From fa9221f28d6f490f0843e94b623d5ac3c8c5f86e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Danie=CC=88l=20de=20Kok?= Date: Thu, 25 Jul 2024 09:39:42 +0000 Subject: [PATCH] Fix GPTQ autotune data type to be compatible with Torch 2.4.0 --- server/text_generation_server/layers/gptq/custom_autotune.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/server/text_generation_server/layers/gptq/custom_autotune.py b/server/text_generation_server/layers/gptq/custom_autotune.py index 1eb40f1e..0388ef20 100644 --- a/server/text_generation_server/layers/gptq/custom_autotune.py +++ b/server/text_generation_server/layers/gptq/custom_autotune.py @@ -91,7 +91,7 @@ class Autotuner(triton.KernelInterface): kernel_call, quantiles=(0.5, 0.2, 0.8), rep=40 ) except triton.OutOfResources: - return (float("inf"), float("inf"), float("inf")) + return [float("inf"), float("inf"), float("inf")] def run(self, *args, **kwargs): self.nargs = dict(zip(self.arg_names, args))