Compute

Compute in amics is a full runtime bundle, not just a raw GPU rental. A launch includes accelerator selection, CPU, memory, local disk, persistent storage, region, logs, access details, and budget guardrails.

Quote Capacity

from amics import Client

amics = Client()

quote = amics.compute.quote(
    hardware="H100",
    gpu_count=1,
    region="us-east-1",
)

Launch A Workspace

session = amics.compute.launch(
    project_id="fine-tune",
    name="notebook",
    hardware="H100",
    ttl_minutes=120,
)

print(session["login_command"])
print(session["jupyter_url"])

Run A Training Job

job = amics.compute.run(
    "reward-tune",
    base_model="DeepSeek V4 Pro",
    hardware="H100",
    gpu_count=1,
)

Useful Methods

• amics.compute.types()
• amics.compute.quote(hardware="H100", gpu_count=1)
• amics.compute.launch(project_id="default", name="notebook")
• amics.compute.list(project_id="default")
• amics.compute.get(project_id, compute_id)
• amics.compute.stop(project_id, compute_id)
• amics.compute.run(name, base_model="DeepSeek V4 Pro")
• amics.compute.jobs()
• amics.compute.stop_job(job_id)