ent d3688cc833 load speed: persist compile cache at $HF_HOME/.vllm_cache (was wiped each container boot, -42s/boot); propagate --load-format to mtp_serve; document startup time

2026-06-24 16:32:02 +10:00

..

mtp_serve.sh

load speed: persist compile cache at $HF_HOME/.vllm_cache (was wiped each container boot, -42s/boot); propagate --load-format to mtp_serve; document startup time

2026-06-24 16:32:02 +10:00

patch_fla_shmem.py

qwen3.5-122B-A10B on DGX Spark: vLLM + DFlash + dense-bandwidth stack, one-shot installer

2026-06-24 13:02:35 +10:00

patch_inc_hybrid.py

qwen3.5-122B-A10B on DGX Spark: vLLM + DFlash + dense-bandwidth stack, one-shot installer

2026-06-24 13:02:35 +10:00

patch_int8_lmhead_v3.py

int8 lm-head: output bf16 (match stock lm-head dtype) instead of fp32 — halves the logits buffer that was inflating vLLM's profiled activation reserve and shrinking the dense KV pool

2026-06-24 15:53:58 +10:00

patch_unify2.py

qwen3.5-122B-A10B on DGX Spark: vLLM + DFlash + dense-bandwidth stack, one-shot installer

2026-06-24 13:02:35 +10:00

serve.sh

load speed: --load-format fastsafetensors (read straight to device, no mmap) — weight load 463s -> 32s (14.5x), time-to-READY ~12min -> ~3min on GB10. Falls back to nogds automatically.

2026-06-24 16:30:33 +10:00