EvolvingLMMs-Lab/LLaVA-OneVision-2

llava-onevision2-consistency

Bilingual guide for running and interpreting LLaVA-OneVision2 HF vs Megatron consistency checks across TP and PP settings

Langue du texte source : Plusieurs langues

merge-ov2

Bilingual guide for merging ViT + LLM into LlavaOnevision2 HF checkpoint and validating weight/inference consistency

Langue du texte source : Plusieurs langues

commit-message

Guide for writing clear, consistent git commit messages following this repository's conventions

Langue du texte source : anglais

offline-packing-env-vars

Bilingual guide for the OFFLINE_PACKING_BMR and OFFLINE_PACKED_DATA environment variables that control LLaVA-OneVision2 training-side packing — what each gate does, why both must be enabled together, MBS=1 requirement, and the dead OFFLINE_PACKING_VQA branch

Langue du texte source : Plusieurs langues

distributed-offline-packing

Bilingual guide for running offline_packing/auto_pipe.sh across multiple nodes to produce padding-free packed WebDataset shards for SFT, with Energon Metadataset assembly

Langue du texte source : anglais

cu-lengths-attention-flow

Bilingual guide for understanding how cu_lengths controls attention behavior across ViT and LLM stages, and how patch_positions scope differs between the two

Langue du texte source : Plusieurs langues

length-pool-sort-dataset

Bilingual guide for understanding LengthPoolSortDataset cross-rank length synchronization mechanism in multi-GPU training

Langue du texte source : Plusieurs langues

megatron-checkpoint-layout

Bilingual guidance for Megatron checkpoint 1D 2D 3D mp_rank layouts across tensor pipeline and expert parallel dimensions

Langue du texte source : Plusieurs langues