Outputs = nextBatch.ctx.Input().FromInts(batchOutputs, len(batchOutputs))
nextBatch.modelOutput, err = model.Forward(nextBatch.ctx, s.model, batch)
if err != nil {
err = fmt.Errorf("failed to build graph: %w", err)
@@ -692,7 +692,7 @@ func (s *Server) computeBatch(activeBatch batchState) {
// At this point the seqs are ready for forwardBatch to move forward so unblock
s.mu.Unlock()
activeBatch.batch.Inputs.SetValueFromIntSlice(batchInputs)
activeBatch.batch.Inputs.FromInts(batchInputs)
activeBatch.ctx.ComputeWithNotify(
func() {
logutil.Trace("computeBatch: signaling computeStartedCh", "batchID", activeBatch.id)
@@ -1090,7 +1090,7 @@ func (s *Server) reserveWorstCaseGraph() error {
batch.Positions[i] = int32(i)
}
batch.Inputs = ctx.Input().FromIntSlice(batchInputs, len(batchInputs))
batch.Inputs = ctx.Input().FromInts(batchInputs, len(batchInputs))
batch.Outputs = ctx.Input().Empty(ml.DTypeI32, s.parallel)
cache := s.model.Config().Cache