LLaVA-OneVision-2
Train and validate open multimodal AI models end to end
All Skills in This Repository (8)
Pure Emerald Level Indicators
EvolvingLMMs-Labcu-lengths-attention-flow
Analyze cu_lengths attention boundaries across ViT and LLM stages in LLaVA-OneVision2.
EvolvingLMMs-Labdistributed-offline-packing
Pack SFT JSONL samples into Energon WebDataset shards across multiple nodes.
EvolvingLMMs-Labmegatron-checkpoint-layout
Diagnose and convert Megatron checkpoint shard layouts across TP, PP, and EP dimensions.
EvolvingLMMs-Labllava-onevision2-consistency
Compare HuggingFace and Megatron/MCore checkpoints for LLaVA-OneVision2 behavioral consistency.
EvolvingLMMs-Lablength-pool-sort-dataset
Explain cross-rank length synchronization in LengthPoolSortDataset for multi-GPU training.
EvolvingLMMs-Labmerge-ov2
Merge standalone ViT encoder and LLM into a unified LlavaOnevision2 HuggingFace checkpoint.
EvolvingLMMs-Laboffline-packing-env-vars
Configure OFFLINE_PACKING_BMR and OFFLINE_PACKED_DATA for LLaVA-OneVision2 training.
EvolvingLMMs-Labcommit-message
Enforce Conventional Commits formatting with type: subject and body guidelines.
Frequently Asked Questions
FAQPage SchemaHow to install LLaVA-OneVision-2?▼
Run `npx skills add EvolvingLMMs-Lab/LLaVA-OneVision-2 --all -g -y` in your terminal to install all skills in this suite globally.
What does LLaVA-OneVision-2 do?▼
It is a fully open 8B multimodal model that understands images, long videos, and 3D spatial scenes in one architecture, with all training code, data, and checkpoints released.
How do I convert checkpoints between HuggingFace and Megatron?▼
The included skills guide your AI agent through HF-to-Megatron conversion, TP/PP re-sharding, and round-trip validation that is verified bitwise identical.
How do I pack training data for multi-GPU runs?▼
The distributed offline packing skill walks the agent through splitting JSONL data, packing it into WebDataset shards across nodes, and assembling the final training dataset.
Do I need deep ML infrastructure experience to use these skills?▼
No. The skills give your AI agent step-by-step commands, pitfall warnings, and validation checklists so it can execute the workflows from plain-English requests.
Related Repositories in Software Engineering
View All in Software Engineering→
openclawopenclaw
Run a personal AI assistant across your devices and chat apps
obrasuperpowers
Gives coding agents a disciplined workflow from idea to merged code
facebookreact
AI agent skills for building, testing, and porting React core