From d6072b96d0f0519b3d905168cff9064ca44d15f1 Mon Sep 17 00:00:00 2001 From: Ajinkya Tejankar Date: Sat, 8 Feb 2025 00:31:33 +0530 Subject: [PATCH] Fix compile bug for stella (#757) --- server/lorax_server/models/flash_causal_lm.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/server/lorax_server/models/flash_causal_lm.py b/server/lorax_server/models/flash_causal_lm.py index 4f5cc842..c064cbe5 100644 --- a/server/lorax_server/models/flash_causal_lm.py +++ b/server/lorax_server/models/flash_causal_lm.py @@ -1352,7 +1352,7 @@ def warmup(self, batch: FlashCausalLMBatch, max_new_tokens: int, embedding_model torch.cuda.synchronize(self.device) logger.info(f'Post warmup cuda memory: {get_cuda_free_memory(self.device, 1) / (1024 ** 3):.2f} GB') - del self.model_graph_wrapper + self.model_graph_wrapper = None self.kv_cache = [] torch.cuda.synchronize(self.device) torch.cuda.empty_cache()