[TPU] support fp8 kv cache quantization (#19292)

Signed-off-by: Chengji Yao <chengjiyao@google.com>
This commit is contained in:
Chengji Yao
2025-07-20 03:01:00 +00:00
committed by GitHub
parent 2b504eb770
commit 3a1d8940ae
6 changed files with 94 additions and 27 deletions
+3 -1
View File
@@ -35,7 +35,9 @@ class TpuPlatform(Platform):
device_control_env_var: str = "TPU_VISIBLE_CHIPS"
simple_compile_backend: str = "openxla"
supported_quantization: list[str] = ["tpu_int8", "compressed-tensors"]
supported_quantization: list[str] = [
"fp8", "tpu_int8", "compressed-tensors"
]
additional_env_vars: list[str] = [
"TPU_CHIPS_PER_HOST_BOUNDS", "TPU_HOST_BOUNDS"