Using the platform

Since 2.0, quantization runs on the R3AL platform at platform.r3al.ai. The SDK is the thin client that uploads your model, submits the job, and downloads the result. This page is the bridge between the two.

Create an account and an API key

Create an account at platform.r3al.ai. Verify your email to activate the account.

Mint an API key

Open the SDK / API keys page and create a key. Live keys look like r3l_live_.... The secret is shown once, so copy it immediately. If you lose it, revoke it and mint a new one.

Connect the SDK

The SDK resolves your key from, in order: the argument you pass, platform.login(...), then the R3AL_API_KEY environment variable.

import r3alai.platform as platform

platform.login("r3l_live_...")

export R3AL_API_KEY="r3l_live_..."

To point at a non-default platform host (for example a staging URL), use configure:

import r3alai.platform as platform

platform.configure(base_url="https://platform.r3al.ai", api_key="r3l_live_...")

Monitor jobs

Every quantize() or train_qat() call creates a job on the platform. The SDK prints live progress while it waits:

[r3al] job 1a2b3c4d · running · 62% · calibrating · 00:48 elapsed

Open the platform’s Jobs page to watch the same job in real time: its status (pending, running, completed, error), a progress bar, and a history of every run you have submitted. From the SDK you can inspect a job directly:

from r3alai.platform import PlatformClient

client = PlatformClient()
job = client.get_job("job_...")
print(job.status, job.progress)

Your plan and the free-run limit

The free plan includes 3 runs. Each quantize or QAT job consumes one run; local benchmark(), load(), and export do not. Check what you have left from the SDK or the platform’s usage/plan page:

from r3alai.platform import PlatformClient

print(PlatformClient().usage())   # plan + remaining free runs

A run beyond your quota raises PlanLimitError (HTTP 402):

from r3alai.platform import PlanLimitError
from r3alai.quant import Quantizer

try:
    Quantizer().quantize(
        "model.onnx",
        calibration_data=["img1.jpg", "img2.jpg", "img3.jpg"],
        output_dir="./out",
    )
except PlanLimitError:
    print("Free plan limit reached. Upgrade at platform.r3al.ai.")

Upgrade your plan on the platform to lift the limit.

Download results

quantize() and train_qat() download the deliverable automatically into output_dir, and result.path points at it. If you submit a job with the lower-level PlatformClient and want to download later, use the Job handle:

from r3alai.platform import PlatformClient

client = PlatformClient()
job = client.quantize(
    "model.onnx",
    calibration_data=["img1.jpg", "img2.jpg", "img3.jpg"],  # required for the default ptq_static
)                                     # uploads + submits, returns a Job
job.wait()                            # blocks with live progress
bundle_dir = job.download("./out")    # downloads + unpacks the bundle

The unpacked bundle contains the quantized model and r3alai_manifest.json, so local load() and benchmark() work on it directly.

Errors

All platform errors subclass R3ALPlatformError:

Error	When
`AuthenticationError`	Missing or invalid API key (HTTP 401)
`PlanLimitError`	Free-run quota reached (HTTP 402)
`UploadError`	A model or calibration upload failed
`JobFailedError`	The job reached a terminal error state

Get Started

Core Concepts

Guides

Using the platform

Create an account and an API key

Connect the SDK

Monitor jobs

Your plan and the free-run limit

Download results

Errors

Next steps

End-to-end workflow

Python SDK

​Create an account and an API key

​Connect the SDK

​Monitor jobs

​Your plan and the free-run limit

​Download results

​Errors

​Next steps

End-to-end workflow

Python SDK

Create an account and an API key

Connect the SDK

Monitor jobs

Your plan and the free-run limit

Download results

Errors

Next steps