All commands run from your regular terminal. There is no REPL — tccp gives you a persistent shell on the compute node instead.
| command | what it does |
|---|---|
| tccp start | Full startup: allocate a SLURM node, pull container, sync files, run init, start persistent shell via dtach. Requires tccp.yaml in the current directory. |
| tccp shell | Attach to the persistent shell. You're inside the container on the compute node with GPU access. Ctrl+S detaches, syncs, and reattaches. Ctrl+D exits. |
| tccp stop | Pull output, cancel the SLURM job, clear session state. |
$ tccp start $ tccp shell tccp> python train.py tccp> exit $ tccp stop
| command | what it does |
|---|---|
| tccp sync | Push changed files to compute node + pull output back to laptop. Also happens automatically on Ctrl+S during tccp shell. |
| tccp exec <cmd> | Run a one-off command inside the container on the compute node and print the result. Useful for quick checks without attaching to the shell. |
$ tccp sync $ tccp exec "python -c 'import torch; print(torch.cuda.is_available())'"
| command | what it does |
|---|---|
| tccp status | Show session info: job ID, compute node, scratch path, container, start time, ports. |
| tccp allocs | List your SLURM allocations with job ID, name, partition, GPU, time, and state. |
| command | what it does |
|---|---|
| tccp gpus | Live GPU availability across all partitions. Shows available vs total count per GPU type. |
| tccp gpus <type> | Per-node breakdown for a specific GPU type. e.g. tccp gpus a100 |
| tccp gpus info | Static GPU guide with VRAM, use cases, and recommendations. See also GPUs page. |
| command | what it does |
|---|---|
| tccp dealloc <id> | Cancel a SLURM allocation by job ID. |
| tccp dealloc all | Cancel all your SLURM allocations. |
| tccp setup | Save or update your cluster credentials (username and password). |
v