[BUG] Verify v0.2 IPEX-LLM Support and Hotfix (#16)

# CHANGES
- bump to `v0.2.0`

---------

Co-authored-by: tjtanaa <tunjian.tan@embeddedllm.com>

Author: tjtanaa

ellm_api_server.spec

@@ -61,6 +61,7 @@ if backend in ('directml', 'cpu', 'cuda'):
     add_package('onnxruntime')
     add_package('onnxruntime_genai')
 elif backend == 'ipex':
+    print(f"Backend IPEX")
     add_package('ipex_llm')
     add_package('torch')
     add_package('torchvision')

setup.py

@@ -60,7 +60,7 @@ def run(self):
                     "install",
                     "--pre",
                     "--upgrade",
-                    "ipex-llm[xpu]",
+                    "ipex-llm[xpu]==2.1.0b20240731",
                     "--extra-index-url",
                     "https://pytorch-extension.intel.com/release-whl/stable/xpu/us/",
                 ],
@@ -74,6 +74,18 @@ def run(self):
                 text=True,
             )
 
+            result = subprocess.run(
+                ["pip", "uninstall", "numpy", "-y"],
+                capture_output=True,
+                text=True,
+            )
+
+            result = subprocess.run(
+                ["pip", "install", "numpy==1.26.4", "--no-cache"],
+                capture_output=True,
+                text=True,
+            )
+
         if _is_directml():
             result = subprocess.run(
                 ["conda", "install", "conda-forge::vs2015_runtime", "-y"],
@@ -93,7 +105,7 @@ def run(self):
                     "install",
                     "--pre",
                     "--upgrade",
-                    "ipex-llm[xpu]",
+                    "ipex-llm[xpu]==2.1.0b20240731",
                     "--extra-index-url",
                     "https://pytorch-extension.intel.com/release-whl/stable/xpu/us/",
                 ],
@@ -107,6 +119,18 @@ def run(self):
                 text=True,
             )
 
+            result = subprocess.run(
+                ["pip", "uninstall", "numpy", "-y"],
+                capture_output=True,
+                text=True,
+            )
+
+            result = subprocess.run(
+                ["pip", "install", "numpy==1.26.4", "--no-cache"],
+                capture_output=True,
+                text=True,
+            )
+
         if _is_directml():
             result = subprocess.run(
                 ["conda", "install", "conda-forge::vs2015_runtime", "-y"],

src/embeddedllm/entrypoints/api_server.py

@@ -99,21 +99,6 @@ def main():
 
     print(f"PATH: {os.environ['PATH']}")
 
-    os.environ["SYCL_CACHE_PERSISTENT"] = "1"
-    os.environ["BIGDL_LLM_XMX_DISABLED"] = "1"
-    import sys
-
-    if getattr(sys, "frozen", False):
-        # Running as compiled executable
-        bundle_dir = sys._MEIPASS
-    else:
-        # Running in a normal Python environment
-        bundle_dir = os.path.dirname(os.path.abspath(__file__))
-
-    os.environ["PATH"] = bundle_dir + os.pathsep + os.environ["PATH"]
-
-    print(f"PATH: {os.environ['PATH']}")
-
     import uvicorn
 
     if os.name == "nt":

src/embeddedllm/version.py

@@ -1 +1 @@
-__version__ = "0.0.1"
+__version__ = "0.2.0"