Andrew Park
1a843d1ece
orchestrator: SFT training + eval harness
...
Tokenize the cleaned routing data (all-turns masking for ChatML),
full-parameter FSDP SFT (fsdp4/8 configs for Lambda 8xH100), and an eval
backend that drives the fine-tuned orchestrator as the scored model over
GAIA/MMLU-Pro/SuperGPQA. Includes split-making, Braintrust upload, eval
sample rendering, scorer fixes, and a boxed-tool-call salvage for SFT'd
checkpoints that emit their action inside \boxed{}.
2026-07-15 13:47:06 -07:00
..
2026-05-25 23:51:29 +00:00
2026-07-15 13:45:07 -07:00
2026-05-27 09:36:26 -07:00
2026-03-20 18:35:18 -07:00
2026-07-02 14:49:18 -07:00
2026-07-02 14:49:18 -07:00
2026-07-02 14:49:18 -07:00
2026-07-02 14:49:18 -07:00
2026-03-17 00:13:49 +00:00
2026-05-25 23:56:55 +00:00
2026-07-01 13:02:13 -07:00
2026-07-06 14:10:22 -07:00
2026-07-12 14:48:22 -07:00
2026-04-17 15:22:17 -07:00
2026-05-20 03:36:55 +00:00
2026-06-03 16:59:29 -07:00
2026-03-20 18:35:18 -07:00
2026-04-13 15:03:03 -07:00
2026-07-02 14:49:18 -07:00
2026-07-02 14:49:18 -07:00
2026-06-24 19:58:49 -07:00
2026-05-14 20:11:24 -07:00
2026-04-17 11:02:56 -07:00
2026-05-14 20:11:24 -07:00
2026-06-24 19:58:49 -07:00
2026-03-20 18:35:18 -07:00
2026-03-20 18:35:18 -07:00
2026-04-03 11:05:41 -07:00
2026-06-11 13:49:37 -07:00
2026-06-01 11:57:44 -07:00
2026-07-02 14:49:18 -07:00
2026-03-31 19:25:29 +05:30
2026-04-13 15:03:39 -07:00
2026-06-20 17:08:24 -07:00
2026-07-02 14:49:18 -07:00
2026-03-20 18:35:18 -07:00
2026-07-15 13:47:06 -07:00
2026-07-02 14:49:18 -07:00
2026-07-02 14:49:18 -07:00
2026-03-20 18:35:18 -07:00
2026-03-12 17:29:39 +00:00
2026-06-24 19:58:49 -07:00
2026-04-03 11:05:41 -07:00
2026-04-17 11:02:56 -07:00
2026-06-14 18:19:00 -07:00