Daniel Hiltgen dba1e27fa8 llama: enable FA on CUDA CC 6.x GPUs (#16994) ...

Recent upstream Pascal kernel fixes let us compile native SM60/SM61 kernels again instead of relying on PTX JIT, so allow Flash Attention auto at runtime for CC 6.x devices.

Fixes #16591

Fixes #16754

2026-07-02 17:11:39 -07:00

..

backend/ggml/ggml/src/ggml-cuda/template-instances

runner: Remove CGO engines, use llama-server exclusively for GGML models (#16031)

2026-05-29 13:35:47 -07:00

backend.go

Add support for gemma4 (#15214)

2026-04-02 11:33:33 -07:00

device_test.go

llama: enable FA on CUDA CC 6.x GPUs (#16994)

2026-07-02 17:11:39 -07:00

device.go

llama: enable FA on CUDA CC 6.x GPUs (#16994)

2026-07-02 17:11:39 -07:00

path_test.go

runner: Remove CGO engines, use llama-server exclusively for GGML models (#16031)

2026-05-29 13:35:47 -07:00

path.go

runner: Remove CGO engines, use llama-server exclusively for GGML models (#16031)

2026-05-29 13:35:47 -07:00