Function quant_fp8_vllm

lightx2v/utils/quant_utils.py:239–241 · view source on GitHub ↗

(input_tensor)

Source from the content-addressed store, hash-verified

237
238
239	def quant_fp8_vllm(input_tensor):
240	input_tensor_fp8, input_tensor_scale = ops.scaled_fp8_quant(input_tensor, scale=None, scale_ub=None, use_per_token_if_dynamic=True)
241	return input_tensor_fp8, input_tensor_scale
242
243
244	def dequant_fp8_vllm(input_tensor_fp8, input_tensor_scale, dtype):

_quant_and_sendMethod · 0.90

applyMethod · 0.90

_reshape_img_attnMethod · 0.90

applyMethod · 0.90

_reshape_img_attnMethod · 0.90

_fp8_all_to_allFunction · 0.90

quant_utils.pyFile · 0.85

no outgoing calls

no test coverage detected