yzj623/pplab/atomic-loss-training

Claude Code and agent skill by yzj623/pplab

Trust tier 3
  • community
Source check pending

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

Build and optimize the finite-atom APT contrastive loss with proposal posterior correction.

Install

npx vskill@latest install yzj623/pplab/atomic-loss-training

SKILL.md

The same SKILL.md was published earlier as yzj623/pplab/atomic-loss-training.

Atomic Loss Training

Use this skill to implement the APT finite-atom objective from Section 3.2. It is appropriate for reduced recovery, ablations, and deterministic tests where the posterior estimator is represented by a simple differentiable score function.

Inputs

  • Atom parameter values and the index of the positive simulated parameter.
  • Observation value and affine score parameters.
  • Prior and proposal log-density values for the atoms.
  • Learning rate for one optimizer step.

Outputs

  • Cross-entropy loss before and after update.
  • Probability assigned to the positive atom.
  • Parameter values before and after the optimizer step.

Workflow

  1. Score each atom with an observation-conditioned affine model.
  2. Apply the proposal-posterior transform before computing probabilities.
  3. Compute cross-entropy against the positive atom.
  4. Compute analytic gradients and update trainable parameters.
  5. Record a trace compatible with recovery validation.

Validation

Run python scripts/atomic_loss.py --self-test. The deterministic test verifies that the optimizer changes parameters and reduces loss.

Limitations

The provided script is a compact recovery implementation. Full APT normally uses neural density estimators such as mixtures or flows.

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
5

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/yzj623/pplab/atomic-loss-training/badge)](https://verified-skill.com/skills/yzj623/pplab/atomic-loss-training)

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