vox_serve.flashinfer_utils

Functions

apply_rope_pos_ids(query_states, key_states, ...)	Wrapper for FlashInfer RoPE application with position IDs.
rms_norm(hidden_states, weight[, eps])	Wrapper for FlashInfer RMSNorm operation.

Classes

FlashInferDecodeWrapper(attn_buffer, ...[, ...])
FlashInferPrefillWrapper(attn_buffer, ...[, ...])