Skip to content

[CI] Fix test_inference_engines_generation after vllm 0.16.0 upgrade; Use the correct GSM8k path for test_generator_multi_turn_gsm8k_router_replay #1339

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
SumanthRH merged 2 commits into from
Mar 18, 2026
Merged

[CI] Fix test_inference_engines_generation after vllm 0.16.0 upgrade; Use the correct GSM8k path for test_generator_multi_turn_gsm8k_router_replay #1339

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
4a6afda
fix ci faiilures
SumanthRH Mar 18, 2026
79da8fb
x
SumanthRH Mar 18, 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
27 changes: 13 additions & 14 deletions tests/backends/skyrl_train/gpu/gpu_ci/test_engine_generation.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,10 +25,10 @@
MOE_MODEL = "Qwen/Qwen1.5-MoE-A2.7B"


def get_test_actor_config() -> SkyRLTrainConfig:
def get_test_actor_config(model: str = MODEL) -> SkyRLTrainConfig:
"""Get base config with test-specific overrides."""
cfg = SkyRLTrainConfig()
cfg.trainer.policy.model.path = MODEL
cfg.trainer.policy.model.path = model

cfg.generator.sampling_params.temperature = 0.0
cfg.generator.sampling_params.top_p = 1
Expand Down Expand Up @@ -90,25 +90,25 @@ async def run_single_generation_with_tokens(client, prompt_token_ids, sampling_p

@pytest.mark.skipif(_SKYRL_USE_NEW_INFERENCE, reason="New inference pathway doesn't support text based generation")
@pytest.mark.parametrize(
"tp_size,pp_size,dp_size",
"tp_size,pp_size,dp_size,model",
[
pytest.param(2, 1, 1),
pytest.param(2, 1, 2),
pytest.param(2, 2, 1), # TP=2, PP=2
pytest.param(2, 1, 1, MODEL),
pytest.param(2, 1, 2, MOE_MODEL),
pytest.param(2, 2, 1, MODEL), # TP=2, PP=2
],
ids=["tp2_pp1_dp1", "tp2_pp1_dp2", "tp2_pp2_dp1"],
ids=["tp2_pp1_dp1", "tp2_pp1_dp2_moe", "tp2_pp2_dp1"],
)
def test_inference_engines_generation(ray_init_fixture, tp_size: int, pp_size: int, dp_size: int):
def test_inference_engines_generation(ray_init_fixture, tp_size: int, pp_size: int, dp_size: int, model: str):
"""
Tests generation with both remote and ray-wrapped engines.
"""
cfg = get_test_actor_config()
cfg = get_test_actor_config(model)

prompts = get_test_prompts(MODEL)
tokenizer = AutoTokenizer.from_pretrained(MODEL)
prompts = get_test_prompts(model)
tokenizer = AutoTokenizer.from_pretrained(model)

try:
llm_client, remote_server_process = init_remote_inference_servers(tp_size, "vllm", tokenizer, cfg, MODEL)
llm_client, remote_server_process = init_remote_inference_servers(tp_size, "vllm", tokenizer, cfg, model)
sampling_params = get_sampling_params_for_backend(
cfg.generator.inference_engine.backend, cfg.generator.sampling_params
)
Expand Down Expand Up @@ -210,8 +210,7 @@ def test_token_based_generation(
):
"""Test generation using prompt_token_ids."""

cfg = get_test_actor_config()
cfg.trainer.policy.model.path = model
cfg = get_test_actor_config(model)

prompts = get_test_prompts(model, 3)
tokenizer = AutoTokenizer.from_pretrained(model)
Expand Down
2 changes: 1 addition & 1 deletion tests/backends/skyrl_train/gpu/gpu_ci/test_skyrl_gym_generator.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -488,7 +488,7 @@ async def test_generator_multi_turn_gsm8k_router_replay(ray_init_fixture):
max_prompt_length=2048,
max_input_length=max_input_length,
max_generate_length=1000,
data_path=os.path.expanduser("/mnt/cluster_storage/data/gsm8k/validation.parquet"),
data_path=os.path.expanduser("~/data/gsm8k/validation.parquet"),
Copy link
Copy Markdown
Contributor

@gemini-code-assist gemini-code-assist Bot Mar 18, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

This line explicitly sets data_path to the same value as its default in the run_generator_end_to_end function signature. To reduce redundancy and improve maintainability, you can remove this line and rely on the default value.

env_class="gsm8k_multi_turn",
num_prompts=num_prompts,
max_turns=2,
Expand Down
Loading