hao-ai-lab
- 31 skills
- 0 followers
- 11 hours ago last updated
- ▌ Kernel Bottleneck Tuning · hao-ai-labWorkflow for finding and fixing instruction-throughput bottlenecks in warp-specialized CuTeDSL kernels (FA4-style). Use when a kernel underperforms its roofline and you need to decide between reordering, eliminating, or re-pipelining work.
- ▌ Plan Mode Planning Assistant · hao-ai-labGenerates a structured experiment plan with compute-aware recommendations and saves it via the plan API endpoint.
- ▌ Wild V2 Idea · hao-ai-labIdeation prompt for Wild Loop V2 - generate high-quality, falsifiable research ideas with proposal and plan artifacts
- ▌ Alert Handler · hao-ai-labSystem prompt for handling experiment alerts. Provides diagnosis guidance, GPU wrapper context, action suggestions, and structured response from allowed choices.
- ▌ Agent Mode Research Assistant · hao-ai-labDefault system prompt for agent chat mode. Provides identity, environment context, compute awareness, API-driven job submission, and workflow reflection.
- ▌ Wild V2 Steer · hao-ai-labWraps user steering input with context signals for the model during a wild loop session
- ▌
- ▌ Wild V2 Planning · hao-ai-labPlanning prompt for Wild Loop V2 - iteration 0 phased planning with reflection, experiment ops, and analytics
- ▌ Wild V2 Iteration · hao-ai-labIteration prompt for Wild Loop V2 - execution iterations 1+ with phase-aware task tracking and reflection
- ▌ Wild V2 Reflection · hao-ai-labReflection prompt for Wild Loop V2 - runs after DONE to evaluate whether to continue or finalize
- ▌ Wild V2 Execution Ops Protocol · hao-ai-labSingle source of truth protocol for Wild V2 preflight, sweep/run auditability, GPU discovery, and parallel scheduling
- ▌ Wild V2 Sweep Run Audit Protocol · hao-ai-labMandatory protocol for creating auditable experiments via sweep/run endpoints in Wild V2
- ▌ Fastvideo Model Porting Alignment · hao-ai-labPorts new models into FastVideo with strict numerical alignment to official implementations. Use when adding a FastVideo model/pipeline, porting an official or Diffusers checkpoint, or debugging parity/alignment.
- ▌ Wild V2 Gpu Discovery Parallel Scheduling · hao-ai-labProtocol for GPU discovery and parallel run scheduling across local GPU and Slurm clusters
- ▌ Add Model · hao-ai-lab bundleManual /add-model workflow for implementing a FastVideo model or first-class component port after add-model-01-prep has staged reference code and weights. Organizes the port into numbered phases with conversion rules, component policies, parity gates, and handoff checks.
- ▌ CI Runner · hao-ai-labWork on FastVideo's Slurm-only, change-aware GPU CI lanes, static Buildkite graph, trusted ci-runner policy, lane scripts, and GB200 validation.
- ▌ Add Model 01 Prep · hao-ai-lab bundleUse at the start of a FastVideo model port to gather required inputs, inspect/download HF weights, clone and install the official reference repo in the current environment, create a local_tests README skeleton, and produce a handoff before conversion or implementation.
- ▌ Dreamverse Deploy · hao-ai-lab bundleUse when redeploying the migrated Dreamverse app backend and frontend on a chosen local GPU; tears down existing ports, launches services, and waits for readiness checks.
- ▌ Add Model 08 Trace · hao-ai-lab bundleUse during /add-model Phase 6 when component parity has failed and root cause requires layer-by-layer divergence analysis. Uses FastVideo activation trace first, falling back to custom hooks only for boundaries or stats the utility cannot observe.
- ▌ Add Model 02 Parity · hao-ai-lab bundleUse during /add-model after reference/architecture study to scaffold and later activate local FastVideo component parity tests. Emphasizes early test creation, official-reference loading, standardized FastVideo loading, and non-skip handoff gates.
- ▌ Seed Ssim References · hao-ai-labSeed HF reference artefacts for a single newly-added SSIM test (pixel `.mp4` for `run_text_to_video_similarity_test`-style tests, or latent `.pt` for `run_text_to_latent_similarity_test`-style tests). Runs the test on Modal L40S, downloads the generated artefacts via `modal volume get`, pauses for the user to verify (visual eyeball for mp4, numerics dump for pt), then uploads only that test's files to `FastVideo/ssim-reference-videos`. Use when a new `fastvideo/tests/ssim/test_*_similarity.py` has just been added and has no references on HF yet.
- ▌ Add Model 03 Port Dit · hao-ai-labUse during /add-model Phase 4 or Phase 6 to prototype or parity-debug one FastVideo-native DiT/transformer component.
- ▌ Add Model 04 Port Vae · hao-ai-labUse during /add-model Phase 4 or Phase 6 to prototype or parity-debug one FastVideo-native VAE component.
- ▌ Add Model 09 Pipeline · hao-ai-lab bundleUse during /add-model Phase 7 after all required component parity tests pass to define FastVideo pipeline wiring, configs, presets, registry entries, examples, smoke tests, and pipeline parity tests.
- ▌ Decompose Pipeline Pr · hao-ai-labDecompose an oversized FastVideo pipeline PR into a stack of independently-reviewable PRs. Tiers the diff by blast radius (invisible / dead code / cross-cutting infra / activation), produces a branch graph and worktree bootstrap, drafts the AGENTS.md manifest, flags missing tests on cross-cutting infra changes, and extracts lessons from the PR body.
- ▌ Add Model 10 Pr Review · hao-ai-labReview rubric for FastVideo PRs that add or modify model families, variants, first-class components, checkpoint conversion, pipelines, parity coverage, or generated-media quality baselines. Use when reviewing a PR whose diff touches fastvideo/models/, fastvideo/pipelines/basic/, fastvideo/registry.py, scripts/checkpoint_conversion/, fastvideo/tests/ssim/, or related model-port surfaces. Pairs with review-pr-link as a project-scoped review pass; produces findings, not fixes.
- ▌ Reseed Ssim References · hao-ai-labRe-seed HF reference videos for a single existing SSIM test on Modal L40S. Always backs up current refs locally first, regenerates on Modal, pauses for the user to eyeball before-vs-after quality, then overwrites the targeted model subtree on `FastVideo/ssim-reference-videos` with `--force`. Use when an intentional code change (model port fix, attention backend swap, kernel upgrade, hyperparameter change) has invalidated existing refs and they need to be regenerated. Pairs with `seed-ssim-references`, which is for first-time seeding only.
- ▌ Add Model 07 Conversion · hao-ai-lab bundleUse during /add-model Phase 5 to write and verify a FastVideo checkpoint conversion script after native component prototypes expose FastVideo state-dict keys/shapes.
- ▌ Add Model 05 Port Encoder · hao-ai-labUse during /add-model Phase 4 or Phase 6 to prototype or parity-debug one FastVideo-native text, image, audio, or compound encoder component.
- ▌ Add Model 06 Port Generic · hao-ai-labUse during /add-model Phase 4 or Phase 6 to prototype or parity-debug one non-DiT, non-VAE, non-encoder FastVideo component.
- ▌ Reseed Performance Baseline · hao-ai-labRe-seed the HF performance-tracking baseline for an intentional runtime, dependency, environment-caused benchmark shift, or reviewed v2 calibration using one or more reviewed normalized performance JSONs. Use when performance CI fails because metrics such as latency, throughput, component time, or peak memory changed for an accepted reason and the rolling median baseline in FastVideo/performance-tracking must be advanced, or when a new v2 exact comparable identity needs its first approved baseline. The workflow backs up existing history under /tmp, validates all source JSONs for the same legacy (model_id, gpu_type) target or the same v2 exact identity, rejects internally inconsistent source batches, uploads one success=true baseline record per accepted source JSON, and offers to clean local temp state after a successful upload.