Repository navigation

S26 E6: FP8 quantized rollout and full-parameter dedicated-engine island - #290

Closed

michaellchung wants to merge 23 commits into

s25-finalfrom

s26-e6-quant

Closed

S26 E6: FP8 quantized rollout and full-parameter dedicated-engine island#290
michaellchung wants to merge 23 commits into
s25-finalfrom
s26-e6-quant

Commits on Oct 20, 2026
- E6: quantized rollout tiers with an FFN/routed-expert scope rule
  
  chowhsu
  and
  claude
  committed
  ce3822d
- E6: FP8 card allowlist from the GPU spec; keep tier-0 argv golden
  
  chowhsu
  and
  claude
  committed
  20f603e
- E6: engine sampler tier settings, dense expert count, CPU proofs
  
  chowhsu
  and
  claude
  committed
  6b74178
- E6: load Miles process_utils from its file in the env test
  
  chowhsu
  and
  claude
  committed
  a215af7
Commits on Oct 21, 2026
- E6: fixes from the H100 gate runs (fused ignore aliases, token scoring, BF16 KV)
  
  chowhsu
  and
  claude
  committed
  4854dda
- E6: refuse tier 1 on verl, set verl max_model_len, native-FP8 gate probes
  
  chowhsu
  and
  claude
  committed
  ffe86d0
- E6: pin the FP8-fixed ports image; tier 1 trainer FP8 source (Miles native + ignore list)
  
  chowhsu
  and
  claude
  committed
  d9a2703
- E6: Miles RL drivers for the tier groups (Modal H100!)
  
  chowhsu
  and
  claude
  committed
  7f100f4
- E6: full-parameter pair needs a fixed-partition sampler island
  
  chowhsu
  and
  claude
  committed
  8d1df81
- E6: Dockerfile and tests follow the new image pins; neutral flag help
  
  chowhsu
  and
  claude
  committed
  e3da116
- E6: pair islands are 2xH100 each (fixed partition applies to every island)
  
  chowhsu
  and
  claude
  committed
  2011906
- E6: tier 2 also runs as one launch per role (--rl-sample-role); pair driver uses two launches
  
  chowhsu
  and
  claude
  committed
  aa3b97c
- E6: per-round parser of the Miles RL runs
  
  chowhsu
  and
  claude
  committed
  d7c79be
- Modal store: no reload while a writer holds uncommitted files
  
  chowhsu
  and
  claude
  committed
  be553cf
- E6: pair trainer keeps the Adam state on the host (OOM on one H100)
  
  chowhsu
  and
  claude
  committed
  beaf4e0
- Modal store: the write guard is a container-wide flock
  
  chowhsu
  and
  claude
  committed
  1fe2c76
- E6: tier 1 on Miles defaults to the trainer FP8 source
  
  chowhsu
  and
  claude
  committed
  f4cd14e
- RL: full parameters on one no-sync island with dedicated engine cards
  
  chowhsu
  and
  claude
  committed
  bce94af
- E6: pin the ports image with the SGLang compressed-tensors FP8 reload
  
  chowhsu
  and
  claude
  committed
  8fe48c9
- E6: neutral fp8-source help; full-parameter dedicated-engine group is 1 train + 1 engine card
  
  chowhsu
  and
  claude
  committed
  f74ac99
- E6: full-parameter dedicated-engine group runs without round cuts
  
  chowhsu
  and
  claude
  committed
  dfa4a58
- E6: hold the store writer guard across a round cut; cuts with host Adam
  
  chowhsu
  and
  claude
  committed
  84011ee
Commits on Oct 22, 2026
- E6: 3c rerun cuts every 5 rounds; stop both apps after 15 min without a new step
  
  chowhsu
  and
  claude
  committed
  2ab01f8

Commits

Commits on Oct 20, 2026

Commits on Oct 21, 2026

Commits on Oct 22, 2026