How It Works

Overview

Autoresearch operates in three phases:

1. Initialization – Instrument the pipeline with metrics logging, establish a baseline, and measure headspace.

2. Seed experiments – Launch structural optimizations that are known to be high-impact (baseline measurement, headspace analysis, subprocess pipeline).

3. Iterative optimization – The coding agent analyzes results, proposes follow-up experiments, and the engine executes them concurrently until metrics plateau.

The system is built as two tiers: an interactive supervisor agent that gathers configuration and monitors progress, and a non-interactive async engine that runs the experiment loop. The supervisor starts the engine in the background and can inspect workdir state while experiments run.

The following screenshot shows the supervisor agent reporting on an active autoresearch run — experiment table, running jobs, findings, and diagnostics are all visible in one status query:

Getting Started

The recommended entry point is cli.py, which launches an interactive supervisor agent that gathers missing configuration, starts the engine, and monitors progress:

python spdl/tools/autoresearch/cli.py <workdir> \
  --pipeline-script <path/to/pipeline.py> \
  --source-dir <path/to/source/> \
  --build-command "<build command>" \
  --base-launch-command "<launch command with \$IMAGE>"

The supervisor asks for any missing values interactively. Once configuration is complete, it launches the engine in the background and reports progress as experiments complete.

To run the engine directly without a supervisor (non-interactive mode):

python run.py <workdir> \
  --pipeline-script <path/to/pipeline.py> \
  --source-dir <path/to/source/> \
  --build-command "<build command>" \
  --base-launch-command "<launch command with \$IMAGE>" \
  --max-iterations 10 \
  --patience 3 \
  --max-concurrency 3 \
  --job-timeout 1800

The engine requires the following inputs:

• A pipeline script – the Python file containing the SPDL pipeline to optimize.

• A source directory – the directory containing the pipeline code. The engine modifies files in this directory during experiments.

• A build command – how to build the job image (e.g., docker build).

• A launch command template – the command to launch a training job. Use $IMAGE as a placeholder for the image name.

Configuration is persisted to <workdir>/config.json on the first run. To resume after an interruption, simply re-run with the workdir alone:

python spdl/tools/autoresearch/cli.py <workdir>

# or directly:
python spdl/tools/autoresearch/run.py <workdir>

How It Works

Instrumentation

On the first run, the engine automatically instruments the pipeline script with TTFB (time to first batch) and per-step timing. The pipeline source is sent to the coding agent with instructions to add lightweight logging that records when each training step starts and ends.

The instrumented code is committed to source control (Sapling or Git), creating a clean baseline for subsequent experiments to branch from.

Seed Experiments

The engine schedules three seed experiments before entering the iterative loop. Each addresses a known high-impact area and forms a root node in the hypothesis tree.

Baseline

The unmodified pipeline is run to establish baseline metrics: step time, GPU SM utilization, data readiness, and throughput. All subsequent experiments are compared against this baseline.

Headspace analysis

As described in Headspace Analysis, the pipeline is wrapped with CacheDataLoader to measure the upper bound of improvement achievable by optimizing data loading. If the headspace is near zero, the bottleneck is model compute, not data loading. The engine uses this information to decide which optimizations to prioritize.

Subprocess pipeline (MTP)

The pipeline is moved to a subprocess to eliminate GIL contention between the data loading threads and the training loop. This is often a high-impact optimization, as discussed in resolution. By running the pipeline in a separate process, the data loading threads no longer compete with PyTorch for the GIL.

Iterative Optimization

After the seed experiments, autoresearch enters an iterative loop:

1. Analyze – When a job completes, the workflow collects system metrics (GPU SM utilization, CPU utilization) and SPDL pipeline statistics (per-stage execution time, queue occupancy, throughput). These are sent to the coding agent, which produces a structured analysis identifying the bottleneck and evaluating the experiment’s hypothesis.

2. Plan – The coding agent receives the full experiment history, the current best metrics, and the pipeline source code. Based on this context, it proposes follow-up experiments. Each proposal includes a hypothesis, the specific changes to make, and whether the image needs rebuilding.

3. Execute – The workflow applies code changes (if any), builds the image, and launches jobs. Up to max_concurrency jobs run simultaneously. Each job is monitored for completion and timeout.

4. Repeat – The loop continues until the stopping conditions are met: metrics have not improved for patience consecutive planning sessions and all known best practices have been tried.

sequenceDiagram participant Engine as AsyncWorkEngine participant Adapter as AutoresearchAdapter participant Agent as Coding Agent participant Platform as Platform Engine->>Adapter: Start WorkSpec coroutine Adapter->>Platform: Launch job Platform-->>Adapter: Job completed Adapter->>Adapter: Collect metrics Adapter->>Agent: Analyze (metrics + pipeline code) Agent-->>Adapter: Bottleneck analysis + structured metrics Adapter->>Agent: Plan (history + analysis + code) Agent-->>Adapter: Experiment proposals Adapter-->>Engine: Return child WorkSpecs Engine->>Adapter: Start next WorkSpec coroutines

Monitoring

All experiment state lives in the workdir. The following files are useful for monitoring progress.

engine/engine_state.json

Engine status (running, interrupted, or stopped) and experiment counts (queued, running, completed, failed).

engine/checkpoint.json

Runner checkpoint containing the serialized queued and running _WorkSpec objects. This is the source of truth for resume.

engine/queue.json

Compatibility view of pending experiments in priority order.

engine/active.json

Compatibility view of currently running remote jobs.

summary.md

Human-readable progress summary updated after each job completion.

master_table.tsv

Tab-separated table of all experiments with key metrics: step time, SM utilization, data readiness, and duration.

progress.png

Scatter plot showing job duration and SM utilization over time. Green dots indicate improvements over the previous best.

hypothesis_tree.png

Tree visualization of the experiment hierarchy. Nodes are color-coded: green for improved, gray for no improvement, red for failed, blue for running, and dashed white for queued.

runs/<run_id>/analysis.md

The coding agent’s detailed analysis for each completed experiment, including per-stage pipeline metrics and bottleneck identification.

Generating a Report

After the engine finishes, generate a final summary report:

python cmd.py report <workdir>

This collects all per-run analyses and the master table, sends them to the coding agent for synthesis, and writes the output to final_report.md.

Stopping and Resuming

To stop autoresearch gracefully, send SIGINT (Ctrl+C) to the process. The engine cancels local coroutines and persists queued and running specs to engine/checkpoint.json with status interrupted. The workflow also keeps the monitoring files under engine/ up to date.

Warning

Do not send SIGKILL (kill -9) to the engine process. This prevents state persistence and you may lose the queue and in-progress analysis.

Running jobs on the cluster are not cancelled when autoresearch stops. They continue independently. When autoresearch resumes from engine/checkpoint.json, the workflow re-checks their status and collects results.

To resume, simply re-run with the workdir:

python spdl/tools/autoresearch/cli.py <workdir>

# or directly:
python spdl/tools/autoresearch/run.py <workdir>

Modifying the Queue

To manually adjust the experiment queue, stop the engine and edit engine/checkpoint.json. The queued list contains serialized _WorkSpec objects; change their priority values or remove specs as needed. Lower values run first. engine/queue.json is a compatibility view for monitoring and should not be treated as the resume source of truth.

{
  "status": "interrupted",
  "queued": [
    {
      "id": "010_compile_fused_bs6",
      "priority": -100,
      "kind": "experiment",
      "payload": {"node": {"node_id": "010_compile_fused_bs6"}}
    }
  ],
  "running": []
}

Workdir Structure

<workdir>/
├── config.json                 # Experiment configuration
├── state.json                  # Persistent state (history, best metrics)
├── master_table.tsv            # All experiments and their metrics
├── summary.md                  # Human-readable progress summary
├── progress.png                # Duration/SM scatter plot
├── hypothesis_tree.png         # Experiment tree visualization
├── engine/
│   ├── checkpoint.json         # Runner checkpoint for resume
│   ├── engine_state.json       # Engine status and counts
│   ├── tree.json               # Full hypothesis tree
│   ├── queue.json              # Pending experiments (compat view)
│   ├── active.json             # Currently running jobs (compat view)
│   └── nodes/
│       └── <node_id>/
│           ├── spec.json       # Experiment specification
│           ├── status.txt      # Current status
│           └── result.json     # Analysis results
├── runs/
│   └── <run_id>/
│       ├── analysis.md         # Coding agent's analysis
│       └── metrics/            # Raw metrics data
└── logs/
    ├── autoresearch.log        # Full execution log
    ├── *_prompt.md             # Prompts sent to the coding agent
    ├── *_output.md             # Coding agent's responses
    └── *_raw.json              # Raw response JSON with cost

Example

For a detailed walkthrough of an autoresearch run on a real pipeline, see Example: Video Classification.