Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
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
11 changes: 9 additions & 2 deletions .github/actions/action.yml
Original file line number Diff line number Diff line change
Expand Up @@ -48,7 +48,9 @@ inputs:
is_ci_workload:
description: "Is CI workload"
required: true

is_merge_group:
description: "Is merge group"
required: true
runs:
using: "composite"
steps:
Expand Down Expand Up @@ -179,7 +181,12 @@ runs:
#!/bin/bash
set -euxo pipefail

if [ "${{ steps.has-run-tests-label.outputs.main }}" == "true" ]; then
if [ "${{ inputs.is_merge_group }}" == "true" ]; then
ARGS=(
--scope mr-github
--n-repeat 1
)
elif [ "${{ steps.has-run-tests-label.outputs.main }}" == "true" ]; then
ARGS=(
--scope mr-github
--enable-lightweight-mode
Expand Down
9 changes: 3 additions & 6 deletions .github/workflows/cicd-main.yml
Original file line number Diff line number Diff line change
Expand Up @@ -18,8 +18,6 @@ on:
- cron: 0 0 * * *
push:
branches:
- dev
- main
- "pull-request/[0-9]+"
- "deploy-release/*"
merge_group:
Expand Down Expand Up @@ -181,7 +179,7 @@ jobs:
cicd-wait-in-queue:
runs-on: ubuntu-latest
needs: [pre-flight, linting]
environment: ${{ needs.pre-flight.outputs.is_merge_group == 'true' && 'merge-gate' || 'test' }}
environment: "test"
if: |
!(needs.pre-flight.outputs.is_ci_workload == 'true'
|| needs.pre-flight.outputs.is_deployment_workflow == 'true'
Expand Down Expand Up @@ -404,7 +402,6 @@ jobs:
success()
|| needs.pre-flight.outputs.is_ci_workload == 'true'
|| needs.pre-flight.outputs.force_run_all == 'true'
|| needs.pre-flight.outputs.is_merge_group == 'true'
)
&& !cancelled()
outputs:
Expand Down Expand Up @@ -442,7 +439,7 @@ jobs:
id: main
env:
HAS_RUN_TESTS_LABEL: ${{ steps.has-run-tests-label.outputs.main }}
HAS_RUN_FUNCTIONAL_TESTS_LABEL: ${{ steps.has-run-functional-tests-label.outputs.main }}
HAS_RUN_FUNCTIONAL_TESTS_LABEL: ${{ steps.has-run-functional-tests-label.outputs.main == 'true' || needs.pre-flight.outputs.is_merge_group == 'true' }}
run: |
export PYTHONPATH=$(pwd)

Expand Down Expand Up @@ -505,7 +502,6 @@ jobs:
|| needs.pre-flight.outputs.is_ci_workload == 'true'
|| needs.pre-flight.outputs.force_run_all == 'true'
)
&& needs.pre-flight.outputs.is_merge_group == 'false'
&& !cancelled()
steps:
- name: Checkout
Expand All @@ -521,6 +517,7 @@ jobs:
PAT: ${{ secrets.PAT }}
container-image: ${{ env.container-registry }}/megatron-lm:${{ github.sha }}
is_ci_workload: ${{ needs.pre-flight.outputs.is_ci_workload }}
is_merge_group: ${{ needs.pre-flight.outputs.is_merge_group }}

Nemo_CICD_Test:
needs:
Expand Down
78 changes: 39 additions & 39 deletions tests/test_utils/recipes/gpt.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -110,14 +110,14 @@ products:
- test_case: [gpt3_mcore_te_tp1_pp1_dist_optimizer_no_mmap_bin_files]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
- test_case: [gpt3_mcore_te_tp1_pp1_dist_optimizer_fim_dataset]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- test_case: [gpt3_mcore_te_tp1_pp1_resume_torch_dist_dist_optimizer]
products:
Expand All @@ -129,7 +129,7 @@ products:
- test_case: [gpt3_mcore_te_tp1_pp1_resume_torch_dist_uniform_full_recompute]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
Expand All @@ -142,28 +142,28 @@ products:
- test_case: [gpt3_mcore_te_tp1_pp2_resume_torch_dist_rope_embeddings]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
# - environment: [lts]
# scope: [nightly] # outdated TE: #501
- test_case: [gpt3_mcore_te_tp1_pp2_resume_torch_dist_rope_embeddings_interleaved_no_fusion]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
- test_case: [gpt3_mcore_te_tp1_pp4_resume_torch_dist_disable_bias_linear]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
# - environment: [lts]
# scope: [nightly] # non-determinism: #436
- test_case: [gpt3_mcore_te_tp1_pp4_resume_torch_dist_persistent_disable_bias_linear]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
Expand All @@ -177,28 +177,28 @@ products:
- test_case: [gpt3_mcore_te_tp1_pp4_resume_torch_dist_untie_embeddings_and_outputs]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
- test_case: [gpt3_mcore_te_tp1_pp4_vp1]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
- test_case: [gpt3_mcore_te_tp1_pp4_vp1_resume_torch_decoupled_lr]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
- test_case: [gpt3_mcore_te_tp1_pp4_vp1_resume_torch_dist_calculate_per_token_loss]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
Expand All @@ -212,7 +212,7 @@ products:
- test_case: [gpt3_mcore_te_tp1_pp4_vp1_resume_torch_dist_dist_optimizer_overlap_grad_reduce]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
Expand Down Expand Up @@ -247,98 +247,98 @@ products:
- test_case: [gpt3_mcore_te_tp1_pp4_vp1_uneven_pipeline]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
- test_case: [gpt3_mcore_te_tp1_pp4_vp2_account_for_embedding_loss_in_pipeline_split]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
- test_case: [gpt3_mcore_te_tp2_pp1_resume_torch_dist_cp2_nondeterministic]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
- test_case: [gpt3_mcore_te_tp2_pp1_resume_torch_dist_multi_dist_optimizer_instances]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
- test_case: [gpt3_mcore_te_tp2_pp2_cp2]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- test_case: [gpt3_mcore_te_tp2_pp2_cp2_etp4_dp_last]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- test_case: [gpt3_mcore_te_tp2_pp2_cp2_calculate_per_token_loss]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- test_case: [gpt3_mcore_te_tp2_pp2_cp2_etp4_calculate_per_token_loss_dp_last]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- test_case: [gpt3_mcore_te_tp2_pp2_cp2_nondeterministic]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
- test_case: [gpt3_mcore_te_tp2_pp2_cp2_etp4_nondeterministic_dp_last]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
- test_case: [gpt3_mcore_te_tp2_pp2_cp2_calculate_per_token_loss_nondeterministic]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
- test_case: [gpt3_mcore_te_tp2_pp2_cp2_etp4_calculate_per_token_loss_nondeterministic_dp_last]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
- test_case: [gpt3_mcore_te_tp2_pp2_cross_entropy_loss_fusion]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]

- test_case: [gpt3_mcore_te_tp2_pp2_resume_torch_dist_cp2_nondeterministic]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
- test_case: [gpt3_mcore_te_tp2_pp2_resume_torch_dist_cross_entropy_loss_fusion]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
Expand All @@ -350,26 +350,26 @@ products:
- test_case: [gpt3_mcore_te_tp2_pp2_mla]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- test_case: [gpt3_mcore_te_tp2_pp2_resume_torch_dist_ddp_average_in_collective]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
- test_case: [gpt3_mcore_te_tp2_pp2_resume_torch_dist_defer_embedding_wgrad_compute]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
- test_case: [gpt3_mcore_te_tp2_pp2_resume_torch_dist]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
Expand All @@ -383,28 +383,28 @@ products:
- test_case: [gpt3_mcore_te_tp2_pp2_resume_torch_dist_reshard_1x4xNone]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
- test_case: [gpt3_mcore_te_tp4_pp1_resume_torch_dist_dist_optimizer_overlap_grad_reduce]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
- test_case: [gpt3_mcore_te_tp4_pp1_resume_torch_dist_qk_layernorm_test_mode]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
- test_case: [gpt3_mcore_tp2_pp2_uninstall_te]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
Expand All @@ -425,14 +425,14 @@ products:
- test_case: [gpt3_mcore_te_tp2_zp_z3_resume_fsdp_dtensor]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
# - environment: [lts]
# scope: [nightly]
- test_case: [gpt3_mcore_te_tp2_pp1_modelopt_distill_resume]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
# - environment: [lts]
# scope: [nightly] # Outdated: #502
Expand Down Expand Up @@ -469,7 +469,7 @@ products:
- test_case: [gpt3_mcore_te_tp4_pp1_dist_optimizer_overlap_grad_reduce_param_gather]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
Expand All @@ -483,7 +483,7 @@ products:
- test_case: [gpt3_mcore_te_tp4_pp2_resume_torch_dist_reshard_8x1xNone]
products:
- environment: [dev]
scope: [mr]
scope: [mr, mr-github]
platforms: [dgx_h100]
- environment: [lts]
scope: [nightly]
Expand Down
Loading
Loading