From b9ad3acc4e324834ebe0e65a93b898467dfe2c01 Mon Sep 17 00:00:00 2001 From: OlivierDehaene <23298448+OlivierDehaene@users.noreply.github.com> Date: Fri, 12 May 2023 15:53:56 +0200 Subject: [PATCH] clean dtype --- server/text_generation_server/utils/logits_process.py | 1 - server/text_generation_server/utils/tokens.py | 2 +- 2 files changed, 1 insertion(+), 2 deletions(-) diff --git a/server/text_generation_server/utils/logits_process.py b/server/text_generation_server/utils/logits_process.py index 47f3a33c..9a738190 100644 --- a/server/text_generation_server/utils/logits_process.py +++ b/server/text_generation_server/utils/logits_process.py @@ -197,7 +197,6 @@ class HeterogeneousTopKLogitsWarper(LogitsWarper): def __init__( self, top_k: List[int], - dtype: torch.dtype, device: torch.device, filter_value: float = -math.inf, min_tokens_to_keep: int = 1, diff --git a/server/text_generation_server/utils/tokens.py b/server/text_generation_server/utils/tokens.py index c4e45b72..118129ae 100644 --- a/server/text_generation_server/utils/tokens.py +++ b/server/text_generation_server/utils/tokens.py @@ -187,7 +187,7 @@ class HeterogeneousNextTokenChooser: if any([x != 0 for x in top_k]): do_sample = [sample or x != 0 for x, sample in zip(top_k, do_sample)] - warpers.append(HeterogeneousTopKLogitsWarper(top_k, dtype, device)) + warpers.append(HeterogeneousTopKLogitsWarper(top_k, device)) if any([x < 1.0 for x in top_p]): do_sample = [sample or x < 1.0 for x, sample in zip(top_p, do_sample)]