[TPU][Quantization] TPU W8A8 (#11785)

Co-authored-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
This commit is contained in:
Robert Shaw
2025-01-08 19:33:29 +00:00
committed by GitHub
co-authored by Woosuk Kwon
parent 47de8821d3
commit 56fe4c297c
18 changed files with 565 additions and 190 deletions
+3 -1
View File
@@ -19,7 +19,9 @@ class TpuPlatform(Platform):
device_name: str = "tpu"
device_type: str = "tpu"
dispatch_key: str = "XLA"
supported_quantization: list[str] = ["tpu_int8"]
supported_quantization: list[str] = [
"tpu_int8", "compressed-tensors", "compressed_tensors"
]
@classmethod
def get_default_attn_backend(cls, selected_backend: _Backend) -> _Backend: