All temperature validation gates use comparison operators (<, >), which silently evaluate to False for NaN and for positive Infinity in Python's IEEE 754 float semantics. Both values pass every guard and propagate to GPU sampling kernels, where they produce undefined behavior or CUDA errors that can crash the inference worker. Note: -Infinity is correctly caught.

sampling params.py:384:

if 0 < self.temperature < MAX TEMP: # NaN → False; +Inf → False

sampling params.py:462:

if self.temperature < 0.0:      # NaN → False; +Inf → False
  raise VLLMValidationError(...)

No math.isnan() or math.isinf() check exists anywhere in sampling params.py.

Python semantics (verified): float('nan') < 0.0 → False, float('inf') < 0.0 → False.

Crash of inference worker on GPU kernel execution with NaN/Inf softmax input, degrading service for all concurrent users.

Add math.isfinite(self.temperature) check in verify args(). Reject non-finite float values with a 400 error.

A fix for this vulnerability was merged here: https://github.com/vllm-project/vllm/pull/45116