generic --cloud-type not working

[casperdcl]

cml runner ... --cloud=aws --cloud-type=m+t4 doesn't work:

{"level":"error",
 "message":"terraform error: Error: Failed creating the machine: Not able to decode: operation error EC2: RunInstances,
https response error StatusCode: 400,
RequestID: 622a6bc7-75c5-47d1-83ae-76f8f9029186,
api error InvalidParameterValue: Invalid value 'm+t4' for InstanceType."}

meanwhile cml runner ... --cloud=aws --cloud-type=m --cloud-gpu=t4:

{"level":"error",
 "message":"terraform error: Error: Failed creating the machine: Not able to decode: operation error EC2: RunInstances,
https response error StatusCode: 400,
RequestID: dcc5278c-7eda-467a-b142-20c4d3de10f1,
api error InvalidParameterValue: Invalid value 'm' for InstanceType."}

does CML not work with TPI generic-machine-types?

[dacbd]

yes, and no.

Something does need to be improved as what exists is clearly not clear 🙃

I was previously in favor of splitting out these mappings to something more consolidated but unfortunately, that was cut and these "helper types" are all completely separate in their respective providers. So something like m+t4 which may work on one does not work on others.

for example, m+t4 is missing from all the providers see: AWS and azure:

additionally the --cloud-gpu is essentially only functional with GCP in a way that I would expect since azure and AWS have separate machine types with predetermined GPUs but GCP allows you to more explicitly set something, see: #1049

[casperdcl]

...and iterative/cml.dev#251

[dacbd]

@casperdcl with iterative/terraform-provider-iterative#718 the next tpi release with resolve the error you found here, with iterative/cml.dev#251 as its own issue that should still be addressed I think we can close this one, but I'll leave that to you.

[casperdcl]

Thanks; happy to close here :)