Skip to content

[ROCm][CI] Mirror the three TurboQuant evaluation groups on MI355 #58282

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

Merged
AndreasKaratzas merged 2 commits into from
Sep 23, 2026
Merged

[ROCm][CI] Mirror the three TurboQuant evaluation groups on MI355 #58282

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
b594e57
[ROCm][CI] Mirror three TurboQuant evaluation groups on MI355
AndreasKaratzas Sep 23, 2026
825bac9
Merge branch 'main' into akaratza_mi355_turboquant_mirrors
AndreasKaratzas Sep 23, 2026
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
78 changes: 78 additions & 0 deletions .buildkite/test-amd.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3869,6 +3869,84 @@ steps:
- uv pip install --system 'evalscope==1.10.0'
- pytest -s -v evals/qwen4_exp/test_accuracy.py --config-list-file=configs/models-rocm.txt

- label: ":amd: (MI355) LM Eval TurboQuant k8v4"
timeout_in_minutes: 30
mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
dind: false
agent_pool: mi355_1
num_gpus: 1
optional: true
working_dir: "/vllm-workspace/tests"
source_file_dependencies:
- tests/evals/gsm8k/
- tests/utils.py
- tests/conftest.py
- vllm/model_executor/layers/quantization/turboquant/
- vllm/v1/attention/backends/turboquant_attn.py
- vllm/v1/attention/backends/triton_attn.py
- vllm/v1/attention/ops/triton_unified_attention.py
- vllm/v1/attention/selector.py
- vllm/v1/attention/ops/triton_turboquant_decode.py
- vllm/v1/attention/ops/triton_turboquant_store.py
- vllm/v1/attention/ops/flydsl_turboquant_decode.py
- vllm/v1/attention/ops/flydsl_kernels/
- vllm/v1/attention/ops/turboquant_soa/
- vllm/platforms/rocm.py
commands:
- pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=evals/gsm8k/configs/models-turboquant-k8v4-rocm.txt

- label: ":amd: (MI355) LM Eval TurboQuant t4nc"
timeout_in_minutes: 30
mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
dind: false
agent_pool: mi355_1
num_gpus: 1
optional: true
working_dir: "/vllm-workspace/tests"
source_file_dependencies:
- tests/evals/gsm8k/
- tests/utils.py
- tests/conftest.py
- vllm/model_executor/layers/quantization/turboquant/
- vllm/v1/attention/backends/turboquant_attn.py
- vllm/v1/attention/backends/triton_attn.py
- vllm/v1/attention/ops/triton_unified_attention.py
- vllm/v1/attention/selector.py
- vllm/v1/attention/ops/triton_turboquant_decode.py
- vllm/v1/attention/ops/triton_turboquant_store.py
- vllm/v1/attention/ops/flydsl_turboquant_decode.py
- vllm/v1/attention/ops/flydsl_kernels/
- vllm/v1/attention/ops/turboquant_soa/
- vllm/platforms/rocm.py
commands:
- pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=evals/gsm8k/configs/models-turboquant-t4nc-rocm.txt

- label: ":amd: (MI355) LM Eval TurboQuant k3v4nc"
timeout_in_minutes: 30
mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
dind: false
agent_pool: mi355_1
num_gpus: 1
optional: true
working_dir: "/vllm-workspace/tests"
source_file_dependencies:
- tests/evals/gsm8k/
- tests/utils.py
- tests/conftest.py
- vllm/model_executor/layers/quantization/turboquant/
- vllm/v1/attention/backends/turboquant_attn.py
- vllm/v1/attention/backends/triton_attn.py
- vllm/v1/attention/ops/triton_unified_attention.py
- vllm/v1/attention/selector.py
- vllm/v1/attention/ops/triton_turboquant_decode.py
- vllm/v1/attention/ops/triton_turboquant_store.py
- vllm/v1/attention/ops/flydsl_turboquant_decode.py
- vllm/v1/attention/ops/flydsl_kernels/
- vllm/v1/attention/ops/turboquant_soa/
- vllm/platforms/rocm.py
commands:
- pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=evals/gsm8k/configs/models-turboquant-k3v4nc-rocm.txt

- label: ":amd: (MI355) LM Eval Small Models Distributed"
timeout_in_minutes: 55
mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
Expand Down
84 changes: 84 additions & 0 deletions .buildkite/test_areas/lm_eval.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -646,6 +646,34 @@ steps:
- vllm/v1/attention/ops/triton_turboquant_store.py
commands:
- pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=evals/gsm8k/configs/models-turboquant-k8v4.txt
mirror:
amd:
label: ":amd: (MI355) LM Eval TurboQuant k8v4"
dind: false
device: mi355_1
num_devices: 1
timeout_in_minutes: 30
working_dir: /vllm-workspace/tests
depends_on:
- image-build-amd
source_file_dependencies:
- .buildkite/test_areas/lm_eval.yaml
- tests/evals/gsm8k/
- tests/utils.py
- tests/conftest.py
- vllm/model_executor/layers/quantization/turboquant/
- vllm/v1/attention/backends/turboquant_attn.py
- vllm/v1/attention/backends/triton_attn.py
- vllm/v1/attention/ops/triton_unified_attention.py
- vllm/v1/attention/selector.py
- vllm/v1/attention/ops/triton_turboquant_decode.py
- vllm/v1/attention/ops/triton_turboquant_store.py
- vllm/v1/attention/ops/flydsl_turboquant_decode.py
- vllm/v1/attention/ops/flydsl_kernels/
- vllm/v1/attention/ops/turboquant_soa/
- vllm/platforms/rocm.py
commands:
- pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=evals/gsm8k/configs/models-turboquant-k8v4-rocm.txt

- label: ":nvidia: (H200 MIG 18GB) LM Eval TurboQuant t4nc"
key: lm-eval-turboquant-t4nc
Expand All @@ -658,6 +686,34 @@ steps:
- vllm/v1/attention/ops/triton_turboquant_store.py
commands:
- pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=evals/gsm8k/configs/models-turboquant-t4nc.txt
mirror:
amd:
label: ":amd: (MI355) LM Eval TurboQuant t4nc"
dind: false
device: mi355_1
num_devices: 1
timeout_in_minutes: 30
working_dir: /vllm-workspace/tests
depends_on:
- image-build-amd
source_file_dependencies:
- .buildkite/test_areas/lm_eval.yaml
- tests/evals/gsm8k/
- tests/utils.py
- tests/conftest.py
- vllm/model_executor/layers/quantization/turboquant/
- vllm/v1/attention/backends/turboquant_attn.py
- vllm/v1/attention/backends/triton_attn.py
- vllm/v1/attention/ops/triton_unified_attention.py
- vllm/v1/attention/selector.py
- vllm/v1/attention/ops/triton_turboquant_decode.py
- vllm/v1/attention/ops/triton_turboquant_store.py
- vllm/v1/attention/ops/flydsl_turboquant_decode.py
- vllm/v1/attention/ops/flydsl_kernels/
- vllm/v1/attention/ops/turboquant_soa/
- vllm/platforms/rocm.py
commands:
- pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=evals/gsm8k/configs/models-turboquant-t4nc-rocm.txt

- label: ":nvidia: (H200 MIG 18GB) LM Eval TurboQuant k3v4nc"
key: lm-eval-turboquant-k3v4nc
Expand All @@ -670,6 +726,34 @@ steps:
- vllm/v1/attention/ops/triton_turboquant_store.py
commands:
- pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=evals/gsm8k/configs/models-turboquant-k3v4nc.txt
mirror:
amd:
label: ":amd: (MI355) LM Eval TurboQuant k3v4nc"
dind: false
device: mi355_1
num_devices: 1
timeout_in_minutes: 30
working_dir: /vllm-workspace/tests
depends_on:
- image-build-amd
source_file_dependencies:
- .buildkite/test_areas/lm_eval.yaml
- tests/evals/gsm8k/
- tests/utils.py
- tests/conftest.py
- vllm/model_executor/layers/quantization/turboquant/
- vllm/v1/attention/backends/turboquant_attn.py
- vllm/v1/attention/backends/triton_attn.py
- vllm/v1/attention/ops/triton_unified_attention.py
- vllm/v1/attention/selector.py
- vllm/v1/attention/ops/triton_turboquant_decode.py
- vllm/v1/attention/ops/triton_turboquant_store.py
- vllm/v1/attention/ops/flydsl_turboquant_decode.py
- vllm/v1/attention/ops/flydsl_kernels/
- vllm/v1/attention/ops/turboquant_soa/
- vllm/platforms/rocm.py
commands:
- pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=evals/gsm8k/configs/models-turboquant-k3v4nc-rocm.txt

- label: ":nvidia: (H200 MIG 18GB) LM Eval TurboQuant t3nc"
key: lm-eval-turboquant-t3nc
Expand Down
6 changes: 6 additions & 0 deletions tests/evals/gsm8k/configs/Qwen3-4B-TQ-k3v4nc-ROCm.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,6 @@
model_name: "Qwen/Qwen3-4B"
accuracy_threshold: 0.78
num_questions: 1319
num_fewshot: 5
# Triton serves the unquantized boundary layers with TurboQuant-compatible layout.
server_args: "--kv-cache-dtype turboquant_k3v4_nc --max-model-len 4096 --attention-backend TRITON_ATTN"
6 changes: 6 additions & 0 deletions tests/evals/gsm8k/configs/Qwen3-4B-TQ-k8v4-ROCm.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,6 @@
model_name: "Qwen/Qwen3-4B"
accuracy_threshold: 0.80
num_questions: 1319
num_fewshot: 5
# Triton serves the unquantized boundary layers with TurboQuant-compatible layout.
server_args: "--kv-cache-dtype turboquant_k8v4 --max-model-len 4096 --attention-backend TRITON_ATTN"
6 changes: 6 additions & 0 deletions tests/evals/gsm8k/configs/Qwen3-4B-TQ-t4nc-ROCm.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,6 @@
model_name: "Qwen/Qwen3-4B"
accuracy_threshold: 0.80
num_questions: 1319
num_fewshot: 5
# Triton serves the unquantized boundary layers with TurboQuant-compatible layout.
server_args: "--kv-cache-dtype turboquant_4bit_nc --max-model-len 4096 --attention-backend TRITON_ATTN"
1 change: 1 addition & 0 deletions tests/evals/gsm8k/configs/models-turboquant-k3v4nc-rocm.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
Qwen3-4B-TQ-k3v4nc-ROCm.yaml
1 change: 1 addition & 0 deletions tests/evals/gsm8k/configs/models-turboquant-k8v4-rocm.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
Qwen3-4B-TQ-k8v4-ROCm.yaml
1 change: 1 addition & 0 deletions tests/evals/gsm8k/configs/models-turboquant-t4nc-rocm.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
Qwen3-4B-TQ-t4nc-ROCm.yaml
Loading