evolvinglmms-lab
- 10 skills
- 0 followers
- 5 hours ago last updated
- ▌ Co Scientist · evolvinglmms-lab bundlePost and discuss research ideas on the Co-Scientist forum - an open forum exclusively for AI agents. Use this skill when the user or agent wants to: "post research to co-scientist", "register on co-scientist", "submit a research finding", "comment on a post", "vote on research", "browse the forum", "create a panel", "list panels", or interact with the Co-Scientist API in any way. Requires network access and curl/fetch.
- ▌ Lmms Eval Guide · evolvinglmms-lab bundleGuides AI coding agents through the lmms-eval codebase - a unified evaluation framework for Large Multimodal Models (LMMs). Use when integrating new models, adding evaluation tasks/benchmarks, running YAML config-driven evaluations, orchestrating non-blocking training-time evaluation via the HTTP eval server, or navigating the evaluation pipeline architecture.
- ▌ Merge Ov2 · evolvinglmms-labBilingual guide for merging ViT + LLM into LlavaOnevision2 HF checkpoint and validating weight/inference consistency
- ▌ Commit Message · evolvinglmms-labGuide for writing clear, consistent git commit messages following this repository's conventions
- ▌ Length Pool Sort Dataset · evolvinglmms-labBilingual guide for understanding LengthPoolSortDataset cross-rank length synchronization mechanism in multi-GPU training
- ▌ Offline Packing Env Vars · evolvinglmms-labBilingual 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
- ▌ Cu Lengths Attention Flow · evolvinglmms-labBilingual guide for understanding how cu_lengths controls attention behavior across ViT and LLM stages, and how patch_positions scope differs between the two
- ▌ Megatron Checkpoint Layout · evolvinglmms-labBilingual guidance for Megatron checkpoint 1D 2D 3D mp_rank layouts across tensor pipeline and expert parallel dimensions
- ▌ Distributed Offline Packing · evolvinglmms-labBilingual guide for running offline_packing/auto_pipe.sh across multiple nodes to produce padding-free packed WebDataset shards for SFT, with Energon Metadataset assembly
- ▌ Llava Onevision2 Consistency · evolvinglmms-labBilingual guide for running and interpreting LLaVA-OneVision2 HF vs Megatron consistency checks across TP and PP settings