ml-mlip-automl

ml-mlip-automl is a skill for Claude Code, Codex from learningmatter-mit/AtomisticSkills. It costs 34 tokens per session (1,000 once invoked), scanned A, original, MIT.

An automated search process for choosing training settings for machine-learning models that predict atomic interactions. It tests settings such as learning rate, batch size, and whether to freeze part of a model.

In plain words
What is it for?
Use it to tune MACE, MatGL, or FairChem models for a particular chemical system or prediction task.
Why use it?
It reduces manual trial and error when adapting a model to a new dataset. The search uses validation error during training to guide the next experiment.

Skill for Claude CodeCodex

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/learningmatter-mit/atomisticskills/ml-mlip-automl
Any agent
npx skills add learningmatter-mit/AtomisticSkills --skill ml-mlip-automl
Clone the repo
git clone --depth 1 https://github.com/learningmatter-mit/AtomisticSkills

Made for: Claude Code, Codex.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for ml-mlip-automl

README.md
[![agentmods](https://agentmods.dev/badge/skills/learningmatter-mit/atomisticskills/ml-mlip-automl.svg)](https://agentmods.dev/skills/learningmatter-mit/atomisticskills/ml-mlip-automl)
Your own site
<a href="https://agentmods.dev/skills/learningmatter-mit/atomisticskills/ml-mlip-automl"><img src="https://agentmods.dev/badge/skills/learningmatter-mit/atomisticskills/ml-mlip-automl.svg" alt="Measured on agentmods" height="20"></a>
Per session 34 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,000 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin unknown No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00034 $0.01000
Opus 5 $0.00017 $0.00500
Sonnet 5 $0.00007 $0.00200
Haiku 4.5 $0.00003 $0.00100

Measured today against content hash 7734439bab62, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

ml-mlip-automl scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured today.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.agents/skills/ml-mlip-automl/SKILL.md · 58 lines

How it starts

The opening of the file, as written. The whole thing — 58 lines — stays where its author put it; the contents beside it link to each section on GitHub.

LLM-Coupled MLIP Hyperparameter Search (AutoML)

Goal

To automate the discovery of optimal training hyperparameters (learning rate, batch size, backbone freezing) by coupling the agent to an iterative execution loop. Instead of conducting exhaustive grid searches, the LLM actively interprets validation MAE trajectories per epoch and strategically deduces the next optimal parameter permutation to evaluate.

The Agentic Workflow

Instead of calling a monolithic training tool, this SKILL dictates how you (the Agent) should orchestrate a search using the individual foundation fine-tuning SKILLs (ml-mace-finetune, ml-matgl-finetune, ml-fairchem-finetune).

Execution Protocol

  1. Initial Anchor: Select a conservative parameter set based on the dataset size (e.g., small dataset: freeze backbone, lower LR).
  2. Execute Fine-Tuning: Create a dedicated subdirectory for the run (e.g., search_run_lr1e-3_frozen) and execute the respective MLIP's prepare_*_data.py and training runner.
  3. Parse and Evaluate: Read the resulting training_history.json directly into your context. Analyze the validation loss and MAE curves (slope, signs of overfitting/forgetting).
  4. Iterate: Deduce the next logical parameter adjustment based on the evaluation (e.g. "Validation MAE plateaued early. Action: Unfreeze backbone & lower LR to 1e-4").
  5. Terminate: Halt and report the best configuration once a performance target is met or degradation occurs repeatedly.

Hyperparameter Search Spaces

Different MLIP frameworks expose different tunable parameters. When deducing your next move, restrict your search space to the available features:

Feature ml-fairchem-finetune ml-matgl-finetune ml-mace-finetune
Freeze backbone --freeze-backbone --freeze-backbone --freeze-backbone
Re-init head N/A (Auto) --reinit-head --reinit-head
Learning Rate --lr (Default: 4e-4) --lr (Default: 1e-3) --lr (Default: 1e-4)
Batch Size --batch-size --batch-size --batch-size
Scheduler Cosine with Warmup Cosine or Plateau Exponential or Plateau
Force Loss Ratio --force-weight --force-weight --forces-weight
Energy Loss Ratio --energy-weight --energy-weight --energy-weight
Loss Criterion Huber/L1/L2 Huber/L1/L2 universal/weighted

Read the full file on GitHub · 58 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. today First seen · 58 lines · 34 tokens per session scan A 7734439bab62

Subscribe to this mod's changes

ml-mlip-automl is a skill published in the GitHub repository learningmatter-mit/AtomisticSkills (158 stars, last pushed yesterday), licensed MIT. It adds 34 tokens to every session and 1,000 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.

Related

Other skills, from other repositories

hugging-face-trackio

Track and visualize ML training experiments with Trackio. Use when logging metrics during training (Python API) or retrieving/analyzing logged metrics (CLI). Supports real-time dashboard visualization, HF Space syncing, and JSON output for automation.

patchy631/ai-engineering-hub · 52 tokens

pysr

Use when fitting equations to data with PySR or SymbolicRegression.jl, when a user wants an interpretable formula, symbolic model, scaling law, or empirical relation discovered from numeric data, or when debugging a PySR search that is slow, stuck, or giving poor equations.

astroautomata/PySR · 61 tokens

cellxgene-census-query

Query CZ CELLxGENE Census (61M+ cells). Filter by cell type/tissue/disease, retrieve expression data, and integrate with scanpy/PyTorch for population-scale single-cell analysis. Use this skill when: (1) Querying single-cell expression data by cell type, tissue, or disease, (2) Exploring available single-cell datasets…

PharMolix/OpenBioMed · 105 tokens

HomeSafe-Bench

VLM indoor safety hazard detection benchmark inspired by HomeSafeBench (arXiv 2509.23690).

SharpAI/DeepCamera · 28 tokens

depth-estimation

Real-time depth map privacy transforms using Depth Anything v2 (CoreML + PyTorch).

SharpAI/DeepCamera · 22 tokens

start-experiment

Start the autoresearch optimization loop for a specific model + lane. Resolves the hierarchical program.md (root → model → lane), asks the user for hardware (local TPU VM or GKE cluster of a specified TPU type + topology), discovers available clusters from .env/, checks occupancy with USERPREFIX-aware attribution…

vlasenkoalexey/tpu_performance_autoresearch_wiki · 103 tokens