Update to ipex 2.2.0 for CPU (#143)

* update * update
intel · Mar 15, 2024 · dcef31b · dcef31b
1 parent e2a6450
commit dcef31b
Show file tree

Hide file tree

Showing 4 changed files with 7 additions and 7 deletions.
diff --git a/llm_on_ray/inference/deepspeed_predictor.py b/llm_on_ray/inference/deepspeed_predictor.py
@@ -142,7 +142,7 @@ def init_model(self, local_rank: int):
                 ipex._C.disable_jit_linear_repack()
             except Exception:
                 pass
-            pipe.model = ipex.optimize_transformers(
+            pipe.model = ipex.llm.optimize(
                 pipe.model.eval(),
                 dtype=torch.bfloat16
                 if self.infer_conf.ipex.precision == PRECISION_BF16

diff --git a/llm_on_ray/inference/mllm_predictor.py b/llm_on_ray/inference/mllm_predictor.py
@@ -43,7 +43,7 @@ def __init__(self, infer_conf: InferenceConfig):
                 ipex._C.disable_jit_linear_repack()
             except Exception:
                 pass
-            model = ipex.optimize_transformers(
+            model = ipex.llm.optimize(
                 model.eval(),
                 dtype=torch.bfloat16
                 if infer_conf.ipex.precision == PRECISION_BF16

diff --git a/llm_on_ray/inference/transformer_predictor.py b/llm_on_ray/inference/transformer_predictor.py
@@ -86,7 +86,7 @@ def __init__(self, infer_conf: InferenceConfig):
                     ipex._C.disable_jit_linear_repack()
                 except Exception:
                     pass
-                model = ipex.optimize_transformers(
+                model = ipex.llm.optimize(
                     model.eval(),
                     dtype=torch.bfloat16
                     if infer_conf.ipex.precision == PRECISION_BF16

diff --git a/pyproject.toml b/pyproject.toml
@@ -39,10 +39,10 @@ dependencies = [
 
 [project.optional-dependencies]
 cpu = [
-    "transformers>=4.35.0",
-    "intel_extension_for_pytorch==2.1.0+cpu",
-    "torch==2.1.0+cpu",
-    "oneccl_bind_pt==2.1.0+cpu"
+    "transformers>=4.35.0, <=4.35.2",
+    "intel_extension_for_pytorch>=2.2.0",
+    "torch>=2.2.0",
+    "oneccl_bind_pt>=2.2.0"
 ]
 
 gpu = [