maxent_grpo.training.runtime.ops.vllm_startup

Detect and classify vLLM server startup stalls from log text.

Functions

_build_parser()
classify_vllm_startup_log(log_text[, ...])	Classify startup progress using marker patterns in log_text.
main()
should_trigger_v0_fallback(log_text, attempt)	Return True when vLLM startup appears stuck and should be relaunched in V0 mode.

Classes

StartupStatus(value)

High-level startup state derived from vLLM log lines.

class maxent_grpo.training.runtime.ops.vllm_startup.StartupStatus(value)

Bases: str, Enum

High-level startup state derived from vLLM log lines.

STARTING = 'starting'

HEALTHY = 'healthy'

CORE_ENGINE_STALL = 'core_engine_stall'

ERROR = 'error'

maxent_grpo.training.runtime.ops.vllm_startup.classify_vllm_startup_log(log_text, stall_threshold=3)

Classify startup progress using marker patterns in log_text.

Parameters:

• log_text (str)

• stall_threshold (int)

Return type:

StartupStatus

maxent_grpo.training.runtime.ops.vllm_startup.should_trigger_v0_fallback(log_text, attempt, min_attempts=20, stall_threshold=3)

Return True when vLLM startup appears stuck and should be relaunched in V0 mode.

Parameters:

• log_text (str)

• attempt (int)

• min_attempts (int)

• stall_threshold (int)

Return type:

bool

maxent_grpo.training.runtime.ops.vllm_startup.main()

Return type:

int