smolgym

A session

Start a run.

There is no account. Send the model, whether the loop is supervised or reinforcement, and what the reward checks.

01

What to send

Three things
  1. 01

    The model

    A small model on the rack, named by one string, or the weights you upload.

  2. 02

    The loop

    Supervised or reinforcement. Supervised training skips sample().

  3. 03

    The reward

    What it checks. An exact match, a test, a rubric, a judge, or an environment you run.

02

What you get

If it fits
  1. 01

    A session

    Arranged by email. There is no account to create.

  2. 02

    The machines

    Your script stays on your machine. The GPUs are ours.

  3. 03

    The weights

    You can download them when you want them.

03

What does not fit

One machine
  1. 01

    A cluster

    The floor is one training machine.

  2. 02

    A model that does not fit

    If the weights need more than that machine, the run is not here.

Write with those three.

If the model fits the floor, you get a session: your script, our GPUs, weights you can download.

Start a run
swapnil@rycerz.es