Views
No views yet
liuhaotian/llava-v1.6-vicuna-7b,
trained on a small (50-row) image+question dataset as a pipeline
proof-of-concept, not a production drafter. Part of the
VL-FLASH project's Phase 5
(LLaVA-1.6 port).DFlashDraftModel, 5 layers, hidden_size=4096,
block_size=16 (config: configs/llava-v1.6-vicuna-7b-dflash.json in
the training fork).min(answer_token_len - 1, block_size - 1) — LLaVA's
real VQA answers are often a single MCQ letter, so accept_len is
capped by each row's own answer length, not by block_size.accept_len.--max-tokens 60
truncating some training answers mid-generation) — fixed in the
pipeline after this checkpoint was trained, not re-trained against the
fix. See the repo's NOTES.md "Export + cold-start replay gate" section
for the full breakdown.scripts/gates/llava_text_only/RUNBOOK.md
in the training fork.