#3771 [Core feature] Imagespec to support GPU dependencies Issue created by zeryx Motivation: Why do you think this is important? Image spec does not support adding GPU dependencies. It's possible with envd to support cuda https://envd.tensorchord.ai/api/starlark/v1/install.html#cuda Goal: What should the final outcome look like, ideally? I should be able to define an imagespec file similar to the following: default = ImageSpec( name="gpu", base_image="ghcr.io/flyteorg/flytekit:py3.8-1.6.0", registry="ghcr.io/some_user", packages=["flytekit>=1.6.0", "bayesian-optimization==1.4.3"], gpu=["cuda:11.04-cudnn7"], python_version="3.9" ) Describe alternatives you've considered An alternative is to build a base image that contains the particular CUDA / ROCm / OpenCL dependency - however that requires building and maintaining Propose: Link/Inline OR Additional context No response Are you sure this issue hasn't been raised already? ☑︎ Yes Have you read the Code of Conduct? ☑︎ Yes flyteorg/flyte

#1688 Add Cuda to ImageSpec Pull request opened by pingsutw TL;DR

from flytekit import task, workflow
from flytekit.image_spec import ImageSpec

new_flytekit = "git+<https://github.com/flyteorg/flytekit@8c7c43ad310ec74b4e8c06c2b4604b6cb9692f86>"
image_spec = ImageSpec(base_image="ubuntu20.04", packages=[new_flytekit], cuda="11.2.2", cudnn="8",
                       registry="pingsutw", apt_packages=["git"], python_version="3.9",)

Type ☐ Bug Fix ☐ Feature ☐ Plugin Are all requirements met? ☐ Code completed ☐ Smoke tested ☐ Unit tests added ☐ Code documentation added ☐ Any pending items have an associated Issue Complete description ^^^ Tracking Issue NA Follow-up issue NA flyteorg/flytekit ✅ All checks have passed 30/30 successful checks

#3772 [BUG] Imagespec envd build skipped when _not_ using pyflyte register Issue created by zeryx Describe the bug Imagespec dependencies, such as those added to a workflow/example.py file are automatically built when one calls

pyflyte register

. https://gist.github.com/zeryx/a3074a2aade6a7263c3d796d08891741 However, when you register and execute with

pyflyte run

,

pyflyte run --copy-all

, or

flytekit.remote.register_workflow

the ImageSpec build step is skipped, and the container_image argument failsover to use the base image. Expected behavior The expected behavior is regardless of the mechanism used to register the workflow onto a flyte cluster, if an imagespec image is required for any tasks; those ImageSpec steps should be built via envd and deployed to the provided registry. If that fails for whatever reason, the registration itself should fail; and return a useful error indicating that an ImageSpec build process failed for some reason; rather than registering the task with the default flyte base image. If one is using flytekit as a library and triggering register_workflow() or register_task() and an imagespec image is provided as a

container_image

argument; that ImageSpec image should attempt to build said image, or validate that the existing image already exists. Additional context to reproduce • pyflyte init, replace your workflow/example.py with https://gist.github.com/zeryx/a3074a2aade6a7263c3d796d08891741 • call pyflyte run --copy-all workflows/example.py • verify that envd is not being interacted with, and no images are written to registry Screenshots

pyflyte --config ~/.uctl/config.yaml run --remote --copy-all workflows/example.py wf
2023-06-12 11:23:46,062144 WARNING  {"asctime": "2023-06-12 11:23:46,062", "name": "flytekit.cli", "levelname": "WARNING", ignore.py:51
                                    "message": "Could not determine ignored files due to:\nb'fatal: not a git repository               
                                    (or any of the parent directories): .git\\n'\nNot applying any filters"}                           
Opening in existing browser session.
libva error: /usr/lib/x86_64-linux-gnu/dri/iHD_drv_video.so init failed
libva error: /usr/lib/x86_64-linux-gnu/dri/i965_drv_video.so init failed
127.0.0.1 - - [12/Jun/2023 11:24:13] "GET /callback?code=WnyxL2kywRTHw3tSSMLVO3T-38OLUwpQbOr5_8i_snY&state=feEPffKlUICJ63Y8bCuEtXBtgcU3kY2Oi1sQNrme5amT17xMnhgNOA HTTP/1.1" 200 -
Go to <https://demo.hosted.unionai.cloud/console/projects/flytesnacks/domains/development/executions/f8a710d7a26b74e77a9e> to see execution in the console.

Are you sure this issue hasn't been raised already? ☑︎ Yes Have you read the Code of Conduct? ☑︎ Yes flyteorg/flyte

Release - v1.1.97 New release published by flyte-bot Changelog • eadf07b Correctly validating error code on aborting terminal launchplans (#574) flyteorg/flytepropeller

Pending review on flyteorg/flytekit-python-template - Manage reminder [#40] ImageSpec overhaul (zeryx) 13 days stale · 13 days old · Waiting on _ @Niels Bantilan_

Pending review on flyteorg/flyteidl - Manage reminder [#297] Proposal : Settings proto (pmahindrakar-oss) 5 months stale · a year old · Waiting on _ @Haytham Abuelfutuh, katrogan_

#3773 [UI Feature] Update FLYTE_NAVIGATION behavior Issue created by jsonporter Motivation: Why do you think this is important? Users have requested two changes to how the navigation handles

FLYTE_NAVIGATION.items[]

. Currently, when a user exports

FLYTE_NAVIGATION

and provides values for custom navigation using

items[]

we render a dropdown with those links. Example from documentation:

{
    color:"white",          // optional - default NavBar text color, if not provided uses Flyte colors
    background:"black",     // optional - default NavBar background color, if not provided uses Flyte colors
    console:"FlyteConsole"  // optional - name of the default navigation, if not provided uses "Console"
    items:[                 // required - if no dropdown needed provide an empty array
        {title:"Remote", url:"<https://remote.site/>"},
        {title:"Dashboard", url:"+/projects/flytesnacks/executions?domain=development&duration=all"},
        {title:"Information", url:"/information"}
    ]
}

1. Add the ability to have dropdown links open in a new tab (current behavior redirects/adds to history) via some optional property (eg,

items[{title:"foo", url:"bar", target:"_blank"}]

) 2. Do not include

BASE_URL

as a default in the dropdown or at least gives users the choice where it appears in the list. This could work a couple of ways (eg, we could just rely on users clicking the logo to return to

console

from some external app or we could even add some sort of logic when creating that list to ensure that

console

is present but users could choose which position in the list

console

appears. Goal: What should the final outcome look like, ideally? 1. Users can choose to have certain links in the dropdown open in new tabs. 2. Users can choose where

console

appears in that list or if it appears at all. Propose: Link/Inline OR Additional context No response Are you sure this issue hasn't been raised already? ☑︎ Yes Have you read the Code of Conduct? ☑︎ Yes flyteorg/flyte

1 new commit pushed to

<https://github.com/flyteorg/flyte/tree/master|master>

by eapolinario

<https://github.com/flyteorg/flyte/commit/1ee95de9fab19af1e058ab02137a68101bf53c38|1ee95de9>

- Add agent to helm chart and single binary (#3454) flyteorg/flyte

#3454 Add agent to helm chart and single binary Pull request opened by pingsutw Tracking issue #3282 Describe your changes • Add a new deployment for agent service • Add a new container in flyte-sandbox deployment for agent service Check all the applicable boxes ☐ I updated the documentation accordingly. ☐ All new and existing tests passed. ☑︎ All commits are signed-off. Screenshots Note to reviewers flyteorg/flyte GitHub Actions: build-and-push-sandbox-bundled-image ✅ 11 other checks have passed 11/12 successful checks

#3774 Release candidate - 1.7.0 Pull request opened by eapolinario Release candidate 1.7.0 as per discussion on slack. Branches starting with the prefix "rc/" are protected and only modifiable by Flyte maintainers. flyteorg/flyte GitHub Actions: build-and-push-single-binary-image GitHub Actions: docs GitHub Actions: compile GitHub Actions: generate_kustomize GitHub Actions: test-bootstrap GitHub Actions: generate_helm ✅ 5 other checks have passed 5/11 successful checks

#3775 [BUG] ImageSpec not leveraging build cache for basic example Issue created by jeevb Describe the bug Repeated invocations of

pyflyte run

involving

ImageSpec

rebuild python and apt dependencies from scratch instead of leveraging cache. Expected behavior When an `ImageSpec`definition is not updated, image rebuilds should reuse cached layers where possible to speed up build times. Additional context to reproduce 1. Create a script,

wf.py

with the following contents:

from flytekit import task, workflow, ImageSpec

image = ImageSpec(
    packages=["smart-open[gcs]"],
    apt_packages=["git"],
)


@task(container_image=image)
def my_task() -> str:
    return "hello, world!"


@workflow
def my_wf() -> str:
    return my_task()

2. Run the following command multiple times:

pyflyte run --remote wf.py my_wf

Screenshots No response Are you sure this issue hasn't been raised already? ☑︎ Yes Have you read the Code of Conduct? ☑︎ Yes flyteorg/flyte

#1689 Add PydanticRemote plugin for creating pythonic interfaces to remote workflows Pull request opened by ggydush TL;DR Adds a PydanticRemote plugin that provides a simpler UX for retrieving and launching workflows from FlyteRemote. See demo for more information. Type ☐ Bug Fix ☑︎ Feature ☐ Plugin Are all requirements met? ☐ Code completed ☑︎ Smoke tested ☑︎ Unit tests added ☐ Code documentation added ☐ Any pending items have an associated Issue Complete description Most of the users of Flyte that I interact with are launching tasks and workflows from Python or Jupyter notebooks. Currently with

FlyteRemote.execute

, the user must know the workflow interface and provide workflow inputs as a Python dictionary when launching a workflow. This works, but often involves the user going back and forth between the FlyteConsole and their Jupyter Notebook to ensure all inputs are provided and are of the correct type. To address this, I created a version of FlyteRemote that leverages Pydantic to dynamically create workflow interfaces. This works by: 1. Querying Flyte for workflows and their input/output interface 2. Converting the interface to a Pydantic model and populating the docstring 3. Returning this model with a

execute

method that enables launching the workflow on a remote cluster flyteorg/flytekit ✅ All checks have passed 30/30 successful checks

#1690 Fix task type version in pytorch and mpi task plugin Pull request opened by yubofredwang TL;DR Previously the pytorch plugin is missing comments. We would suggest user to not set configurations for individual replica groups as they should essentially be the same for worker and master. The only fields user should set is number of worker and restart policies. Fix a bug in pytorch and mpijob to use task type version 1. Should be merged in after: flyteorg/flyteplugins#359 Type ☑︎ Bug Fix ☐ Feature ☐ Plugin Are all requirements met? ☑︎ Code completed ☑︎ Smoke tested ☑︎ Unit tests added ☐ Code documentation added ☐ Any pending items have an associated Issue Complete description Smoke tested on the following with corrected backend plugin 1. Elastic Pytorch Job 2. MPI Task 3. Horovod Job flyteorg/flytekit ✅ All checks have passed 30/30 successful checks

1 new commit pushed to

<https://github.com/flyteorg/flytesnacks/tree/master|master>

by ByronHsu

<https://github.com/flyteorg/flytesnacks/commit/4b031a565c3320d80edc23098f38be15986f29b0|4b031a56>

- Update tasks_and_workflows.md (#997) flyteorg/flytesnacks

#3776 [BUG] - Pod Template + Sync Resources does not delete annotations Issue created by franco-bocci Describe the bug When editing the PodTemplate and removing an annotation, the sync resources job does not delete the annotation from the existing pod templates. For this change to be applied, we had to delete the pod template and let Flyte re-create it with the updated template. Expected behavior PodTemplate changes should be reflected on the pod templates in each namespace, both when adding components as well as when removing components. Additional context to reproduce 1. Define a global pod template with 2 annotations 2. Let Flyte propagate the pod template to project's namespaces 3. Delete an annotation from the global pod template 4. Run the sync resources job 5. Get the pod template from the project's namespaces. There will be 2 annotations and not 1, as expected Screenshots No response Are you sure this issue hasn't been raised already? ☑︎ Yes Have you read the Code of Conduct? ☑︎ Yes flyteorg/flyte

#1691 Support configuring pip index url with the image spec Pull request opened by yini7777 Type: Bug Fix Complete description Currently, we are unable to specify the pip index URL during package installation, although it is supported by envd. The purpose of this PR is to enable configuring the custom pip index url. Tracking Issue flyteorg/flytekit DCO: DCO ✅ 1 other check has passed 1/2 successful checks

#1692 Support configuring pip index url with the image spec Pull request opened by yini7777 Type: Bug Fix Complete description Currently, we are unable to specify the pip index URL during package installation, although it is supported by envd. The purpose of this PR is to enable configuring the custom pip index url. Tracking Issue flyteorg/flyte#3765 flyteorg/flytekit ✅ All checks have passed 30/30 successful checks

#3777 [BUG] AutoIncrement sequence number data type is not consistent and not cycle Issue created by honnix Describe the bug This is probably due to change in gorm over time when handling

uint

and

int64

(I have no proof though, but https://github.com/flyteorg/flyteadmin/blob/master/pkg/repositories/models/resource.go#L17 might indicate that because

resources_id_seq

is

bigint

), but we have seen different sequence types in our database, some of them being

integer

while some of them being

bigint

. For example the newer sequence

description_entities_id_seq

is

bigint

while older things like

node_execution_events_id_seq

is

integer

that is obviously too short. Also the sequences are not cycle which means they will not wrap around when reaching max value. This is somewhat problematic especially for those

integer

typed sequences. We have scheduled data retention job to clean up database so old entries are deleted, but since the sequence number is not reusable, we can hit max value easily when it is an

integer

. This gives a quick overview of what we have in database (note that some of the types have been manually altered:

executions_id_seq

,

node_execution_events_id_seq

)

sequencename               | data_type | min_value |      max_value      | cycle 
-----------------------------------------+-----------+-----------+---------------------+-------
 projects_id_seq                         | integer   |         1 |          2147483647 | f
 description_entities_id_seq             | bigint    |         1 | 9223372036854775807 | f
 tasks_id_seq                            | integer   |         1 |          2147483647 | f
 signals_id_seq                          | bigint    |         1 | 9223372036854775807 | f
 workflows_id_seq                        | integer   |         1 |          2147483647 | f
 launch_plans_id_seq                     | integer   |         1 |          2147483647 | f
 executions_id_seq                       | bigint    |         1 | 9223372036854775807 | f
 execution_events_id_seq                 | integer   |         1 |          2147483647 | f
 node_executions_id_seq                  | integer   |         1 |          2147483647 | f
 node_execution_events_id_seq            | bigint    |         1 | 9223372036854775807 | f
 task_executions_id_seq                  | integer   |         1 |          2147483647 | f
 named_entity_metadata_id_seq            | integer   |         1 |          2147483647 | f
 named_entity_metadata_resource_type_seq | integer   |         1 |          2147483647 | f
 resources_id_seq                        | bigint    |         1 | 9223372036854775807 | f
 schedulable_entities_id_seq             | integer   |         1 |          2147483647 | f
 schedule_entities_snapshots_id_seq      | integer   |         1 |          2147483647 | f

Expected behavior All sequences are typed as

bigint

, and it would be nice to make them cycle (although a quick search didn't find how gorm could support that). Additional context to reproduce No response Screenshots No response Are you sure this issue hasn't been raised already? ☑︎ Yes Have you read the Code of Conduct? ☑︎ Yes flyteorg/flyte

#573 Use GetExecutionData instead Pull request opened by EngHabu TL;DR Removes deprecated code path to retrieve LP's outputs and moves to the separate GetExecutionData() call. Type ☑︎ Bug Fix ☐ Feature ☐ Plugin Are all requirements met? ☑︎ Code completed ☑︎ Smoke tested ☐ Unit tests added ☑︎ Code documentation added ☐ Any pending items have an associated Issue Tracking Issue fixes flyteorg/flyte#3753 flyteorg/flytepropeller GitHub Actions: Goreleaser GitHub Actions: Build & Push Flytepropeller Image GitHub Actions: Bump Version ✅ 10 other checks have passed 10/13 successful checks

1 new commit pushed to

<https://github.com/flyteorg/flytekit/tree/master|master>

by pingsutw

<https://github.com/flyteorg/flytekit/commit/c3b087623a15615a21966fa1641b55a960d22ea2|c3b08762>

- pyflyte run remote file (#1670) flyteorg/flytekit

#3778 enable actions for RC branches Pull request opened by jeevb flyteorg/flyte GitHub Actions: build-and-push-sandbox-bundled-image ✅ 9 other checks have passed 9/10 successful checks

1 new commit pushed to

<https://github.com/flyteorg/flytepropeller/tree/master|master>

by EngHabu

<https://github.com/flyteorg/flytepropeller/commit/89dfd5b3bb694d9a19108aa4c8a6d10a18fd3a32|89dfd5b3>

- Use GetExecutionData instead (#573) flyteorg/flytepropeller

1 new commit pushed to

<https://github.com/flyteorg/flyte/tree/master|master>

by jeevb

<https://github.com/flyteorg/flyte/commit/47fea41ce2f6783217a49f2191f442eda16ce646|47fea41c>

- enable actions for RC branches (#3778) flyteorg/flyte

#998 Feat: Add pod start and finish time in RFC3339 time format to logging link templating variables Pull request opened by fg91 Documentation for flyteorg/flyteplugins#360 flyteorg/flytesnacks GitHub Actions: Mark github pre-release as Release GitHub Actions: Publish artifacts to github release GitHub Actions: Create Prerelease GitHub Actions: Bump Version ✅ 26 other checks have passed 26/30 successful checks

#360 Feat: Add pod start and finish time in RFC3339 time format to logging link templating variables #minor Pull request opened by fg91 TL;DR Currently, the the pod start and finish time in unix time format (seconds since 1970) can be used to template logging links. This time format cannot be used in combination with google cloud logging. Instead, GCP logging requires timestamps in RFC3339 format:

timestamp >= "2016-11-29T23:00:00Z"
timestamp <= "2016-11-29T23:30:00Z"

This PR adds the pod start and finish time in this format to the variables available for templating logging links. Type ☐ Bug Fix ☑︎ Feature ☐ Plugin Are all requirements met? ☑︎ Code completed ☑︎ Smoke tested ☑︎ Unit tests added ☑︎ Code documentation added: flyteorg/flytesnacks#998 ☑︎ Any pending items have an associated Issue Complete description NA Tracking Issue NA Follow-up issue NA flyteorg/flyteplugins ✅ All checks have passed 7/7 successful checks

Release - v1.1.98 New release published by flyte-bot Changelog • 89dfd5b Use GetExecutionData instead (#573) flyteorg/flytepropeller

#3779 Update agent endpoint in the single binary Pull request opened by pingsutw Tracking issue Describe your changes update endpoint.

flyte-sandbox-http

->

flyte-binary-http

Check all the applicable boxes ☐ I updated the documentation accordingly. ☐ All new and existing tests passed. ☐ All commits are signed-off. Screenshots Note to reviewers flyteorg/flyte GitHub Actions: build-and-push-sandbox-bundled-image GitHub Actions: sandbox-bundled-functional-tests ✅ 10 other checks have passed 10/12 successful checks

#1693 Upload 'deck.html' only when the deck is enabled Pull request opened by Yicheng-Lu-llll TL;DR the file 'deck.html' is always uploaded regardless of the deck setting in my previous PR. This PR fix this issue by modifying the upload logic to only upload 'deck.html' when the deck setting is enabled. Type ☐ Bug Fix ☐ Feature ☐ Plugin Are all requirements met? ☐ Code completed ☐ Smoke tested ☐ Unit tests added ☐ Code documentation added ☐ Any pending items have an associated Issue Complete description How did you fix the bug, make the feature etc. Link to any design docs etc Tracking Issue https://github.com/flyteorg/flyte/issues/ Follow-up issue NA OR https://github.com/flyteorg/flyte/issues/ flyteorg/flytekit ✅ All checks have passed 30/30 successful checks

1 new commit pushed to

<https://github.com/flyteorg/flyte/tree/master|master>

by eapolinario

<https://github.com/flyteorg/flyte/commit/1ae254f8683699b68ecddc89d775fc5d39cc3d84|1ae254f8>

- Update agent endpoint in the single binary (#3779) flyteorg/flyte

1 new commit pushed to

<https://github.com/flyteorg/flyteplugins/tree/master|master>

by hamersaw

<https://github.com/flyteorg/flyteplugins/commit/53e63bc0291fcc1a1f58a0adb8e963b2710a0777|53e63bc0>

- Feat: Add pod start and finish time in RFC3339 time format to logging link templating variables #minor (#360) flyteorg/flyteplugins