Examples
Examples
Complete, runnable workloads: save the files, run the commands.
Each page is a self-contained project: a README with the exact commands, and
every source file inline. super login suggests five of these at random.
| Example | What it does |
|---|---|
| qwen3.5-9b | Chat with Qwen3.5-9B on a single L4. |
| qwen-image-text-to-image | Generate images from text with Qwen-Image on an H100. |
| trellis2-image-to-3d | Turn one image into a textured 3D mesh with TRELLIS.2. |
| jax-mnist | Train a JAX + Flax CNN on MNIST in a couple of minutes. |
| modded-nanogpt | Run the GPT-2 speedrun on 8×H100. |
| jax-distributed | Multi-process JAX across 8 GPUs with jax.distributed. |
| vla-jepa-libero-finetune | Fine-tune a VLA-JEPA robot policy on LIBERO for about $10. |
| boltzgen | Design a GLP-1 receptor binder with boltzgen. |
| mujoco-mjx-locomotion | Legged-robot rollouts with MuJoCo MJX on the GPU. |
| nerfstudio-splatfacto | Train a 3D Gaussian Splat with nerfstudio on an L4. |
| pufferlib-rl | High-throughput reinforcement learning with PufferLib. |
| map-train-sweep | Sweep a learning rate across GPUs with super map. |
| volume-train-checkpoint | Checkpoint training to a volume that outlives the session. |