From 433cfa93d7f54b5f2666caed3263e56935c62125 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 7 May 2026 06:19:36 +0000
Subject: [PATCH] CI(consolidated): skip false-positive patches in runtime
 ledger; drop job-level continue-on-error

Two cleanups derived from review of the matrix output:

1. Skip false-positive zero-arg patches in the runtime ledger.
   Three patches have all-defaulted signatures but require either
   runtime args or real CUDA, so calling them in isolation produces
   a meaningless failure:
     - patch_linear_scaling: defaults are None placeholders;
       body starts with `assert rope_module is not None` etc.
     - patch_llama_rope_scaling: same shape.
     - patch_unsloth_smart_gradient_checkpointing: legitimately
       allocates CUDA tensors via aten::empty.memory_format inside
       initialize_unsloth_gradient_checkpointing(); the torch.cuda.*
       Python spoof can't intercept that at the dispatcher level.
   Add NEEDS_PRECONDITION = {...} to the shim and skip those by name.
   Symbol presence is still verified via REQUIRED.

2. Drop the job-level `continue-on-error: true`.
   Previously the cell reported SUCCESS even when steps failed, which
   made the PR check UI lie. Real failures now turn the cell red.
   Per-step `continue-on-error: true` stays so a single failed step
   does not cascade and skip the rest of the ledger.

Three other failures the matrix surfaced are addressed by separate PRs
to source:
  - unslothai/unsloth#5319 (patch_fast_lora missing import,
    patch_sft_trainer_tokenizer Union NameError, openenv OSError)
  - unslothai/unsloth-zoo#628 (skip MoE coverage on older transformers)
---
 .github/workflows/consolidated-tests-ci.yml | 38 ++++++++++++++++++---
 1 file changed, 33 insertions(+), 5 deletions(-)

diff --git a/.github/workflows/consolidated-tests-ci.yml b/.github/workflows/consolidated-tests-ci.yml
index 005b034b22..14cd75f4fa 100644
--- a/.github/workflows/consolidated-tests-ci.yml
+++ b/.github/workflows/consolidated-tests-ci.yml
@@ -82,9 +82,14 @@ jobs:
     name: "Consolidated CPU (${{ matrix.combo.label }})"
     runs-on: ubuntu-latest
     timeout-minutes: 35
-    # First-pass posture: surface results in the PR check UI without blocking
-    # merge. Flip to false (or delete this line) once every cell is green.
-    continue-on-error: true
+    # NOTE: previously had a job-level `continue-on-error: true` so cells
+    # reported SUCCESS even when individual steps failed. That made the
+    # PR check UI lie. The job-level flag is removed; per-step
+    # `continue-on-error: true` remains so a single failed step does not
+    # cascade and skip the rest of the ledger. Real failures now show up
+    # as a red cell on the PR. Patches with known preconditions are now
+    # explicitly skipped via NEEDS_PRECONDITION in the runtime check
+    # shim, not silenced via blanket continue-on-error.
     env:
       UNSLOTH_ZOO_REF: ${{ inputs.unsloth_zoo_ref || 'main' }}
       MATRIX_TRANSFORMERS_SPEC: ${{ matrix.combo.transformers_spec }}
@@ -434,9 +439,26 @@ jobs:
               "patch_unsloth_smart_gradient_checkpointing",
               "patch_gradient_accumulation_fix",
           }
+          # Patches whose signature looks zero-arg (`()` or all-defaulted)
+          # but which actually require either runtime args or real CUDA.
+          # Calling these in isolation is meaningless, so skip the
+          # invocation. Symbol presence (REQUIRED above) is still verified.
+          #   patch_linear_scaling / patch_llama_rope_scaling: defaults are
+          #     None placeholders; the bodies start with
+          #     `assert <param> is not None`.
+          #   patch_unsloth_smart_gradient_checkpointing: legitimately
+          #     allocates CUDA tensors via aten::empty.memory_format inside
+          #     initialize_unsloth_gradient_checkpointing(); the
+          #     torch.cuda.* spoof can't intercept that at the dispatcher
+          #     level.
+          NEEDS_PRECONDITION = {
+              "patch_linear_scaling",
+              "patch_llama_rope_scaling",
+              "patch_unsloth_smart_gradient_checkpointing",
+          }
 
           def test_zero_arg_patch_invocations():
-              ok, fail, args, miss_imports = 0, [], [], {}
+              ok, fail, args, skipped, miss_imports = 0, [], [], [], {}
               seen_required = set()
               for mod_name in MODULES:
                   try:
@@ -459,6 +481,10 @@ jobs:
                           need = []
                       if need:
                           args.append((mod_name, name, need)); continue
+                      if name in NEEDS_PRECONDITION:
+                          skipped.append(f"{mod_name}.{name}")
+                          print(f"  SKIP {mod_name}.{name} (needs precondition / CUDA)")
+                          continue
                       try:
                           fn()
                           ok += 1
@@ -466,10 +492,12 @@ jobs:
                       except Exception as e:
                           fail.append((mod_name, name, type(e).__name__, str(e)[:200]))
                           print(f"  FAIL {mod_name}.{name} -> {type(e).__name__}: {str(e)[:200]}")
-              print(f"\nzero-arg patch_*: ok={ok} fail={len(fail)}")
+              print(f"\nzero-arg patch_*: ok={ok} fail={len(fail)} skipped={len(skipped)}")
               print(f"arg-required patch_* (skipped, listed for review): {len(args)}")
               for m, n, r in args:
                   print(f"    needs={r}: {m}.{n}")
+              if skipped:
+                  print(f"explicitly skipped (needs precondition / CUDA): {skipped}")
               if miss_imports:
                   print("\nmodules failed to import (skipped):")
                   for k, v in miss_imports.items():