Merge branch 'main' into ahmadki/simulated-user-rec

Signed-off-by: Ahmad Kiswani <[email protected]>

Author: ahmadki

.github/ISSUE_TEMPLATE/bug_report.md

@@ -20,7 +20,7 @@ on:
 
 jobs:
   cherry-pick:
-    uses: NVIDIA-NeMo/FW-CI-templates/.github/workflows/_cherry_pick.yml@v0.31.0
+    uses: NVIDIA-NeMo/FW-CI-templates/.github/workflows/_cherry_pick.yml@v0.63.0
     secrets:
       PAT: ${{ secrets.PAT }}
       SLACK_WEBHOOK_ADMIN: ${{ secrets.SLACK_WEBHOOK_ADMIN }}

.github/ISSUE_TEMPLATE/feature_request.md

@@ -288,7 +288,7 @@ jobs:
           is_fork_pr: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.repo.full_name != github.event.pull_request.base.repo.full_name }}
 
   CI_QA_Gate:
-    name: CI quality check
+    name: "CI quality check${{ needs.pre-flight.outputs.test_level == 'none' && ' (No tests run: Label CI:L*)' || '' }}"
     if: always()
     runs-on: ubuntu-latest
     needs:
@@ -310,17 +310,14 @@ jobs:
               needs.lint-check.result == 'success' &&
               (needs.pr-branch-up-to-date-check.result == 'success' || needs.pr-branch-up-to-date-check.result == 'skipped') &&
               (
-                needs.pre-flight.outputs.test_level == 'none' ||
+                needs.pre-flight.outputs.test_level != 'none' &&
+                needs.sphinx-build.result == 'success' &&
+                needs.build-container.result == 'success' &&
                 (
-                  needs.pre-flight.outputs.test_level != 'none' &&
-                  needs.sphinx-build.result == 'success' &&
-                  needs.build-container.result == 'success' &&
                   (
-                    (
-                      (needs.cicd-doc-tests.result == 'skipped' || needs.cicd-doc-tests.result == 'success') &&
-                      (needs.cicd-unit-tests.result == 'skipped' || needs.cicd-unit-tests.result == 'success') &&
-                      (needs.cicd-functional-tests.result == 'skipped' || needs.cicd-functional-tests.result == 'success')
-                    )
+                    needs.cicd-doc-tests.result == 'success' &&
+                    (needs.cicd-unit-tests.result == 'skipped' || needs.cicd-unit-tests.result == 'success') &&
+                    (needs.cicd-functional-tests.result == 'skipped' || needs.cicd-functional-tests.result == 'success')
                   )
                 )
               )

.github/workflows/cherry-pick-release-commit.yml

@@ -1,6 +1,6 @@
 # Adding to .gitignore helps reduce the size of your working_dir
 
-.git
+/.git
 *.out
 *.log
 *.tar
@@ -17,7 +17,6 @@ dist/
 *.vscode/
 release_run*
 ckpts/
-3rdparty/vllm
 
 # Test
 coverage.json
@@ -33,7 +32,7 @@ uv_cache/
 hf_home/
 hf_datasets_cache/
 *logs/
-datasets/
+/datasets/
 docker/*
 !docker/Dockerfile
 !docker/Dockerfile.ngc_pytorch

.github/workflows/cicd-main.yml

@@ -1,7 +1,7 @@
 [submodule "3rdparty/Megatron-LM"]
 	path = 3rdparty/Megatron-LM-workspace/Megatron-LM
 	url = https://github.com/terrykong/Megatron-LM.git
-	branch = yuya/nemo-rl-use-2
+	branch = guyueh/yuya/nemo-rl-use-2
 	shallow = true
 [submodule "3rdparty/Megatron-Bridge"]
 	path = 3rdparty/Megatron-Bridge-workspace/Megatron-Bridge

.gitignore

@@ -46,3 +46,38 @@ repos:
         require_serial: true
         additional_dependencies: []
         minimum_pre_commit_version: "2.9.2"
+
+  # This pre-commit hook ensures that the config file is minimized and reflects exactly what you
+  # intend to merge. Without it, you might run experiments with one config, but when merging upstream,
+  # the config could silently fall back to the base defaults—resulting in different hyperparameters.
+  #
+  # For example, we’ve seen cases where an SFT recipe runs without a custom chat_template. When merged,
+  # it unexpectedly picks up the default recommended chat_template from upstream, which doesn’t match
+  # the original experiment setup.
+  #
+  # If this check is disruptive, you can disable the pre-commit hook locally. However, before a recipe
+  # is accepted upstream, we expect the config to be minimized.
+  - repo: local
+    hooks:
+      - id: configs-minimize-check-llm
+        name: minimize-check llm recipes
+        language: system
+        pass_filenames: false
+        entry: bash
+        args:
+          - -lc
+          - |
+            set -euo pipefail
+            base="examples/configs/dpo.yaml"; for f in examples/configs/recipes/llm/dpo-*.yaml; do [ -e "$f" ] && ./tools/config_cli.py minimize-check "$base" "$f"; done
+            base="examples/configs/grpo_math_1B.yaml"; for f in examples/configs/recipes/llm/grpo-*.yaml; do [ -e "$f" ] && ./tools/config_cli.py minimize-check "$base" "$f"; done
+            base="examples/configs/sft.yaml"; for f in examples/configs/recipes/llm/sft-*.yaml; do [ -e "$f" ] && ./tools/config_cli.py minimize-check "$base" "$f"; done
+      - id: configs-minimize-check-vlm
+        name: minimize-check vlm recipes
+        language: system
+        pass_filenames: false
+        entry: bash
+        args:
+          - -lc
+          - |
+            set -euo pipefail
+            base="examples/configs/vlm_grpo_3B.yaml"; for f in examples/configs/recipes/vlm/vlm_grpo-*.yaml; do [ -e "$f" ] && ./tools/config_cli.py minimize-check "$base" "$f"; done

.gitmodules

@@ -33,7 +33,7 @@
     "packaging",
     "tensorboard>=2.19.0",
     "torch",
-    "transformers>=4.51.3",
+    "transformers>=4.55.0",
     "typing-extensions",
     "rich",
     "wandb>=0.19.10",