Managed workflows

Managed workflows run your own Python code on managed classical compute that IBM Quantum® Platform provides. Use a managed workflow when no pre-built Catalog Function covers your task, or when you want to package your own method as a reusable function.

What Functions manages for you

You provide the code. IBM Quantum Functions manages the rest of the run, so you do not provision machines or build an environment yourself:

• Compute. Functions allocates the CPU, memory, and GPUs that you choose with a compute profile.
• Environment. Functions installs and pre-compiles all of the workflow's packages before the job starts.
• Storage. Functions sets up storage for the files that the workflow reads and writes.

What you can do

• Run long jobs in the cloud. A managed workflow keeps running when you close your laptop, and you can reconnect to the job later.
• Use large-memory and GPU nodes. Choose a compute profile that matches the memory, CPU, and GPU needs of the workload.
• Run tasks in parallel. Split classical pre-processing and post-processing across workers.
• Share a reusable program. Upload the workflow once and run it with different inputs.

Compute profiles

Each workflow runs with one compute profile. A profile name lists the CPU cores, the memory in GB, and, when present, the GPUs. For example, 24x120x1l40s is 24 CPU cores, 120 GB of memory, and one NVIDIA L40S GPU.

A workflow can define several sizes and a default size. If the workflow defines no sizes, it uses a default size. When you run a workflow, pass function_size to choose a size, or omit it to use the default.

Next steps