training_hub: Skill for Claude Code

.claude/skills/memory-estimation/SKILL.md

memory-estimation is a skill for Claude Code from Red-Hat-AI-Innovation-Team/training_hub. It costs 40 tokens per session (413 once invoked), scanned A, original, Apache-2.0.

A tool for estimating how much GPU memory, also called VRAM, a language-model training setup will need.

In plain words
What is it for?
Use it to estimate memory for SFT, GRPO, OSFT, LoRA, or QLoRA configurations and assess options such as shorter sequences, smaller batches, LoRA, or additional GPUs.
Why use it?
It lets you check whether a model and configuration will fit on the available GPUs before committing to a training run.

Skill for Claude Code

Written for Claude Code: allowed-tools in frontmatter.

This is Red-Hat-AI-Innovation-Team/training_hub's own configuration. It tells Claude Code how to work on training_hub itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything training_hub configures →

Runs only inside its plugin — its command needs a path that Claude Code sets for a plugin’s own hooks and for nothing else. Install the plugin, not this.

Part of the training-hub plugin — 4 skills shipped together

Reuse

Borrowing it

Nothing to install: this file belongs to Red-Hat-AI-Innovation-Team/training_hub. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/Red-Hat-AI-Innovation-Team/training_hub/main/.claude/skills/memory-estimation/SKILL.md
Clone the repo
git clone --depth 1 https://github.com/Red-Hat-AI-Innovation-Team/training_hub

Made for: Claude Code.

Or install training-hub, the plugin that ships this one along with the rest of its 4 skills.

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 memory-estimation

README.md
[![agentmods](https://agentmods.dev/badge/skills/red-hat-ai-innovation-team/training_hub/memory-estimation/github.svg)](https://agentmods.dev/skills/red-hat-ai-innovation-team/training_hub/memory-estimation)
Your own site
<a href="https://agentmods.dev/skills/red-hat-ai-innovation-team/training_hub/memory-estimation"><img src="https://agentmods.dev/badge/skills/red-hat-ai-innovation-team/training_hub/memory-estimation/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for memory-estimation

Your own site · 80×15
<a href="https://agentmods.dev/skills/red-hat-ai-innovation-team/training_hub/memory-estimation"><img src="https://agentmods.dev/badge/skills/red-hat-ai-innovation-team/training_hub/memory-estimation.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 40 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 413 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original 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 5.1 $0.00040 $0.00413
Opus 5 $0.00020 $0.00206
Sonnet 5 $0.00008 $0.00083
Haiku 4.5 $0.00004 $0.00041

Measured 10d ago against content hash c817ec7dbef1, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-09, from the pricing page.

Security

Grade A, and why

memory-estimation 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 10d ago.

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.

.claude/skills/memory-estimation/SKILL.md · 49 lines

What it actually says

GPU Memory Estimation

Estimate GPU VRAM requirements before committing to a training run.

Step 1: Check Environment

"${CLAUDE_PLUGIN_ROOT}/scripts/th_detect.sh"

If library=missing, tell the user to install training_hub first via the setup-guide skill.

Step 2: Run Estimation

Execute the estimation script with user-provided parameters or config defaults:

"${CLAUDE_PLUGIN_ROOT}/scripts/th_estimate.sh" $ARGUMENTS

Step 3: Present Results

Parse the JSON output and present clearly:

  1. Memory estimates — Show low/mid/high VRAM estimates in GB
  2. GPU fit — Report whether the configuration fits on the available GPU(s)
  3. Recommendations — If memory is tight, suggest:
    • Reduce max_seq_len (e.g., 4096 -> 2048)
    • Reduce effective_batch_size
    • Switch to LoRA or QLoRA for lower memory
    • Add more GPUs for data parallelism

Estimation Methods

Method For Estimator
basic SFT, GRPO BasicEstimator
osft OSFT OSFTEstimator
lora LoRA-SFT, LoRA-GRPO LoRAEstimator
qlora Quantized LoRA QLoRAEstimator

If no method is specified, the script infers it from the configured algorithm.

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. 10d ago First seen · 49 lines · 40 tokens per session scan A c817ec7dbef1

Subscribe to this mod's changes

memory-estimation is a skill published in the GitHub repository Red-Hat-AI-Innovation-Team/training_hub (95 stars, last pushed today), licensed Apache-2.0. It adds 40 tokens to every session and 413 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 2026-08-30.

Related

Other skills, from other repositories

ml-training-recipes

Battle-tested PyTorch training recipes for all domains — LLMs, vision, diffusion, medical imaging, protein/drug discovery, spatial omics, genomics. Covers training loops, optimizer selection (AdamW, Muon), LR scheduling, mixed precision, debugging, and systematic experimentation. Use when training or fine-tuning…

Orchestra-Research/AI-Research-SKILLs · 88 tokens

tao-run-automl

Run container-backed AutoML / hyperparameter optimization (HPO) for NVIDIA TAO networks using AutoMLRunner. Handles algorithm selection (bayesian, hyperband, asha, bohb, llm, hybrid, autoresearch), WandB experiment tracking, job execution on any TAO SDK platform, result interpretation, and per-rec custom evaluation…

NVIDIA-TAO/tao-skill-bank · 235 tokens

unsloth

Expert guidance for fast fine-tuning with Unsloth - 2-5x faster training, 50-80% less memory, LoRA/QLoRA optimization.

davila7/claude-code-templates · 39 tokens

unsloth

Expert guidance for fast fine-tuning with Unsloth - 2-5x faster training, 50-80% less memory, LoRA/QLoRA optimization.

OpenLAIR/dr-claw · 39 tokens

hugging-face-model-trainer

This skill should be used when users want to train or fine-tune language models using TRL (Transformer Reinforcement Learning) on Hugging Face Jobs infrastructure. Covers SFT, DPO, GRPO and reward modeling training methods, plus GGUF conversion for local deployment. Includes guidance on the TRL Jobs package, UV…

synthetic-sciences/openscience · 131 tokens

Training

Guides fine-tuning and post-training work with explicit data, objective, hardware, and rollback assumptions.

agentic-in/elephant-agent · 23 tokens