Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Refactoring Workload Stable Stack Images #398

Merged
merged 6 commits into from
Sep 23, 2024
Merged

Refactoring Workload Stable Stack Images #398

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
15e5e19
Refactoring Workload Stable Stack Images
parambole Sep 19, 2024
901d898
Merge branch 'master' into parambole/jax_stable_stack
parambole Sep 19, 2024
c44f560
Merge branch 'master' into parambole/jax_stable_stack
parambole Sep 19, 2024
9be0984
Merge branch 'master' into parambole/jax_stable_stack
parambole Sep 20, 2024
e9ec1aa
Merge branch 'master' into parambole/jax_stable_stack
parambole Sep 20, 2024
51769e5
Merge branch 'master' into parambole/jax_stable_stack
parambole Sep 23, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion dags/examples/maxtext_sweep_gke_example_dag.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -48,7 +48,7 @@
time_out_in_min=60,
base_output_directory=base_output_directory,
num_slices=[1],
docker_image=DockerImage.MAXTEXT_TPU_JAX_STABLE.value,
docker_image=DockerImage.MAXTEXT_TPU_JAX_STABLE_STACK.value,
run_name_prefix="maxtext-16b",
base_run_model_cmds=base_run_model_cmds,
sweep_params={"M_PER_DEVICE_BATCH_SIZE": [2, 4, 8]},
Expand Down
4 changes: 2 additions & 2 deletions dags/imagegen_devx/jax_stable_stack_e2e.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -58,7 +58,7 @@
f"base_output_directory={gcs_bucket.BASE_OUTPUT_DIR}/maxtext/jax-ss/automated/{current_datetime}",
),
test_name=f"maxtext-jax-ss-{accelerator}-{slice_num}x",
docker_image=DockerImage.MAXTEXT_TPU_JAX_STABLE.value,
docker_image=DockerImage.MAXTEXT_TPU_JAX_STABLE_STACK.value,
test_owner=test_owner.PARAM_B,
).run()

Expand All @@ -76,6 +76,6 @@
f"output_dir={gcs_bucket.BASE_OUTPUT_DIR}/maxdiffusion/jax-ss/automated/{current_datetime}",
),
test_name=f"maxdiffusion-jax-ss-{accelerator}-{slice_num}x",
docker_image=DockerImage.MAXDIFFUSION_TPU_STABLE.value,
docker_image=DockerImage.MAXDIFFUSION_TPU_JAX_STABLE_STACK.value,
test_owner=test_owner.PARAM_B,
).run()
2 changes: 1 addition & 1 deletion dags/multipod/legacy.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,7 +24,7 @@
# Run once a day at 9 am UTC (1 am PST)
SCHEDULED_TIME = "0 9 * * *" if composer_env.is_prod_env() else None
DOCKER_IMAGE = {
SetupMode.STABLE: DockerImage.MAXTEXT_TPU_JAX_STABLE,
SetupMode.STABLE: DockerImage.MAXTEXT_TPU_JAX_STABLE_STACK,
SetupMode.NIGHTLY: DockerImage.MAXTEXT_TPU_JAX_NIGHTLY,
}

Expand Down
2 changes: 1 addition & 1 deletion dags/multipod/maxtext_checkpointing.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,7 @@
base_output_directory = f"{gcs_bucket.BASE_OUTPUT_DIR}/maxtext_checkpointing"
dataset_path = gcs_bucket.MAXTEXT_DIR
docker_images = [
(SetupMode.STABLE, DockerImage.MAXTEXT_TPU_JAX_STABLE),
(SetupMode.STABLE, DockerImage.MAXTEXT_TPU_JAX_STABLE_STACK),
(SetupMode.NIGHTLY, DockerImage.MAXTEXT_TPU_JAX_NIGHTLY),
]
test_configs = {
Expand Down
2 changes: 1 addition & 1 deletion dags/multipod/maxtext_configs_aot.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,7 +51,7 @@
}
num_slices = [1, 2]
docker_images = [
(SetupMode.STABLE, DockerImage.MAXTEXT_TPU_JAX_STABLE),
(SetupMode.STABLE, DockerImage.MAXTEXT_TPU_JAX_STABLE_STACK),
(SetupMode.NIGHTLY, DockerImage.MAXTEXT_TPU_JAX_NIGHTLY),
]

Expand Down
2 changes: 1 addition & 1 deletion dags/multipod/maxtext_convergence.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -63,7 +63,7 @@
time_out_in_min=300,
test_name=test_name,
run_model_cmds=run_command,
docker_image=DockerImage.MAXTEXT_TPU_JAX_STABLE.value,
docker_image=DockerImage.MAXTEXT_TPU_JAX_STABLE_STACK.value,
test_owner=test_owner.MATT_D,
base_output_directory=base_output_directory,
metric_aggregation_strategy=metric_config.AggregationStrategy.LAST,
Expand Down
6 changes: 3 additions & 3 deletions dags/multipod/maxtext_end_to_end.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -47,7 +47,7 @@
time_out_in_min=60,
test_name=f"{test_name_prefix}-stable-{model}",
run_model_cmds=(f"bash end_to_end/{test_script}.sh",),
docker_image=DockerImage.MAXTEXT_TPU_JAX_STABLE.value,
docker_image=DockerImage.MAXTEXT_TPU_JAX_STABLE_STACK.value,
test_owner=test_owner.JON_B,
).run()
nightly_tpu = gke_config.get_gke_config(
Expand Down Expand Up @@ -129,7 +129,7 @@
f"export BASE_OUTPUT_PATH=$GCS_OUTPUT; bash end_to_end/{test_scripts_details[0]['script_name']}.sh",
),
cluster=test_scripts_details[0]["cluster"],
docker_image=DockerImage.MAXTEXT_TPU_JAX_STABLE.value,
docker_image=DockerImage.MAXTEXT_TPU_JAX_STABLE_STACK.value,
test_owner=test_owner.ANISHA_M,
).run(gcs_location=shared_gcs_location)
stable_tpu = gke_config.get_gke_config(
Expand All @@ -138,7 +138,7 @@
run_model_cmds=(
f"export BASE_OUTPUT_PATH=$GCS_OUTPUT; bash end_to_end/{test_scripts_details[1]['script_name']}.sh",
),
docker_image=DockerImage.MAXTEXT_TPU_JAX_STABLE.value,
docker_image=DockerImage.MAXTEXT_TPU_JAX_STABLE_STACK.value,
test_owner=test_owner.ANISHA_M,
cluster=test_scripts_details[1]["cluster"],
).run(gcs_location=shared_gcs_location)
Expand Down
2 changes: 1 addition & 1 deletion dags/multipod/maxtext_profiling.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,7 @@
base_output_directory = f"{gcs_bucket.BASE_OUTPUT_DIR}/maxtext_profiling"
dataset_path = gcs_bucket.MAXTEXT_DIR
docker_images = [
(SetupMode.STABLE, DockerImage.MAXTEXT_TPU_JAX_STABLE),
(SetupMode.STABLE, DockerImage.MAXTEXT_TPU_JAX_STABLE_STACK),
(SetupMode.NIGHTLY, DockerImage.MAXTEXT_TPU_JAX_NIGHTLY),
]

Expand Down
2 changes: 1 addition & 1 deletion dags/multipod/maxtext_profiling_vertex_ai_tensorboard.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@
)
dataset_path = gcs_bucket.MAXTEXT_DIR
docker_images = [
(SetupMode.STABLE, DockerImage.MAXTEXT_TPU_JAX_STABLE),
(SetupMode.STABLE, DockerImage.MAXTEXT_TPU_JAX_STABLE_STACK),
(SetupMode.NIGHTLY, DockerImage.MAXTEXT_TPU_JAX_NIGHTLY),
]
test_configs = {
Expand Down
2 changes: 1 addition & 1 deletion dags/multipod/maxtext_trillium_configs_perf.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -29,7 +29,7 @@
"gpt3_175b",
]
DOCKER_IMAGES = [
(SetupMode.STABLE, DockerImage.MAXTEXT_TPU_JAX_STABLE),
(SetupMode.STABLE, DockerImage.MAXTEXT_TPU_JAX_STABLE_STACK),
(SetupMode.NIGHTLY, DockerImage.MAXTEXT_TPU_JAX_NIGHTLY),
]
QUANTIZATION_SWEEP = {"M_QUANTIZATION": ["", "int8"]}
Expand Down
2 changes: 1 addition & 1 deletion dags/multipod/maxtext_v5e_configs_perf.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,7 @@
"llama2_70b",
]
DOCKER_IMAGES = [
(SetupMode.STABLE, DockerImage.MAXTEXT_TPU_JAX_STABLE),
(SetupMode.STABLE, DockerImage.MAXTEXT_TPU_JAX_STABLE_STACK),
(SetupMode.NIGHTLY, DockerImage.MAXTEXT_TPU_JAX_NIGHTLY),
]
QUANTIZATION_SWEEP = {"M_QUANTIZATION": ["", "int8"]}
Expand Down
14 changes: 7 additions & 7 deletions dags/vm_resource.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -251,17 +251,17 @@ class DockerImage(enum.Enum):
"us-central1-docker.pkg.dev/tpu-pytorch-releases/docker/"
f"xla:nightly_3.10_tpuvm_{datetime.datetime.today().strftime('%Y%m%d')}"
)
MAXTEXT_TPU_JAX_STABLE = (
"us-docker.pkg.dev/tpu-prod-env-multipod/maxtext-jax-stable-stack/tpu:"
MAXTEXT_TPU_JAX_STABLE_STACK = (
"us-docker.pkg.dev/tpu-prod-env-multipod/maxtext_jax_stable_stack_0.4.33:"
f"jax0.4.30-rev1-{datetime.datetime.today().strftime('%Y-%m-%d')}"
)
MAXDIFFUSION_TPU_STABLE = (
"us-docker.pkg.dev/tpu-prod-env-multipod/maxdiffusion-jax-stable-stack/tpu:"
f"jax0.4.30-rev1-{datetime.datetime.today().strftime('%Y-%m-%d')}"
MAXDIFFUSION_TPU_JAX_STABLE_STACK = (
"gcr.io/tpu-prod-env-multipod/maxdiffusion_jax_stable_stack_0.4.33:"
parambole marked this conversation as resolved.
Show resolved Hide resolved
f"{datetime.datetime.today().strftime('%Y-%m-%d')}"
)
MAXDIFFUSION_TPU_JAX_NIGHTLY = (
"us-docker.pkg.dev/tpu-prod-env-multipod/maxdiffusion-jax-nightly/tpu:"
f"auto-{datetime.datetime.today().strftime('%Y-%m-%d')}"
"gcr.io/tpu-prod-env-multipod/maxdiffusion_jax_nightly:"
f"{datetime.datetime.today().strftime('%Y-%m-%d')}"
)
MAXTEXT_TPU_JAX_NIGHTLY = (
"gcr.io/tpu-prod-env-multipod/maxtext_jax_nightly:"
Expand Down
Loading