From c35810d6f024e7492f72dfc9a68fd9d2d1eb4d22 Mon Sep 17 00:00:00 2001 From: Yuan Wu Date: Fri, 28 Feb 2025 18:20:55 +0800 Subject: [PATCH] Fix the loading issue of 90B (#283) Signed-off-by: yuanwu --- server/text_generation_server/models/vlm_causal_lm.py | 1 + 1 file changed, 1 insertion(+) diff --git a/server/text_generation_server/models/vlm_causal_lm.py b/server/text_generation_server/models/vlm_causal_lm.py index 06c9951e1..f17e9c247 100644 --- a/server/text_generation_server/models/vlm_causal_lm.py +++ b/server/text_generation_server/models/vlm_causal_lm.py @@ -784,6 +784,7 @@ class VlmCausalLM(Model): htorch.core.hpu_set_env() if world_size > 1: + os.environ.setdefault("DEEPSPEED_USE_HABANA_FRAMEWORKS_DETERMINISTIC_API", "1") model = self.get_deepspeed_model( model_class, model_id, dtype, revision )