vllm.kernels.triton.layernorm_batch_invariant ¶

rms_norm_no_var = (
    lambda x, weight, epsilon, variance_size=None: (
        variance_size is None
    )
)