api v1alpha1 #17
api v1alpha1 #17
Conversation
|
@dolfim-ibm Let's prioritize this PR asap! |
dolfim-ibm
force-pushed
the
api-upgrade
branch
from
d64cc33 to
409541d
Compare
Signed-off-by: Guillaume Moutier <[email protected]> Signed-off-by: Michele Dolfi <[email protected]>
Signed-off-by: Michele Dolfi <[email protected]>
Signed-off-by: Michele Dolfi <[email protected]>
Github Action runners are running out of the space while building both the images in parallel. This change will build the image sequentially and also clean up the cpu images before start building gpu image. Signed-off-by: Anil Vishnoi <[email protected]> Signed-off-by: Michele Dolfi <[email protected]>
) Signed-off-by: Anil Vishnoi <[email protected]> Signed-off-by: Michele Dolfi <[email protected]>
dolfim-ibm
force-pushed
the
api-upgrade
branch
from
409541d to
930d3fd
Compare
|
@guimou I started with a few changes
Next, we will look
|
|
Just had a look at the changes, that's great. Thanks a lot!
…On Fri., Jan. 24, 2025, 08:19 Michele Dolfi, ***@***.***> wrote:
@guimou <https://github.com/guimou> I started with a few changes
- used the Annotated syntax for the parameters
- use default values and not Optional (which in Python means "can be
None" and has almost no relation with the concept of optional)
- refactor the request arguments logic. where possible use the native
enum.
- make the gradio part optional (but highly advised)
- rebase with main such that tests build the image
Next, we will look
- the background processing
- input/output schemas
—
Reply to this email directly, view it on GitHub
<#17 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AA6C4YQ7K4TIVK7KTR4SPL32MI4WZAVCNFSM6AAAAABU3IUWYSVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMMJSGUYTINRZGQ>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
Signed-off-by: Michele Dolfi <[email protected]>
Signed-off-by: Michele Dolfi <[email protected]>
Signed-off-by: Michele Dolfi <[email protected]>
Signed-off-by: Michele Dolfi <[email protected]>
Signed-off-by: Michele Dolfi <[email protected]>
Signed-off-by: Michele Dolfi <[email protected]>
Signed-off-by: Michele Dolfi <[email protected]>
Signed-off-by: Michele Dolfi <[email protected]>
|
@guimou I reworked a few parts, I think that now the PR is ready for merge and release. @vishnoianil do you want to also have a quick look? |
|
|
||
| EXPOSE 5000 |
There was a problem hiding this comment.
is there any specific need to run this service on port 8080? I am wondering If we can revert it 5000 or some other non-80** port? @dolfim-ibm
There was a problem hiding this comment.
It was to run on the same "standard" port as other serving runtimes in OpenShift AI. But it's an env var, so can be anything in the end.
|
Awesome work @guimou @dolfim-ibm minor comment, PR looks good to me. I am wondering can we enable the arm64 arc for "cpu only" image ? It's enabled for GPU image, but not for CPU only image. I remember we were discussing about some dependencies not available for arm64, and that was causing issue in building arm64 image, is that still true? |
Signed-off-by: Michele Dolfi <[email protected]>
Signed-off-by: Michele Dolfi <[email protected]>
Signed-off-by: Michele Dolfi <[email protected]>
Signed-off-by: Michele Dolfi <[email protected]>
|
I realized we missed addressing the input http with custom headers. I'm making a pushing a version similar to what we have on origin/main. |
Signed-off-by: Michele Dolfi <[email protected]>
Signed-off-by: Michele Dolfi <[email protected]>
Signed-off-by: Michele Dolfi <[email protected]>
Signed-off-by: Michele Dolfi <[email protected]>
Signed-off-by: Michele Dolfi <[email protected]>
|
@vishnoianil The PORT has been changed to 5001. Additionally I also reworked the input payload to match the format already in |
Rework for API version v1alpha1 as discussed.