aleleba 0cf8bb1df7 Phase 6.4.25: point vllm-eval at v2-bf16 and record gate 2/5-holdout results ...

Gate 2 on v2-bf16: PASS, 192/200 = 96.0% (penpot 60/60 = 100.0%).
Gate 5 holdout mode (60 prompts, fallback trigger): FAIL, forbidden-API
rate 15.0% (9/60) vs threshold <=5% -- model still occasionally invents
penpot.importImage(...) for photo/URL prompts.

2026-08-03 07:31:04 +00:00

data

Phase 6.4.25: point vllm-eval at v2-bf16 and record gate 2/5-holdout results

2026-08-03 07:31:04 +00:00

out

Phase 6.4.21: version the LoRA #2 adapter (Penpot design capability)

2026-07-31 12:43:45 +00:00

scripts

Phase 6.4.20: fix eval OOM by pinning per_device_eval_batch_size

2026-07-31 00:44:54 +00:00

.gitignore

Phase 6.4.21: version the LoRA #2 adapter (Penpot design capability)

2026-07-31 12:43:45 +00:00

constraints.txt

Fase 0: scripts de verificacion de hardware e inspeccion de modulos

2026-07-28 03:52:10 +00:00

docker-compose.eval.yml

Phase 6.4.25: point vllm-eval at v2-bf16 and record gate 2/5-holdout results

2026-08-03 07:31:04 +00:00

docker-compose.yml

Fase 0: estructura base de carpetas y docker-compose de training

2026-07-28 03:42:37 +00:00

requirements.train.txt

Phase 6.2: pin the training environment and make 10_train.py configurable

2026-07-30 17:01:41 +00:00

S

Description

No description provided

465 MiB

Languages

Python 92.8%

Jinja 7.2%