Skip to content

Add distributed test cases #4161

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

Open
littlegy wants to merge 9 commits into
base: main
Choose a base branch
from
Open

Add distributed test cases #4161

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
25a31d5
TEST: update distributed test
littlegy Nov 24, 2025
3f6f989
add proxy test
littlegy Nov 25, 2025
b9ca337
update deep test
littlegy Nov 28, 2025
3b26e42
rm turbomind test
littlegy Nov 28, 2025
c526ff9
Merge branch 'InternLM:main' into multi-node-test
zhulinJulia24 Dec 1, 2025
beace57
update backend
littlegy Dec 1, 2025
70a4c9a
Refactor test configuration
littlegy Dec 5, 2025
c0b25e2
update config
littlegy Dec 5, 2025
241bc98
Merge branch 'InternLM:main' into multi-node-test
zhulinJulia24 Dec 5, 2025
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
16 changes: 8 additions & 8 deletions autotest/benchmark/test_longtext_performance.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,51 +5,51 @@

@pytest.mark.gpu_num_1
@pytest.mark.flaky(reruns=0)
@pytest.mark.parametrize('run_config', get_benchmark_model_list(tp_num=1, kvint_list=[4, 8], is_longtext=True))
@pytest.mark.parametrize('run_config', get_benchmark_model_list(parallel_config=1, kvint_list=[4, 8], is_longtext=True))
def test_longtext_tp1(config, run_id, run_config, worker_id):
result, msg = longtext_throughput_test(config,
run_id,
run_config,
cuda_prefix=get_cuda_prefix_by_workerid(worker_id, tp_num=1),
cuda_prefix=get_cuda_prefix_by_workerid(worker_id, parallel_config=1),
worker_id=worker_id)

assert result, msg


@pytest.mark.gpu_num_2
@pytest.mark.flaky(reruns=0)
@pytest.mark.parametrize('run_config', get_benchmark_model_list(tp_num=2, kvint_list=[4, 8], is_longtext=True))
@pytest.mark.parametrize('run_config', get_benchmark_model_list(parallel_config=2, kvint_list=[4, 8], is_longtext=True))
def test_longtext_tp2(config, run_id, run_config, worker_id):
result, msg = longtext_throughput_test(config,
run_id,
run_config,
cuda_prefix=get_cuda_prefix_by_workerid(worker_id, tp_num=2),
cuda_prefix=get_cuda_prefix_by_workerid(worker_id, parallel_config=2),
worker_id=worker_id)

assert result, msg


@pytest.mark.gpu_num_4
@pytest.mark.flaky(reruns=0)
@pytest.mark.parametrize('run_config', get_benchmark_model_list(tp_num=4, kvint_list=[4, 8], is_longtext=True))
@pytest.mark.parametrize('run_config', get_benchmark_model_list(parallel_config=4, kvint_list=[4, 8], is_longtext=True))
def test_longtext_tp4(config, run_id, run_config, worker_id):
result, msg = longtext_throughput_test(config,
run_id,
run_config,
cuda_prefix=get_cuda_prefix_by_workerid(worker_id, tp_num=4),
cuda_prefix=get_cuda_prefix_by_workerid(worker_id, parallel_config=4),
worker_id=worker_id)

assert result, msg


@pytest.mark.gpu_num_8
@pytest.mark.flaky(reruns=0)
@pytest.mark.parametrize('run_config', get_benchmark_model_list(tp_num=8, kvint_list=[4, 8], is_longtext=True))
@pytest.mark.parametrize('run_config', get_benchmark_model_list(parallel_config=8, kvint_list=[4, 8], is_longtext=True))
def test_longtext_tp8(config, run_id, run_config, worker_id):
result, msg = longtext_throughput_test(config,
run_id,
run_config,
cuda_prefix=get_cuda_prefix_by_workerid(worker_id, tp_num=8),
cuda_prefix=get_cuda_prefix_by_workerid(worker_id, parallel_config=8),
worker_id=worker_id)

assert result, msg
16 changes: 8 additions & 8 deletions autotest/benchmark/test_prefixcache_performance.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,51 +5,51 @@

@pytest.mark.gpu_num_1
@pytest.mark.flaky(reruns=0)
@pytest.mark.parametrize('run_config', get_benchmark_model_list(tp_num=1, kvint_list=[4, 8]))
@pytest.mark.parametrize('run_config', get_benchmark_model_list(parallel_config=1, kvint_list=[4, 8]))
def test_prefixcache_throughput_tp1(config, run_id, run_config, worker_id):
result, msg = prefixcache_throughput_test(config,
run_id,
run_config,
cuda_prefix=get_cuda_prefix_by_workerid(worker_id, tp_num=1),
cuda_prefix=get_cuda_prefix_by_workerid(worker_id, parallel_config=1),
worker_id=worker_id)

assert result, msg


@pytest.mark.gpu_num_2
@pytest.mark.flaky(reruns=0)
@pytest.mark.parametrize('run_config', get_benchmark_model_list(tp_num=2, kvint_list=[4, 8]))
@pytest.mark.parametrize('run_config', get_benchmark_model_list(parallel_config=2, kvint_list=[4, 8]))
def test_prefixcache_throughput_tp2(config, run_id, run_config, worker_id):
result, msg = prefixcache_throughput_test(config,
run_id,
run_config,
cuda_prefix=get_cuda_prefix_by_workerid(worker_id, tp_num=2),
cuda_prefix=get_cuda_prefix_by_workerid(worker_id, parallel_config=2),
worker_id=worker_id)

assert result, msg


@pytest.mark.gpu_num_4
@pytest.mark.flaky(reruns=0)
@pytest.mark.parametrize('run_config', get_benchmark_model_list(tp_num=4, kvint_list=[4, 8]))
@pytest.mark.parametrize('run_config', get_benchmark_model_list(parallel_config=4, kvint_list=[4, 8]))
def test_prefixcache_throughput_tp4(config, run_id, run_config, worker_id):
result, msg = prefixcache_throughput_test(config,
run_id,
run_config,
cuda_prefix=get_cuda_prefix_by_workerid(worker_id, tp_num=4),
cuda_prefix=get_cuda_prefix_by_workerid(worker_id, parallel_config=4),
worker_id=worker_id)

assert result, msg


@pytest.mark.gpu_num_8
@pytest.mark.flaky(reruns=0)
@pytest.mark.parametrize('run_config', get_benchmark_model_list(tp_num=4, kvint_list=[4, 8]))
@pytest.mark.parametrize('run_config', get_benchmark_model_list(parallel_config=4, kvint_list=[4, 8]))
def test_prefixcache_throughput_tp8(config, run_id, run_config, worker_id):
result, msg = prefixcache_throughput_test(config,
run_id,
run_config,
cuda_prefix=get_cuda_prefix_by_workerid(worker_id, tp_num=8),
cuda_prefix=get_cuda_prefix_by_workerid(worker_id, parallel_config=8),
worker_id=worker_id)

assert result, msg
18 changes: 9 additions & 9 deletions autotest/benchmark/test_throughput_performance.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,51 +7,51 @@

@pytest.mark.gpu_num_1
@pytest.mark.flaky(reruns=0)
@pytest.mark.parametrize('run_config', get_benchmark_model_list(tp_num=1, kvint_list=[4, 8]))
@pytest.mark.parametrize('run_config', get_benchmark_model_list(parallel_config=1, kvint_list=[4, 8]))
def test_throughput_tp1(config, run_id, run_config, worker_id):
result, msg = throughput_test(config,
run_id,
run_config,
cuda_prefix=get_cuda_prefix_by_workerid(worker_id, tp_num=1),
cuda_prefix=get_cuda_prefix_by_workerid(worker_id, parallel_config=1),
worker_id=worker_id)

assert result, msg


@pytest.mark.gpu_num_2
@pytest.mark.flaky(reruns=0)
@pytest.mark.parametrize('run_config', get_benchmark_model_list(tp_num=2, kvint_list=[4, 8]))
@pytest.mark.parametrize('run_config', get_benchmark_model_list(parallel_config=2, kvint_list=[4, 8]))
def test_throughput_tp2(config, run_id, run_config, worker_id):
result, msg = throughput_test(config,
run_id,
run_config,
cuda_prefix=get_cuda_prefix_by_workerid(worker_id, tp_num=2),
cuda_prefix=get_cuda_prefix_by_workerid(worker_id, parallel_config=2),
worker_id=worker_id)

assert result, msg


@pytest.mark.gpu_num_4
@pytest.mark.flaky(reruns=0)
@pytest.mark.parametrize('run_config', get_benchmark_model_list(tp_num=4, kvint_list=[4, 8]))
@pytest.mark.parametrize('run_config', get_benchmark_model_list(parallel_config=4, kvint_list=[4, 8]))
def test_throughput_tp4(config, run_id, run_config, worker_id):
result, msg = throughput_test(config,
run_id,
run_config,
cuda_prefix=get_cuda_prefix_by_workerid(worker_id, tp_num=4),
cuda_prefix=get_cuda_prefix_by_workerid(worker_id, parallel_config=4),
worker_id=worker_id)

assert result, msg


@pytest.mark.gpu_num_8
@pytest.mark.flaky(reruns=0)
@pytest.mark.parametrize('run_config', get_benchmark_model_list(tp_num=4, kvint_list=[4, 8]))
@pytest.mark.parametrize('run_config', get_benchmark_model_list(parallel_config=4, kvint_list=[4, 8]))
def test_throughput_tp8(config, run_id, run_config, worker_id):
result, msg = throughput_test(config,
run_id,
run_config,
cuda_prefix=get_cuda_prefix_by_workerid(worker_id, tp_num=8),
cuda_prefix=get_cuda_prefix_by_workerid(worker_id, parallel_config=8),
worker_id=worker_id)

assert result, msg
Expand All @@ -73,7 +73,7 @@ def test_throughput_func_tp2(config, run_id, run_config, worker_id):
result, msg = throughput_test(config,
run_id,
run_config,
cuda_prefix=get_cuda_prefix_by_workerid(worker_id, tp_num=2),
cuda_prefix=get_cuda_prefix_by_workerid(worker_id, parallel_config=2),
worker_id=worker_id,
is_smoke=True)

Expand Down
86 changes: 46 additions & 40 deletions autotest/config-3090.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,56 +9,62 @@ tp_config:
empty: 2

turbomind_chat_model:
- meta-llama/Llama-3.2-3B-Instruct
- meta-llama/Llama-3.2-1B-Instruct
- internlm/internlm3-8b-instruct
- OpenGVLab/InternVL3-8B
- OpenGVLab/InternVL3-2B-Instruct
- OpenGVLab/InternVL3-1B-Instruct
- OpenGVLab/InternVL2_5-1B
- Qwen/Qwen3-8B
- Qwen/Qwen3-4B
- Qwen/Qwen3-1.7B
- Qwen/Qwen3-0.6B
- Qwen/Qwen2.5-7B-Instruct
tp:
- meta-llama/Llama-3.2-3B-Instruct
- meta-llama/Llama-3.2-1B-Instruct
- internlm/internlm3-8b-instruct
- OpenGVLab/InternVL3-8B
- OpenGVLab/InternVL3-2B-Instruct
- OpenGVLab/InternVL3-1B-Instruct
- OpenGVLab/InternVL2_5-1B
- Qwen/Qwen3-8B
- Qwen/Qwen3-4B
- Qwen/Qwen3-1.7B
- Qwen/Qwen3-0.6B
- Qwen/Qwen2.5-7B-Instruct

pytorch_chat_model:
- meta-llama/Llama-3.2-3B-Instruct
- meta-llama/Llama-3.2-1B-Instruct
- internlm/internlm3-8b-instruct
- OpenGVLab/InternVL3-8B
- OpenGVLab/InternVL3-2B-Instruct
- OpenGVLab/InternVL3-1B-Instruct
- OpenGVLab/InternVL2_5-1B
- Qwen/Qwen3-8B
- Qwen/Qwen3-4B
- Qwen/Qwen3-1.7B
- Qwen/Qwen3-0.6B
- Qwen/Qwen2.5-7B-Instruct
- Qwen/Qwen2.5-VL-3B-Instruct
- Qwen/Qwen2.5-VL-7B-Instruct
tp:
- meta-llama/Llama-3.2-3B-Instruct
- meta-llama/Llama-3.2-1B-Instruct
- internlm/internlm3-8b-instruct
- OpenGVLab/InternVL3-8B
- OpenGVLab/InternVL3-2B-Instruct
- OpenGVLab/InternVL3-1B-Instruct
- OpenGVLab/InternVL2_5-1B
- Qwen/Qwen3-8B
- Qwen/Qwen3-4B
- Qwen/Qwen3-1.7B
- Qwen/Qwen3-0.6B
- Qwen/Qwen2.5-7B-Instruct
- Qwen/Qwen2.5-VL-3B-Instruct
- Qwen/Qwen2.5-VL-7B-Instruct

turbomind_vl_model:
- OpenGVLab/InternVL3-8B
- OpenGVLab/InternVL3-2B-Instruct
- OpenGVLab/InternVL3-1B-Instruct
- OpenGVLab/InternVL2_5-1B
tp:
- OpenGVLab/InternVL3-8B
- OpenGVLab/InternVL3-2B-Instruct
- OpenGVLab/InternVL3-1B-Instruct
- OpenGVLab/InternVL2_5-1B

pytorch_vl_model:
- OpenGVLab/InternVL3-8B
- OpenGVLab/InternVL3-2B-Instruct
- OpenGVLab/InternVL3-1B-Instruct
- OpenGVLab/InternVL2_5-1B
- Qwen/Qwen2.5-VL-3B-Instruct
- Qwen/Qwen2.5-VL-7B-Instruct
tp:
- OpenGVLab/InternVL3-8B
- OpenGVLab/InternVL3-2B-Instruct
- OpenGVLab/InternVL3-1B-Instruct
- OpenGVLab/InternVL2_5-1B
- Qwen/Qwen2.5-VL-3B-Instruct
- Qwen/Qwen2.5-VL-7B-Instruct

turbomind_base_model:
- internlm/internlm3-8b-instruct
- Qwen/Qwen3-8B
tp:
- internlm/internlm3-8b-instruct
- Qwen/Qwen3-8B

pytorch_base_model:
- internlm/internlm3-8b-instruct
- Qwen/Qwen3-8B
tp:
- internlm/internlm3-8b-instruct
- Qwen/Qwen3-8B

turbomind_quatization:
no_awq:
Expand Down
58 changes: 32 additions & 26 deletions autotest/config-5080.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,42 +9,48 @@ tp_config:
empty: 2

turbomind_chat_model:
- meta-llama/Llama-3.2-3B-Instruct
- meta-llama/Llama-3.2-1B-Instruct
- OpenGVLab/InternVL3-2B-Instruct
- OpenGVLab/InternVL3-1B-Instruct
- OpenGVLab/InternVL2_5-1B
- Qwen/Qwen3-4B
- Qwen/Qwen3-1.7B
- Qwen/Qwen3-0.6B
tp:
- meta-llama/Llama-3.2-3B-Instruct
- meta-llama/Llama-3.2-1B-Instruct
- OpenGVLab/InternVL3-2B-Instruct
- OpenGVLab/InternVL3-1B-Instruct
- OpenGVLab/InternVL2_5-1B
- Qwen/Qwen3-4B
- Qwen/Qwen3-1.7B
- Qwen/Qwen3-0.6B

pytorch_chat_model:
- meta-llama/Llama-3.2-3B-Instruct
- meta-llama/Llama-3.2-1B-Instruct
- OpenGVLab/InternVL3-2B-Instruct
- OpenGVLab/InternVL3-1B-Instruct
- OpenGVLab/InternVL2_5-1B
- Qwen/Qwen3-4B
- Qwen/Qwen3-1.7B
- Qwen/Qwen3-0.6B
- Qwen/Qwen2.5-VL-3B-Instruct
tp:
- meta-llama/Llama-3.2-3B-Instruct
- meta-llama/Llama-3.2-1B-Instruct
- OpenGVLab/InternVL3-2B-Instruct
- OpenGVLab/InternVL3-1B-Instruct
- OpenGVLab/InternVL2_5-1B
- Qwen/Qwen3-4B
- Qwen/Qwen3-1.7B
- Qwen/Qwen3-0.6B
- Qwen/Qwen2.5-VL-3B-Instruct

turbomind_vl_model:
- OpenGVLab/InternVL3-2B-Instruct
- OpenGVLab/InternVL3-1B-Instruct
- OpenGVLab/InternVL2_5-1B
tp:
- OpenGVLab/InternVL3-2B-Instruct
- OpenGVLab/InternVL3-1B-Instruct
- OpenGVLab/InternVL2_5-1B

pytorch_vl_model:
- OpenGVLab/InternVL3-2B-Instruct
- OpenGVLab/InternVL3-1B-Instruct
- OpenGVLab/InternVL2_5-1B
- Qwen/Qwen2.5-VL-3B-Instruct
tp:
- OpenGVLab/InternVL3-2B-Instruct
- OpenGVLab/InternVL3-1B-Instruct
- OpenGVLab/InternVL2_5-1B
- Qwen/Qwen2.5-VL-3B-Instruct

turbomind_base_model:
- Qwen/Qwen3-4B
tp:
- Qwen/Qwen3-4B

pytorch_base_model:
- Qwen/Qwen3-4B
tp:
- Qwen/Qwen3-4B

turbomind_quatization:
no_awq:
Expand Down
Loading
Loading