vectorspacelab/arex-skill/aevb-vae-core

Claude Code and agent skill by vectorspacelab/arex-skill

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Run a compact AEVB/VAE training step with encoder statistics, reparameterization, decoder reconstruction, analytic KL, and optimizer evidence.

Install

npx vskill@latest install vectorspacelab/arex-skill/aevb-vae-core

SKILL.md

The same SKILL.md was published earlier as vectorspacelab/arex-skill/aevb-vae-core.

AEVB VAE Core

Use this skill when a recovery experiment must exercise the core Auto-Encoding Variational Bayes mechanism on image-like data. It is appropriate for reduced/proxy recovery when full dataset training is blocked, provided the proxy is declared.

Inputs

  • Proxy image batch JSON with values in [0, 1].
  • latent_dim, learning_rate, and seed.
  • Optional output path for a training trace.

Outputs

  • Training trace JSON with reconstruction loss, KL, total loss before and after one optimizer step; positive loss_delta is desirable recovery evidence but callers should inspect the numeric trace because tiny stochastic steps can vary by seed and learning rate.
  • params_before and params_after snapshots.
  • Mechanism checks: encoder, reparameterization, decoder, reconstruction loss, KL, optimizer step.

Workflow

  1. Load image observations and flatten them.
  2. Encode to mu and logvar.
  3. Compute z = mu + exp(0.5 * logvar) * epsilon.
  4. Decode z to Bernoulli reconstruction logits.
  5. Compute BCE plus analytic Gaussian KL.
  6. Update trainable parameters and record before/after signals.

Validation

Run:

bash
python scripts/run_vae_step.py --batch-json /tmp/vae_proxy_batch.json --output /tmp/vae_trace.json
python tests/test_vae_step.py

Limitations

A tiny proxy trace is not a full paper reproduction. Keep full_training_executed false unless a real full dataset/model run was executed.

Security

  • Security scanPassed, 90 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
0

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/vectorspacelab/arex-skill/aevb-vae-core/badge)](https://verified-skill.com/skills/vectorspacelab/arex-skill/aevb-vae-core)

Something wrong with this skill or its page? Tell us.