PR #31742 [NPUW]Optimize token rate for dynamic LoRA. - SemanticDiff

Commits

Reuse infer request buffer for LoRA.

intelgaoxiong committed 303 days ago
Solved review comments in pr#31433.

intelgaoxiong committed 303 days ago
Fixed for CI.

intelgaoxiong committed 303 days ago
Set remote tensor for kvcache and prefill

intelgaoxiong committed 303 days ago
Choose device for pre-allocation.

intelgaoxiong committed 303 days ago
Move lora name matching functions to utils.

intelgaoxiong committed 303 days ago
Move allocMem to util.

intelgaoxiong committed 303 days ago
Minor change.

intelgaoxiong committed 303 days ago
No need to bump the serialization version.

intelgaoxiong committed 303 days ago

Loading