What is run-on-databricks?

A process for submitting a machine-learning training job to a Databricks GPU cluster and bringing its MLflow metrics back to the local project. A GPU cluster is a group of remote computers suited to parallel numerical work, and MLflow records experiment results.

Running a local smoke test, launching full training on Databricks, optionally using a temporary job cluster, retrieving MLflow results, and reporting whether the run succeeded.

It checks the training code before sending it to the remote cluster and then provides status, metrics, runtime, hardware, and comparison with earlier runs.

run-on-databricks — Skill for Claude Code and Codex by duonginspace: A process for submitting a machine-learning training job to…

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/duonginspace/claude-code-databricks-ml/run-on-databricks

Any agent

npx skills add duonginspace/claude-code-databricks-ml --skill run-on-databricks

Clone the repo

git clone --depth 1 https://github.com/duonginspace/claude-code-databricks-ml

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.

README.md

[![agentmods](https://agentmods.dev/badge/skills/duonginspace/claude-code-databricks-ml/run-on-databricks.svg)](https://agentmods.dev/skills/duonginspace/claude-code-databricks-ml/run-on-databricks)

Your own site

<a href="https://agentmods.dev/skills/duonginspace/claude-code-databricks-ml/run-on-databricks"><img src="https://agentmods.dev/badge/skills/duonginspace/claude-code-databricks-ml/run-on-databricks.svg" alt="Measured on agentmods" height="20"></a>

Per session 80 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.

When invoked 608 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. Scan, not verified.

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.

Model	Per session	Once invoked
Fable 5	$0.00080	$0.00608
Opus 5	$0.00040	$0.00304
Sonnet 5	$0.00016	$0.00122
Haiku 4.5	$0.00008	$0.00061

Measured 5d ago against content hash e30678ba33a0, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

run-on-databricks 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 5d 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.

skills/run-on-databricks/SKILL.md · 43 lines

What it actually says

Databricks training task

Script and args: $ARGUMENTS

Steps

Read CLAUDE.md for cluster ID, experiment name, and environment setup.
Run a local smoke test first if the user hasn't confirmed the code runs locally: uv run python scripts/train.py --datasets 1 --epochs 2 2>&1 | head -40 If it fails, stop and report the error. Don't submit broken code to the cluster.
Submit to Databricks (this builds a wheel, uploads it + scripts/train.py to DBFS, and runs): uv run python scripts/submit_to_databricks.py scripts/train.py $ARGUMENTS The script passes --wheel-path and --experiment args automatically.
- To use an ephemeral job cluster (lower DBU rate) instead of the existing cluster, add --job-cluster
- Job clusters have ~5-10min startup overhead but cost less per DBU
Pull MLflow results: uv run python scripts/pull_results_on_databricks.py
Read mlflow_results/latest_run.json and mlflow_results/all_runs.csv.
Report:
- Run status (SUCCESS / FAILED)
- Key metrics (loss, accuracy, or whatever the primary metric is)
- Training time and GPU type used
- Comparison with the previous best run from all_runs.csv
If the run failed, read mlflow_results/job_logs.txt and diagnose the error. Common DBR 15+ issues:
- OSError: Operation not supported on Workspace paths — scripts must be on DBFS
- pydantic has no model_validator — stale system pydantic; check that the bootstrap clears sys.modules
- cannot import Sentinel from typing_extensions — same stale module issue
- BAD_REQUEST: For input string: "None" — MLflow experiment name must be a /Users/... workspace path
- DBFS library installations are not supported on DBR 15 — wheel must be pip-installed at runtime, not via compute.Library

Output

Summarize results and propose the next experiment with specific hyperparameter or architecture changes. ultrathink

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.

5d ago First seen · 43 lines · 80 tokens per session scan A e30678ba33a0

Subscribe to this mod's changes

run-on-databricks is a skill published in the GitHub repository duonginspace/claude-code-databricks-ml (5 stars, last pushed 5mo ago), licensed MIT. It adds 80 tokens to every session and 608 once invoked, about $0.0004 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 2026-08-31.

Data and AI

agent-platform-rag-engine-management

Manage and query Agent Platform RAG Engine Corpora and retrieve grounded contexts using the Google GenAI SDK. Use when listing RAG corpora or files, inspecting a corpus, retrieving contexts, or generating content grounded in a RAG corpus. Do not use for standard database queries (use SQL/Spanner skills), Google…

google/skills · 85 tokens

agent-platform-model-registry

Agent Platform Model Registry Management. Use when you need to upload, list, describe, update, or delete machine learning models (and their versions) in the Agent Platform Model Registry. Don't use for model training, model deployment to endpoints, or managing non-Agent Platform models.

google/skills · 60 tokens

foundry-config-setup

Resolve missing setup caused by a hardcoded Foundry project endpoint or model in a sample. Use when a sample fails because it uses a placeholder/hardcoded projectendpoint (for example "https://your-project.services.ai.azure.com") or a hardcoded model instead of reading them from the environment.

microsoft/agent-framework · 65 tokens

google-cloud-solution-agentic-analytics-spark-knowledge-catalog

Discovers requirements and generates guidance to design and deploy a governed, secure agentic-analytics solution for data that's distributed across Google Cloud, other cloud providers, or on-premises. Data that's outside Google Cloud (such as data from Databricks, Snowflake, Salesforce, SAP, or Oracle systems) is…

google/skills · 138 tokens

training-check

Interactively monitor training metrics from the current Codex session, periodically checking WandB or fallback logs for NaN, divergence, plateaus, and broken runs.

wanshuiyin/Auto-claude-code-research-in-sleep · 35 tokens

nemo-automodel-launcher-config

Configure NeMo AutoModel job launches for interactive runs, Slurm clusters, and SkyPilot cloud execution.

NVIDIA/skills · 30 tokens