From a7353c35e8af9564c218929ab35d0e341211d41d Mon Sep 17 00:00:00 2001 From: Mohit Sharma Date: Fri, 11 Apr 2025 15:10:19 +0000 Subject: [PATCH] fix bt --- server/text_generation_server/models/transformers_flash_vlm.py | 1 + 1 file changed, 1 insertion(+) diff --git a/server/text_generation_server/models/transformers_flash_vlm.py b/server/text_generation_server/models/transformers_flash_vlm.py index b20eae624..c1852a4f0 100644 --- a/server/text_generation_server/models/transformers_flash_vlm.py +++ b/server/text_generation_server/models/transformers_flash_vlm.py @@ -769,6 +769,7 @@ class TransformersLlama4VlmCausalLM(TransformersFlashVlmCausalLM): block_tables = self.cuda_graphs[max_bs]["block_tables"][: bs * max_bt] else: block_tables = self.cuda_graphs[max_bs]["block_tables"][:bs] + block_tables_local = self.cuda_graphs[max_bs]["block_tables_local"][:bs] slots = self.cuda_graphs[max_bs]["slots"][:bs] input_lengths_tensor = self.cuda_graphs[max_bs]["input_lengths"][:bs] cache_lengths_tensor = self.cuda_graphs[max_bs]["cache_lengths"][:bs]