Files
ollama37/runner/llamarunner/runner.go
Jesse Gross b2a465296d runner: Release semaphore and improve error messages on failures
If we have an error after creating a new sequence but before
finding a slot for it, we return without releasing the semaphore.
This reduces our parallel sequences and eventually leads to deadlock.

In practice this should never happen because once we have acquired
the semaphore, we should always be able to find a slot. However, the
code is clearly not correct.
2025-03-30 19:21:54 -07:00

23 KiB