ci: enable model-backed ROCm tests

[howard0su]

[cubic-dev-ai]

3 issues found across 1 file

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name=".github/workflows/ci.yml">

<violation number="1" location=".github/workflows/ci.yml:291">
P3: `LUCE_TEST_MODEL_QWEN3` is not consumed by any current test, so this added environment variable has no effect. Remove it or wire it into a test that loads Qwen3.</violation>

<violation number="2" location=".github/workflows/ci.yml:292">
P3: `LUCE_TEST_MODEL_DEEPSEEK4` is not consumed by any current test, so it does not configure the DeepSeek test. Remove it or change the test to use this variable instead of the separate `DS4_TEST_MODEL`.</violation>

<violation number="3" location=".github/workflows/ci.yml:366">
P2: When `/opt/models` is absent or misconfigured, both model-backed tests can be reported as skipped and the ROCm job still passes. Fail the Strix job before `ctest` when the configured model files are unavailable.</violation>
</file>

<sub>Reply with feedback, questions, or to request a fix.

Re-trigger cubic</sub>

[cubic-dev-ai]

P2: When /opt/models is absent or misconfigured, both model-backed tests can be reported as skipped and the ROCm job still passes. Fail the Strix job before ctest when the configured model files are unavailable.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At .github/workflows/ci.yml, line 366:

<comment>When `/opt/models` is absent or misconfigured, both model-backed tests can be reported as skipped and the ROCm job still passes. Fail the Strix job before `ctest` when the configured model files are unavailable.</comment>

<file context>
@@ -352,12 +356,19 @@ jobs:
           ctest --test-dir "$RUNNER_TEMP/rocmfp-build" \
             --output-on-failure \
             -R 'cuda_pool_shutdown|rocmfp4_reference|rocmfpx_reference|rocmfp4_hip_tail|rocmfpx_mmq|deepseek4_mmid_grouped_cuda|deepseek4_unit|recurrent_snapshot|ChainRollbackPolicy|rocmfp3_mix_registry|rocmfp_mix_slice_matvec|rocmfp_mix_gateup_glu|ds4_mix_registry_teardown|test_model_smoke\.PagedAttention\.|batched_gdn$|concat_transpose$'
+          if [[ "$EXPECTED_HIP_ARCH" == "gfx1151" ]]; then
+            ctest --test-dir "$RUNNER_TEMP/rocmfp-build" \
+              --output-on-failure \
</file context>

Suggested change

	if [[ "$EXPECTED_HIP_ARCH" == "gfx1151" ]]; then
	if [[ "$EXPECTED_HIP_ARCH" == "gfx1151" ]]; then
	test -f "$LUCE_TEST_MODEL_QWEN35"
	test -f "$DS4_TEST_MODEL"

[cubic-dev-ai]

P3: LUCE_TEST_MODEL_DEEPSEEK4 is not consumed by any current test, so it does not configure the DeepSeek test. Remove it or change the test to use this variable instead of the separate DS4_TEST_MODEL.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At .github/workflows/ci.yml, line 292:

<comment>`LUCE_TEST_MODEL_DEEPSEEK4` is not consumed by any current test, so it does not configure the DeepSeek test. Remove it or change the test to use this variable instead of the separate `DS4_TEST_MODEL`.</comment>

<file context>
@@ -287,6 +287,10 @@ jobs:
       EXPECTED_HIP_ARCH: ${{ matrix.arch }}
+      LUCE_TEST_MODEL_QWEN35: /opt/models/Qwen3.6-27B-Q4_K_M.gguf
+      LUCE_TEST_MODEL_QWEN3: /opt/models/Qwen3-0.6B-BF16.gguf
+      LUCE_TEST_MODEL_DEEPSEEK4: /opt/models/DeepSeek-V4-Flash-0731-ROCMFPX-MIX-STRIX.gguf
+      DS4_TEST_MODEL: /opt/models/DeepSeek-V4-Flash-0731-ROCMFPX-MIX-STRIX.gguf
     concurrency:
</file context>

[cubic-dev-ai]

P3: LUCE_TEST_MODEL_QWEN3 is not consumed by any current test, so this added environment variable has no effect. Remove it or wire it into a test that loads Qwen3.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At .github/workflows/ci.yml, line 291:

<comment>`LUCE_TEST_MODEL_QWEN3` is not consumed by any current test, so this added environment variable has no effect. Remove it or wire it into a test that loads Qwen3.</comment>

<file context>
@@ -287,6 +287,10 @@ jobs:
       HIP_VISIBLE_DEVICES: ${{ matrix.device_index }}
       EXPECTED_HIP_ARCH: ${{ matrix.arch }}
+      LUCE_TEST_MODEL_QWEN35: /opt/models/Qwen3.6-27B-Q4_K_M.gguf
+      LUCE_TEST_MODEL_QWEN3: /opt/models/Qwen3-0.6B-BF16.gguf
+      LUCE_TEST_MODEL_DEEPSEEK4: /opt/models/DeepSeek-V4-Flash-0731-ROCMFPX-MIX-STRIX.gguf
+      DS4_TEST_MODEL: /opt/models/DeepSeek-V4-Flash-0731-ROCMFPX-MIX-STRIX.gguf
</file context>