input stringlengths 1 18.7k | output stringlengths 1 18.7k |
|---|---|
sorry what is native paritioning?
| basically multiple partitions with the same table name
really a secondary concern
does the above match the intended usage of the api?
|
basically multiple partitions with the same table name
really a secondary concern
does the above match the intended usage of the api?
| what is `schema_task_out()`?
so i understand correctly… this is the project to get schemas to read/write from bigquery instead of gcs right?
and yeah same, could you elaborate on what you mean by native partitioning?
|
what is `schema_task_out()`?
so i understand correctly… this is the project to get schemas to read/write from bigquery instead of gcs right?
and yeah same, could you elaborate on what you mean by native partitioning?
| let's not worry about native partitiong
otherwise yes
|
let's not worry about native partitiong
otherwise yes
| I think i’m gonna need Ketan Umare to confirm or deny here, but I think there’s a couple ways to do this. so first off the choice of the transformer that the type engine picks is based on the signature of the task, not the actual return type. so if you
```def t1() -> FlyteSchema:
return df```
it’ll still be the ... |
I think i’m gonna need Ketan Umare to confirm or deny here, but I think there’s a couple ways to do this. so first off the choice of the transformer that the type engine picks is based on the signature of the task, not the actual return type. so if you
```def t1() -> FlyteSchema:
return df```
it’ll still be the ... | Yee yes, I just disabled overriding by default, there is actually a comment in the TypeEngine. We could allow overriding, as long as the the overrider knows that they are doing it.
It was done as a security thing
|
Yee yes, I just disabled overriding by default, there is actually a comment in the TypeEngine. We could allow overriding, as long as the the overrider knows that they are doing it.
It was done as a security thing
| Sorry overriding what
|
Sorry overriding what
| overriding the default type transformer for a type
|
overriding the default type transformer for a type
| this is a related thread btw ketan <https://flyte-org.slack.com/archives/CNMKCU6FR/p1621001152011800>
Type transformer stuff -
Registration of type transformers gets called <https://github.com/flyteorg/flytekit/blob/26baa890a5190ff61401ea627b149b59f8e22520/flytekit/types/schema/types.py#L399|like so> and ends up trigge... |
this is a related thread btw ketan <https://flyte-org.slack.com/archives/CNMKCU6FR/p1621001152011800>
Type transformer stuff -
Registration of type transformers gets called <https://github.com/flyteorg/flytekit/blob/26baa890a5190ff61401ea627b149b59f8e22520/flytekit/types/schema/types.py#L399|like so> and ends up trigge... | yep :slightly_smiling_face:
a couple questions:
1. Is there anyway to control the init order of type registration? i suppose to I can just import the others first so that mine runs last but that's a bit unwieldy
2. when discussing the notebook reader idea, someone pointed out that for local runs there's no serialized t... |
yep :slightly_smiling_face:
a couple questions:
1. Is there anyway to control the init order of type registration? i suppose to I can just import the others first so that mine runs last but that's a bit unwieldy
2. when discussing the notebook reader idea, someone pointed out that for local runs there's no serialized t... | Ya for local there is no serialization as a perf thing. You can always serialize everything if needed
|
Ya for local there is no serialization as a perf thing. You can always serialize everything if needed
| 1. As in you want to ensure that your transformer gets picked up? I think that’s fine. cuz there’ll be a flag right? overwrite okay. and flytekit will promise to never use it :crossed_fingers: . And given python loading semantics (as you suggest), as long as your custom transformer file imports the regular type en... |
1. As in you want to ensure that your transformer gets picked up? I think that’s fine. cuz there’ll be a flag right? overwrite okay. and flytekit will promise to never use it :crossed_fingers: . And given python loading semantics (as you suggest), as long as your custom transformer file imports the regular type en... | 1. my concern was more about my depending spark plugin (or other interplugin deps)
2. ketan answered i think, but the scheme to transfer data between flyte <-> notebook requires data to be pushed to disk. that doesn't happen in local runs
|
Howdy! Is there are way for a type transformer/reader/writer to get hold of the task that is producing/ingesting it?
I’m trying to convey metadata to the type code using <task>.with_overrides(…)
| the transformers are independent entities and currently have no knowledge of the task itself. This allows easy decoupling
what do you want to convey?
usually the best way to convey metadata about a task, is to create a new type (derivative of an internal type) and that adds additional metadata in the serialization
|
the transformers are independent entities and currently have no knowledge of the task itself. This allows easy decoupling
what do you want to convey?
usually the best way to convey metadata about a task, is to create a new type (derivative of an internal type) and that adds additional metadata in the serialization
| Hi Ketan! I'm trying to tell the Bigquery schema type where to write the result in a way that's agnostic to the task producing the data, ie leaving that decision to the consumers of the task.
|
Hi Ketan! I'm trying to tell the Bigquery schema type where to write the result in a way that's agnostic to the task producing the data, ie leaving that decision to the consumers of the task.
| is it the `bq://…..` ?
Dylan Wilder and I had a chat yesterday. not sure if this is related?
|
is it the `bq://…..` ?
Dylan Wilder and I had a chat yesterday. not sure if this is related?
| related but not the same
basically i think what stefan is trying to do is access the kwargs from `with_overrides()` from within a task as a way to provide config (eg what bq table to write too)
alterantive is to have tasks hardcode output locations, or pass as an arg which breaks "tasks being agnostic of IO" principle... |
related but not the same
basically i think what stefan is trying to do is access the kwargs from `with_overrides()` from within a task as a way to provide config (eg what bq table to write too)
alterantive is to have tasks hardcode output locations, or pass as an arg which breaks "tasks being agnostic of IO" principle... | so the idea of `with_overrides` is to override `task configuration`. So if you model bq tasks - for queries then, the with_overrides can do what you want
|
so the idea of `with_overrides` is to override `task configuration`. So if you model bq tasks - for queries then, the with_overrides can do what you want
| yea that's what i figured
|
is there an easy way to get a python type from a literal :sweat_smile: ? I'm just here writing a mapping function
one literal at a time
| haha ya, there is a guesser
<https://github.com/flyteorg/flytekit/blob/26baa890a5190ff61401ea627b149b59f8e22520/flytekit/core/type_engine.py#L67>
rather - <https://github.com/flyteorg/flytekit/blob/26baa890a5190ff61401ea627b149b59f8e22520/flytekit/core/type_engine.py#L325>
|
haha ya, there is a guesser
<https://github.com/flyteorg/flytekit/blob/26baa890a5190ff61401ea627b149b59f8e22520/flytekit/core/type_engine.py#L67>
rather - <https://github.com/flyteorg/flytekit/blob/26baa890a5190ff61401ea627b149b59f8e22520/flytekit/core/type_engine.py#L325>
| wait what is a `LiteralType` it's different than a `Literal`
|
wait what is a `LiteralType` it's different than a `Literal`
| ya
LiteralType is how you declare/define the type of a literal
so e.g. `int a = 5` -> int is the LiteralType and 5 is the actual literal
a is the name of the literal
|
ya
LiteralType is how you declare/define the type of a literal
so e.g. `int a = 5` -> int is the LiteralType and 5 is the actual literal
a is the name of the literal
| so from a `Literal` i still have to figure out it's type
|
so from a `Literal` i still have to figure out it's type
| hmm wait
why do you not know the type?
i mean the python type and you know the literal
Yee is there a helper to do this?
from a literal it should be easy to create a python type
we do this a lot in golang
|
hmm wait
why do you not know the type?
i mean the python type and you know the literal
Yee is there a helper to do this?
from a literal it should be easy to create a python type
we do this a lot in golang
| i don’t have that.
should be easy to do in most cases (harder/impossible for stuff like generics)
this hasn’t come up because in all our use-cases we always have the user code or the flyte typedinterface
|
i don’t have that.
should be easy to do in most cases (harder/impossible for stuff like generics)
this hasn’t come up because in all our use-cases we always have the user code or the flyte typedinterface
| Yea I don’t have that. I slogged my way through a mapping :sweat_smile:
|
Yea I don’t have that. I slogged my way through a mapping :sweat_smile:
| :disappointed: sorry about that. feel free to commit that back!
|
:disappointed: sorry about that. feel free to commit that back!
| yea was thinking that
is this insane or does it look right?
```def unbox_type(lit) -> type:
"""
Unbox a literal to reveal the magic inside
"""
if isinstance(lit, _lit.Literal):
return unbox_type(lit.value)
elif isinstance(lit, _lit.LiteralCollection):
if len(lit.literals):
... |
yea was thinking that
is this insane or does it look right?
```def unbox_type(lit) -> type:
"""
Unbox a literal to reveal the magic inside
"""
if isinstance(lit, _lit.Literal):
return unbox_type(lit.value)
elif isinstance(lit, _lit.LiteralCollection):
if len(lit.literals):
... | Blob
Is files
|
Blob
Is files
| right but there's no root type like FlyteSchema
|
right but there's no root type like FlyteSchema
| There is flytefile
Which is just a shadow type
|
There is flytefile
Which is just a shadow type
| yea the api literally doesn't have an equivalent for Blob though, unless i just can't find it. consider the following
```FlyteSchema == literal2pytype(_lit.Literal(scalar=_lit.Scalar(schema=_lit.Schema(..., ...)))))
??? == literal2pytype(_lit.Literal(scalar=_lit.Scalar(blob=_lit.Blob(..., ...))))```
|
yea the api literally doesn't have an equivalent for Blob though, unless i just can't find it. consider the following
```FlyteSchema == literal2pytype(_lit.Literal(scalar=_lit.Scalar(schema=_lit.Schema(..., ...)))))
??? == literal2pytype(_lit.Literal(scalar=_lit.Scalar(blob=_lit.Blob(..., ...))))```
| What do you mean - you can do this - <https://github.com/flyteorg/flytekit/blob/master/flytekit/types/file/file.py|https://github.com/flyteorg/flytekit/blob/master/flytekit/types/file/file.py>
Line 137 init
Other thing is, the flytekit types
Don’t matter, you can always create a type representation
|
What do you mean - you can do this - <https://github.com/flyteorg/flytekit/blob/master/flytekit/types/file/file.py|https://github.com/flyteorg/flytekit/blob/master/flytekit/types/file/file.py>
Line 137 init
Other thing is, the flytekit types
Don’t matter, you can always create a type representation
| right but how do i know it's not actually a FlyteDirectory which is also blob. unless that's _deterministically_ determined by blob metadata?
|
right but how do i know it's not actually a FlyteDirectory which is also blob. unless that's _deterministically_ determined by blob metadata?
| Yup
Blob metadata type
|
Yup
Blob metadata type
| oh ok that's what i was missing. in general though i think this is a bad idea since types aren't 1-1. any number of TypeTransformers can be made for the same literal
and if someone overrides the guess method the result isn't deterministic
|
oh ok that's what i was missing. in general though i think this is a bad idea since types aren't 1-1. any number of TypeTransformers can be made for the same literal
and if someone overrides the guess method the result isn't deterministic
| They are 1:1 for primitives and lower level types
|
They are 1:1 for primitives and lower level types
| yea
|
yea
| But files, directories are no equivalent in programming languages
They are general purpose - as in an image a csv both are files
|
But files, directories are no equivalent in programming languages
They are general purpose - as in an image a csv both are files
| for sure. but the FLYTE_TASK_NAME approach makes more sense don't you agree? since it is the actual api specified by the user?
the guessing approach has a lot of caveats
|
for sure. but the FLYTE_TASK_NAME approach makes more sense don't you agree? since it is the actual api specified by the user?
the guessing approach has a lot of caveats
| Yup, guessing is dangerous - as Flyte type conversion is lossy - like Java
This makes them
Portable
But Yee added a new type
Of task, which allows having pure python portable tasks- I.e no need of building a container
Look at sqlite3 task
In there he has to guess
Types
Which is ok - as there are no files
|
Yup, guessing is dangerous - as Flyte type conversion is lossy - like Java
This makes them
Portable
But Yee added a new type
Of task, which allows having pure python portable tasks- I.e no need of building a container
Look at sqlite3 task
In there he has to guess
Types
Which is ok - as there are no files
| On a hike. No laptop, but there’s no guessing logic in sqlite task yet - that pr isn’t merged yet (waiting on yuvraj to build the flytekit docker file
|
what purpose does a task name serve when it's not `module.name` eg in the case of notebook tasks.
| its just an identifier. The name should be unique in a project
as it is part of the ID
|
its just an identifier. The name should be unique in a project
as it is part of the ID
| why ever not use the module.name then? Any reason a user would want to specify? It seems to either be required or nonspecifiable
|
why ever not use the module.name then? Any reason a user would want to specify? It seems to either be required or nonspecifiable
| you mean like here? <https://github.com/flyteorg/flytesnacks/blob/master/cookbook/integrations/flytekit_plugins/papermilltasks/simple.py#L32>
what did you mean by non-specifiable
|
you mean like here? <https://github.com/flyteorg/flytesnacks/blob/master/cookbook/integrations/flytekit_plugins/papermilltasks/simple.py#L32>
what did you mean by non-specifiable
| for PythonFunctionTasks you cannot override
this isn't a big deal really, but i got stuck on it for a sec since i was expecting FLYTE_INTERNAL_TASK_NAME to point to something importable
so trying to understand the intent
|
Ketan Umare discussion around `FlyteContext` creation wrt parent processes writing to the environment and subprocesses creating contexts from those variables.
| do you want to talk?
|
do you want to talk?
| sure
Dylan Wilder do you have time?
|
sure
Dylan Wilder do you have time?
| for a slack convo for sure
|
for a slack convo for sure
| let me try to rehash here then.
since the existing `entrypoint.py` already creates (in a very complicated, very nested way) much of the FC settings from env vars, perhaps it makes sense to move them into FlyteContextManager, so we’d have a `FCM.create_from_env()` method on it.The default behavior for FlyteContexts woul... |
let me try to rehash here then.
since the existing `entrypoint.py` already creates (in a very complicated, very nested way) much of the FC settings from env vars, perhaps it makes sense to move them into FlyteContextManager, so we’d have a `FCM.create_from_env()` method on it.The default behavior for FlyteContexts woul... | total devils advocate: what FC was intimately linked to the env? it can be viewed as the python mapping of the current env. this might ensure a few things
1. within any given context the env and repr are in sync
2. The env/config can be easily and safely be overridden in any subcontext
3. The context can seamlessly per... |
total devils advocate: what FC was intimately linked to the env? it can be viewed as the python mapping of the current env. this might ensure a few things
1. within any given context the env and repr are in sync
2. The env/config can be easily and safely be overridden in any subcontext
3. The context can seamlessly per... | oh no worries, thanks for doing this! and for making us deal with these questions, it’s good.
|
oh no worries, thanks for doing this! and for making us deal with these questions, it’s good.
| yea total outsiders perspective. but what if flytecontext was the env (plus derived accessors etc). you could serialize/deserialize, checkpoint and pass around between applications. even across sdks if that makes any sense
|
yea total outsiders perspective. but what if flytecontext was the env (plus derived accessors etc). you could serialize/deserialize, checkpoint and pass around between applications. even across sdks if that makes any sense
| sorry, stepped off… i see what you’re saying, but I think i’ll let ketan answer. part of my hesitation is definitely from past experience when we were reading config values across the codebase. which made it so that doing anything required being cognizant of which config files were active, making sure config files we... |
sorry, stepped off… i see what you’re saying, but I think i’ll let ketan answer. part of my hesitation is definitely from past experience when we were reading config values across the codebase. which made it so that doing anything required being cognizant of which config files were active, making sure config files we... | makes sense, i think this might actually reduce spaghetti config
|
makes sense, i think this might actually reduce spaghetti config
| sorry i was busy with something, but, I like the idea of a portable config. But, I also agree with Yee - I am old-fashioned and prefer - dependency injection. Thus having clear knowledge of what is used where is much better and easier to maintain longer term
we broke it only for the user `current_context` method, that ... |
sorry i was busy with something, but, I like the idea of a portable config. But, I also agree with Yee - I am old-fashioned and prefer - dependency injection. Thus having clear knowledge of what is used where is much better and easier to maintain longer term
we broke it only for the user `current_context` method, that ... | does it make sense to have a `FlyteContextFactory` (speaking of old school). is that what we're talking about here?
in my mind this was kinda what the manager was
|
does it make sense to have a `FlyteContextFactory` (speaking of old school). is that what we're talking about here?
in my mind this was kinda what the manager was
| ya thats what it is
|
ya thats what it is
| yea to be clear i dont' think what i'm suggesting is incompatible with dep injection. i don't think info should be grabbed from global context everywhere. rather the flytecontext is the vehicle for all dependency injection of environmental state
|
yea to be clear i dont' think what i'm suggesting is incompatible with dep injection. i don't think info should be grabbed from global context everywhere. rather the flytecontext is the vehicle for all dependency injection of environmental state
| hey Dylan Wilder just to wrap this up… i think ketan came to a compromise, it’s okay to go the route of FCM.from_xyz(), so feel free to go ahead with that - the constraint however, is that the `from_xyz`, if it’s environment variable based, should not intersect/read from/involve any of the existing configuration settin... |
hey Dylan Wilder just to wrap this up… i think ketan came to a compromise, it’s okay to go the route of FCM.from_xyz(), so feel free to go ahead with that - the constraint however, is that the `from_xyz`, if it’s environment variable based, should not intersect/read from/involve any of the existing configuration settin... | hmm that's not exactly what i was talking about, but i can do whatever you guys want. I can also just pickle the flyte context as part of the papermill plugin
also qq: it looks like different environment is passed as part of a notebook task. is that true? how does that work?
for example i don't see FLYTE_PLATFORM_CLOU... |
hmm that's not exactly what i was talking about, but i can do whatever you guys want. I can also just pickle the flyte context as part of the papermill plugin
also qq: it looks like different environment is passed as part of a notebook task. is that true? how does that work?
for example i don't see FLYTE_PLATFORM_CLOU... | sorry was in a meeting - is that not what you were talking about? maybe i missed something?
and are you saying that EVs in the parent process don’t make it into the notebook process? like `os.environ["FLYTE_PLATFORM_CLOUD_PROVIDER"]` is present in the pod but doesn’t exist in the ipynb?
|
sorry was in a meeting - is that not what you were talking about? maybe i missed something?
and are you saying that EVs in the parent process don’t make it into the notebook process? like `os.environ["FLYTE_PLATFORM_CLOUD_PROVIDER"]` is present in the pod but doesn’t exist in the ipynb?
| i haven't verified the pod. it's present in a PythonFunctionTask but not an NotebookTask
re: above what I was talking about was probably a bit too grandiose and not exactly necessary for the problem at hand
|
Hey all! I’m currently using `map_task` and noticed, that one of my decorated tasks (custom pod spec to access internal things, environment vars, etc) does not receive the task configs we need when I try to map it. (cannot access our internal tools like our other decorated tasks normally do). I tried passing things in ... | I'm curious as to whether `task_config` is intentionally set to `None` for some reason here.
|
I'm curious as to whether `task_config` is intentionally set to `None` for some reason here.
| Yea, I was thinking the same thing
|
Yea, I was thinking the same thing
| shouldn’t the pod config be on the decotared_mappable_task instead?
|
shouldn’t the pod config be on the decotared_mappable_task instead?
| I quickly grabbed the pod yamls for the pods running the mapper_execute_decoared_task_name and by the looks of the yamls, they dont have the volume mounts and environment variables like the rest of our tasks do. (the rest of the tasks have the same task_configs and settings)
|
I quickly grabbed the pod yamls for the pods running the mapper_execute_decoared_task_name and by the looks of the yamls, they dont have the volume mounts and environment variables like the rest of our tasks do. (the rest of the tasks have the same task_configs and settings)
| can you copy/paste the signature for decotared_mappable_task
|
[flyteorg/flyte] Issue opened by wild-endeavor
| Anooj Patel
this was the issue i mentioned.
i assume the resource overriding worked?
(but not the volume mounting)
|
Anooj Patel
this was the issue i mentioned.
i assume the resource overriding worked?
(but not the volume mounting)
| Yea! Setting default requests vis the config worked for now. Thank you for making the issue!
|
Yea! Setting default requests vis the config worked for now. Thank you for making the issue!
| hey Anooj Patel i think Katrina Rogan will be working on something related to this.
not sure on the timing, but we’ll link to that other ticket and will update you when complete
|
hey Anooj Patel i think Katrina Rogan will be working on something related to this.
not sure on the timing, but we’ll link to that other ticket and will update you when complete
| Awesome! Looking forward to helping however I can :)
|
hey Max Hoffman the dolt plugin tests are giving me an error locally but works fine in CI:
```_____________________________________ ERROR at setup of test_dolt_table_write ______________________________________
repo_dir = '/var/folders/4q/frdnh9l10h53gggw1m59gr9m0000gn/T/tmpwt7z0aqy/foo', error = False
@staticmet... | it looks like your tests might be using a local version of dolt that's unconfigured?
I made the tests download a fresh version of dolt basically every time they run to avoid the image dependency
when dolt is downloaded in test setup, this configuration line fixes the error it's giving you
<https://github.com/flyteorg/f... |
it looks like your tests might be using a local version of dolt that's unconfigured?
I made the tests download a fresh version of dolt basically every time they run to avoid the image dependency
when dolt is downloaded in test setup, this configuration line fixes the error it's giving you
<https://github.com/flyteorg/f... | > you’ve locally installed dolt
is this different from installing doltcli and dolt-integrations?
|
> you’ve locally installed dolt
is this different from installing doltcli and dolt-integrations?
| yeah doltcli is pure python, no dependencies except dataclasses for python<3.7
dolt is a program compiled from golang
installed separately
|
yeah doltcli is pure python, no dependencies except dataclasses for python<3.7
dolt is a program compiled from golang
installed separately
| I don’t have a local installation of dolt, and it looks like the `dev_setup.sh` script is being run correctly, so not sure what’s going on
|
I don’t have a local installation of dolt, and it looks like the `dev_setup.sh` script is being run correctly, so not sure what’s going on
| i can hop on a debug session if you want, if we dig through the logs there might be an error relating to the install script exiting
i see the "user.name not found" a lot
but the call itself does not usually fail, it's usually something else upstream preventing the call
|
i can hop on a debug session if you want, if we dig through the logs there might be an error relating to the install script exiting
i see the "user.name not found" a lot
but the call itself does not usually fail, it's usually something else upstream preventing the call
| thanks! I solved the issue locally by doing subprocess calls in the `dolt_install` fixture to set the configs:
```subprocess.call([dolt_path, "config", "--global", "--add", "user.name", "Bojack Horseman"])
subprocess.call([dolt_path, "config", "--global", "--add", "user.email", "<mailto:bojack@horseman.com|boja... |
thanks! I solved the issue locally by doing subprocess calls in the `dolt_install` fixture to set the configs:
```subprocess.call([dolt_path, "config", "--global", "--add", "user.name", "Bojack Horseman"])
subprocess.call([dolt_path, "config", "--global", "--add", "user.email", "<mailto:bojack@horseman.com|boja... | very weird! but thanks for the feedback, if you find other issues or manage to reprod feel free to DM me, i'm always trying to make these things more stable
the final two lines aren't required, only `user.name` and `user.email`
i add the metrics lines to avoid my work directly impacting our usage metrics
which i think ... |
very weird! but thanks for the feedback, if you find other issues or manage to reprod feel free to DM me, i'm always trying to make these things more stable
the final two lines aren't required, only `user.name` and `user.email`
i add the metrics lines to avoid my work directly impacting our usage metrics
which i think ... | Max Hoffman Yee made this change to dolt testing fixture to make tests more stable: <https://github.com/flyteorg/flytekit/pull/489>
again, didn’t end up figuring out why the `dev_setup.sh` script doesn’t work on my local, but it does work on CI.
|
Max Hoffman Yee made this change to dolt testing fixture to make tests more stable: <https://github.com/flyteorg/flytekit/pull/489>
again, didn’t end up figuring out why the `dev_setup.sh` script doesn’t work on my local, but it does work on CI.
| +1 from me.
had issues with that before
|
+1 from me.
had issues with that before
| likewise this looks fine to me
|
is there a way to use previously registered workflows as subworkflows:
1. Register workflow `foo` with image `foo:versionX`
2. Use workflow `foo` with version `versionX` as subworkflow within workflow `bar`
3. Register workflow `bar` with image `bar:versionY`
it seems like this might be technically possible, but unl... | Reference workflows?
But those require you to redeclare the interface.
|
Reference workflows?
But those require you to redeclare the interface.
| oh interesting. is there an example of this?
|
oh interesting. is there an example of this?
| Don’t have my laptop on me rn
|
Don’t have my laptop on me rn
| no rush! thanks for the pointer :)
|
no rush! thanks for the pointer :)
| For some reason we have hidden this feature
Babis Kiosidis was going to add a couple examples
|
For some reason we have hidden this feature
Babis Kiosidis was going to add a couple examples
| :pray:
|
:pray:
| <https://github.com/flyteorg/flytekit/blob/master/tests/flytekit/unit/core/test_references.py#L86> does this help Jeev B?
|
<https://github.com/flyteorg/flytekit/blob/master/tests/flytekit/unit/core/test_references.py#L86> does this help Jeev B?
| holy cow, yes! thanks so much Yee!
|
holy cow, yes! thanks so much Yee!
| Hey :wave: completely forgot about this. I will try to fix it during the week :slightly_smiling_face:
|
Hey :wave: completely forgot about this. I will try to fix it during the week :slightly_smiling_face:
| Yee: seeing this error with reference workflows:
```INFO:root:Skipping entity resource_type: WORKFLOW
project: "flyte-sandbox-jeevb"
domain: "main"
name: "workflows.single_step.compute_square_wf"
version: "ff414a48bba9"
because already registered
Traceback (most recent call last):
File "/fn/lib/python3.8/site-packag... |
Yee: seeing this error with reference workflows:
```INFO:root:Skipping entity resource_type: WORKFLOW
project: "flyte-sandbox-jeevb"
domain: "main"
name: "workflows.single_step.compute_square_wf"
version: "ff414a48bba9"
because already registered
Traceback (most recent call last):
File "/fn/lib/python3.8/site-packag... | which version?
would it be possible for you to try out 0.19.0b7? we should be cutting a non-beta for that today or tomorrow.
|
which version?
would it be possible for you to try out 0.19.0b7? we should be cutting a non-beta for that today or tomorrow.
| I think it might be pointing to this object: <https://github.com/flyteorg/flytekit/blob/b773a2015e83a1991a3f06260620d79ff670f247/flytekit/core/workflow.py#L62>
instead of:
<https://github.com/flyteorg/flytekit/blob/b773a2015e83a1991a3f06260620d79ff670f247/flytekit/models/core/workflow.py#L464>
a bit confusing here:
<... |
I think it might be pointing to this object: <https://github.com/flyteorg/flytekit/blob/b773a2015e83a1991a3f06260620d79ff670f247/flytekit/core/workflow.py#L62>
instead of:
<https://github.com/flyteorg/flytekit/blob/b773a2015e83a1991a3f06260620d79ff670f247/flytekit/models/core/workflow.py#L464>
a bit confusing here:
<... | oh boy
|
oh boy
| am i missing something?
lol
this is with 0.19.0b7
|
am i missing something?
lol
this is with 0.19.0b7
| okay… let me think things through. i think i confused myself earlier.
|
okay… let me think things through. i think i confused myself earlier.
| the test probably passes since it doesnt have to serialize the workflow...
|
the test probably passes since it doesnt have to serialize the workflow...
| while i’m un-confusing myself… would you be okay trying the launch plan?
|
while i’m un-confusing myself… would you be okay trying the launch plan?
| yea. is there an example?
|
yea. is there an example?
| it’d be something like this
```import os
from flytekit.core.launch_plan import ReferenceLaunchPlan
from flytekit.core.base_task import kwtypes
rlp = ReferenceLaunchPlan(project=os.environ["FLYTE_INTERNAL_PROJECT"], domain=os.environ["FLYTE_INTERNAL_DOMAIN"], name="workflows.single_step.compute_square_wf",version="ff41... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.