Enabling non-hardcoded run of our test infrastructure #145
+75
−28
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
ajakovljevicTT
force-pushed
the
ajakovljevic/solving_the_twochip_issue
branch
3 times, most recently
from
2fae952 to
7f648a4
Compare
ajakovljevicTT
force-pushed
the
ajakovljevic/solving_the_twochip_issue
branch
2 times, most recently
from
895a20e to
e004717
Compare
|
@ajakovljevicTT please rebase and ping me again to take a look. That bug should be gone now that old infra is gone. |
ajakovljevicTT
force-pushed
the
ajakovljevic/solving_the_twochip_issue
branch
from
e004717 to
2e5c6d5
Compare
ajakovljevicTT
force-pushed
the
ajakovljevic/solving_the_twochip_issue
branch
from
ccd0cf2 to
418f9d4
Compare
kmitrovicTT
reviewed
Jan 10, 2025
| LoadedExecutableInstance::Unwrap(args->executable) | ||
| ->addressable_devices(); | ||
| int num_addressable_devices = | ||
| LoadedExecutableInstance::Unwrap(args->executable) | ||
| ->image_->get_num_addresible_devices(); |
There was a problem hiding this comment.
Maybe add assert num_addressable_devices == addressable_devices.size().
tests/infra/device_runner.py
Outdated
| device = device_connector.connect_device(device_type) | ||
|
|
||
| with jax.default_device(device): | ||
| return device_workload.execute() | ||
|
|
||
| @staticmethod | ||
| def _put_on_device(device_type: DeviceType, workload: Workload) -> Workload: | ||
| def _put_on_device( | ||
| device_type: DeviceType, workload: Workload, num_device: int = 0 |
There was a problem hiding this comment.
Maybe rearrange a bit to keep device_type and device_num next to each other.
tests/infra/device_runner.py
Outdated
| @@ -64,18 +64,22 @@ def put_tensors_on_gpu(*tensors: Tensor) -> Sequence[Tensor]: | |||
| raise NotImplementedError("Support for GPUs not implemented") | |||
|
|
|||
| @staticmethod | |||
| def _run_on_device(device_type: DeviceType, workload: Workload) -> Tensor: | |||
| def _run_on_device( | |||
| device_type: DeviceType, workload: Workload, num_device: int = 0 | |||
ajakovljevicTT
force-pushed
the
ajakovljevic/solving_the_twochip_issue
branch
from
112e32a to
680a3ea
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
As per issue #9, our test infra had a problem when detecting two TT devices, as is the case on the two-chip N300, which was workarounded by hardcoding which devices to use. This PR fixes that, instead just setting that the number of devices that the backend uses is capped to 1. Additionally, there was a bug in our old infra which ran the tests on cpu instead of a TT device, so this PR also fixes that.
Fixes #9.