ProofOfConcept e7e1855b87 research: ORPO — combined SFT + preference in one step, ideal for behavioral training ... ORPO applies 'minor penalty for disfavored response' during SFT. Single learning rate, single pass, both objectives. Implements the bypass mechanism naturally (minor penalty = disfavor, not remove). The loss landscape geometry explains the 40x lr gap: SFT is a valley, DPO is a ridge, ORPO combines both. LLaMA-Factory supports it. Dream loop generates triplets (context + preferred + rejected).		2026-03-31 02:51:26 -04:00
..
checkpoint	checkpoint: sync live weights back into model safetensors in-place	2026-03-30 22:55:23 -04:00
research	research: ORPO — combined SFT + preference in one step, ideal for behavioral training	2026-03-31 02:51:26 -04:00
apollo_mini.py	apollo: rewrite optimizer from paper's math + add research analysis	2026-03-31 00:54:17 -04:00
apollo_worker.py	apollo: make rank configurable (default 1 = Mini, higher ranks for experimentation)	2026-03-30 22:06:31 -04:00
DESIGN.md	DESIGN.md: complete rewrite reflecting validated architecture	2026-03-31 00:42:53 -04:00
export_weights.py	apollo-mini training system: initial implementation	2026-03-30 22:02:37 -04:00
extract_steering_vector.py	steering vector extraction script — answering Q5 experimentally	2026-03-31 02:28:18 -04:00
first_training_step.py	first_training_step.py: ready for Kent to run	2026-03-31 01:59:52 -04:00
start_vllm_with_apollo.sh	vllm launcher with apollo hook	2026-03-30 22:24:02 -04:00
train.py	apollo-mini training system: initial implementation	2026-03-30 22:02:37 -04:00
training_example.py	apollo-mini training system: initial implementation	2026-03-30 22:02:37 -04:00
vllm_export_hook.py	apollo-checkpoint: efficient diff-based GPU weight checkpointing	2026-03-30 22:53:17 -04:00
weight_mapping.py	weight_mapping: strip language_model prefix to match HF text model names	2026-03-30 23:11:03 -04:00