From a24c2cc5e9c26eb3d0331a1fd607db6771c3d743 Mon Sep 17 00:00:00 2001 From: Mohit Sharma Date: Fri, 27 Sep 2024 12:39:12 +0000 Subject: [PATCH] updated default value --- server/text_generation_server/layers/attention/rocm.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/server/text_generation_server/layers/attention/rocm.py b/server/text_generation_server/layers/attention/rocm.py index be6158c1..de7d673f 100644 --- a/server/text_generation_server/layers/attention/rocm.py +++ b/server/text_generation_server/layers/attention/rocm.py @@ -283,7 +283,7 @@ elif ENGINE == "triton": softmax_scale: float, window_size_left: int = -1, causal: bool = True, - softcap: float = 0.0, + softcap: Optional[float] = None, ): if softcap is not None: raise NotImplementedError("softcap is only available with CK flash attn")