iHore/vLLM
8
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

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

Browse Source 
Signed-off-by: Andreas Karatzas <akaratza@amd.com>
Co-authored-by: OpenAI Codex <noreply@openai.com>
This commit is contained in:
Andreas Karatzas
2026-09-23 02:37:23 -05:00
committed by GitHub
co-authored by OpenAI Codex
parent 9127170b88
commit a9f07d0bc5
8 changed files with 183 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.buildkite/test-amd.yaml
+78
View File
@@ -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]
.buildkite/test_areas/lm_eval.yaml
+84
View File
@@ -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
@@ -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
@@ -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
tests/evals/gsm8k/configs/Qwen3-4B-TQ-k3v4nc-ROCm.yaml
+6
View File
@@ -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"
tests/evals/gsm8k/configs/Qwen3-4B-TQ-k8v4-ROCm.yaml
+6
View File
@@ -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"
tests/evals/gsm8k/configs/Qwen3-4B-TQ-t4nc-ROCm.yaml
+6
View File
@@ -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"
tests/evals/gsm8k/configs/models-turboquant-k3v4nc-rocm.txt
+1
View File
@@ -0,0 +1 @@
Qwen3-4B-TQ-k3v4nc-ROCm.yaml
tests/evals/gsm8k/configs/models-turboquant-k8v4-rocm.txt
+1
View File
@@ -0,0 +1 @@
Qwen3-4B-TQ-k8v4-ROCm.yaml
tests/evals/gsm8k/configs/models-turboquant-t4nc-rocm.txt
+1
View File
@@ -0,0 +1 @@
Qwen3-4B-TQ-t4nc-ROCm.yaml