ViperEkura 9d3ccfdffc fix: incremental decode to avoid U+FFFD in streaming ...

- StreamDecoder buffers incomplete multi-byte sequences
- Task.decode_new_token replaces per-token decode in scheduler
- flush_remaining emits final buffered text on task finish

2026-07-18 13:05:36 +08:00

..

__init__.py

perf: replace paged KV cache with contiguous ContiguousCache, decode all groups

2026-07-05 11:34:36 +08:00

cache.py

perf: merge decode batch for 10x throughput

2026-07-18 08:50:46 +08:00

executor.py

perf: merge decode batch for 10x throughput

2026-07-18 08:50:46 +08:00

scheduler.py

fix: incremental decode to avoid U+FFFD in streaming

2026-07-18 13:05:36 +08:00

task.py

fix: incremental decode to avoid U+FFFD in streaming

2026-07-18 13:05:36 +08:00