ipex-llm 2.2.0b20250218__py3-none-win_amd64.whl → 2.2.0b20250219__py3-none-win_amd64.whl

ipex_llm/vllm/xpu/engine/engine.py

@@ -117,7 +117,6 @@ class IPEXLLMClass(LLM):
         Note: if enforce_eager is unset (enforce_eager is None)
         it defaults to False.
         '''
-
         if "disable_log_stats" not in kwargs:
             kwargs["disable_log_stats"] = True
 

ipex_llm/vllm/xpu/model_convert.py

@@ -82,39 +82,49 @@ def get_load_function(low_bit):
         # from vllm.utils import measure_device_memory
         from vllm.utils import DeviceMemoryProfiler
         with DeviceMemoryProfiler() as m:
+            import os
             from dataclasses import replace
             new_device_config = DeviceConfig("cpu")
             new_vllm_config = replace(self.vllm_config, device_config=new_device_config)
+            # We are loading an low-bit model, where all the optimizations should have been
+            # applied...
+            # We can skip the following optimizations
             self.model = get_model(
                 vllm_config=new_vllm_config
             )
-            if "qwen" in self.vllm_config.model_config.model.lower() or \
-                    "baichuan" in self.vllm_config.model_config.model.lower() or \
-                    "codegeex4-all" in self.vllm_config.model_config.model.lower() or \
-                    "chatglm" in self.vllm_config.model_config.model.lower():
-                self.model.apply(padding_mlp)
-            from ipex_llm import optimize_model
-            import os
-            not_convert_last_mlp = os.getenv("IPEX_LLM_NOT_CONVERT_LAST_MLP", None)
-            if not_convert_last_mlp is not None:
-                # only use to avoid nan value in last mlp forward running glm4-9b-chat
-                modules = ["35.mlp", "36.mlp", "37.mlp", "38.mlp", "39.mlp"]
-            else:
-                modules = None
-            if "minicpm" in self.vllm_config.model_config.model.lower():
-                modules = ["vpm", "resampler"]
-            # only for minicpm_2_6
-            if "minicpm-v" in self.vllm_config.model_config.model.lower():
-                from ipex_llm.transformers.models.minicpmv import merge_qkv
-                self.model.vpm.apply(merge_qkv)
-            if "internvl2" in self.vllm_config.model_config.model.lower():
-                modules = ["vision_model", "mlp1"]
-            if "deepseek-v2" in self.vllm_config.model_config.model.lower():
-                modules = ["down_proj"]
-            optimize_model(self.model,
-                           low_bit=low_bit,
-                           torch_dtype=self.vllm_config.model_config.dtype,
-                           modules_to_not_convert=modules)
+            if self.vllm_config.model_config.low_bit_model_path is None:
+                if "qwen" in self.vllm_config.model_config.model.lower() or \
+                        "baichuan" in self.vllm_config.model_config.model.lower() or \
+                        "codegeex4-all" in self.vllm_config.model_config.model.lower() or \
+                        "chatglm" in self.vllm_config.model_config.model.lower():
+                    self.model.apply(padding_mlp)
+                from ipex_llm import optimize_model
+                not_convert_last_mlp = os.getenv("IPEX_LLM_NOT_CONVERT_LAST_MLP", None)
+                if not_convert_last_mlp is not None:
+                    # only use to avoid nan value in last mlp forward running glm4-9b-chat
+                    modules = ["35.mlp", "36.mlp", "37.mlp", "38.mlp", "39.mlp"]
+                else:
+                    modules = None
+                if "minicpm" in self.vllm_config.model_config.model.lower():
+                    modules = ["vpm", "resampler"]
+                if "internvl2" in self.vllm_config.model_config.model.lower():
+                    modules = ["vision_model", "mlp1"]
+                if "deepseek-v2" in self.vllm_config.model_config.model.lower():
+                    modules = ["down_proj"]
+                optimize_model(self.model,
+                               low_bit=low_bit,
+                               torch_dtype=self.vllm_config.model_config.dtype,
+                               modules_to_not_convert=modules)
+            # Guancheng: We have to save the model before moving it to the XPU device.
+            # The `to` method will convert the underlying data.
+            # Saving it before will help to avoid converting two times.
+            if self.vllm_config.model_config.low_bit_save_path is not None:
+                # The local_rank is used for loading models with tensor parallel settings.
+                local_rank = os.environ["LOCAL_RANK"]
+                saved_path = os.path.join(self.vllm_config.model_config.low_bit_save_path,
+                                          str(local_rank))
+                self.model.save_low_bit(saved_path)
+
             self.model = self.model.to(device=self.vllm_config.device_config.device,
                                        dtype=self.vllm_config.model_config.dtype)
 

{ipex_llm-2.2.0b20250218.dist-info → ipex_llm-2.2.0b20250219.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: ipex-llm
-Version: 2.2.0b20250218
+Version: 2.2.0b20250219
 Summary: Large Language Model Develop Toolkit
 Home-page: https://github.com/intel-analytics/ipex-llm
 Author: BigDL Authors
@@ -27,7 +27,7 @@ Requires-Dist: intel-openmp ; (platform_machine == "x86_64" or platform_machine
 Requires-Dist: torch ==2.1.2+cpu ; (platform_system == "Linux") and extra == 'all'
 Requires-Dist: torch ==2.1.2 ; (platform_system == "Windows") and extra == 'all'
 Provides-Extra: cpp
-Requires-Dist: bigdl-core-cpp ==2.6.0b20250218 ; extra == 'cpp'
+Requires-Dist: bigdl-core-cpp ==2.6.0b20250219 ; extra == 'cpp'
 Requires-Dist: setuptools ; extra == 'cpp'
 Requires-Dist: onednn-devel ==2025.0.1 ; (platform_system == "Windows") and extra == 'cpp'
 Requires-Dist: onednn ==2025.0.1 ; (platform_system == "Windows") and extra == 'cpp'
@@ -60,7 +60,7 @@ Requires-Dist: transformers ==4.40.0 ; extra == 'npu'
 Requires-Dist: intel-openmp ; (platform_machine == "x86_64" or platform_machine == "AMD64") and extra == 'npu'
 Requires-Dist: torch ==2.1.2+cpu ; (platform_system == "Linux") and extra == 'npu'
 Requires-Dist: torch ==2.1.2 ; (platform_system == "Windows") and extra == 'npu'
-Requires-Dist: bigdl-core-npu ==2.6.0b20250218 ; (platform_system == "Windows") and extra == 'npu'
+Requires-Dist: bigdl-core-npu ==2.6.0b20250219 ; (platform_system == "Windows") and extra == 'npu'
 Provides-Extra: serving
 Requires-Dist: py-cpuinfo ; extra == 'serving'
 Requires-Dist: fschat[model_worker,webui] ==0.2.36 ; extra == 'serving'
@@ -80,9 +80,9 @@ Requires-Dist: setuptools <70.0.0 ; extra == 'xpu'
 Requires-Dist: torch ==2.1.0a0 ; extra == 'xpu'
 Requires-Dist: torchvision ==0.16.0a0 ; extra == 'xpu'
 Requires-Dist: intel-extension-for-pytorch ==2.1.10+xpu ; extra == 'xpu'
-Requires-Dist: bigdl-core-xe-21 ==2.6.0b20250218 ; extra == 'xpu'
-Requires-Dist: bigdl-core-xe-batch-21 ==2.6.0b20250218 ; extra == 'xpu'
-Requires-Dist: bigdl-core-xe-addons-21 ==2.6.0b20250218 ; extra == 'xpu'
+Requires-Dist: bigdl-core-xe-21 ==2.6.0b20250219 ; extra == 'xpu'
+Requires-Dist: bigdl-core-xe-batch-21 ==2.6.0b20250219 ; extra == 'xpu'
+Requires-Dist: bigdl-core-xe-addons-21 ==2.6.0b20250219 ; extra == 'xpu'
 Provides-Extra: xpu-2-1
 Requires-Dist: py-cpuinfo ; extra == 'xpu-2-1'
 Requires-Dist: protobuf ; extra == 'xpu-2-1'
@@ -97,9 +97,9 @@ Requires-Dist: setuptools <70.0.0 ; extra == 'xpu-2-1'
 Requires-Dist: torch ==2.1.0a0 ; extra == 'xpu-2-1'
 Requires-Dist: torchvision ==0.16.0a0 ; extra == 'xpu-2-1'
 Requires-Dist: intel-extension-for-pytorch ==2.1.10+xpu ; extra == 'xpu-2-1'
-Requires-Dist: bigdl-core-xe-21 ==2.6.0b20250218 ; extra == 'xpu-2-1'
-Requires-Dist: bigdl-core-xe-batch-21 ==2.6.0b20250218 ; extra == 'xpu-2-1'
-Requires-Dist: bigdl-core-xe-addons-21 ==2.6.0b20250218 ; extra == 'xpu-2-1'
+Requires-Dist: bigdl-core-xe-21 ==2.6.0b20250219 ; extra == 'xpu-2-1'
+Requires-Dist: bigdl-core-xe-batch-21 ==2.6.0b20250219 ; extra == 'xpu-2-1'
+Requires-Dist: bigdl-core-xe-addons-21 ==2.6.0b20250219 ; extra == 'xpu-2-1'
 Requires-Dist: intel-openmp ; (platform_machine == "x86_64" or platform_machine == "AMD64") and extra == 'xpu-2-1'
 Requires-Dist: dpcpp-cpp-rt ==2024.0.2 ; (platform_system == "Windows") and extra == 'xpu-2-1'
 Requires-Dist: mkl-dpcpp ==2024.0.0 ; (platform_system == "Windows") and extra == 'xpu-2-1'
@@ -117,7 +117,7 @@ Requires-Dist: setuptools ; extra == 'xpu-2-6'
 Requires-Dist: torch ==2.6.0+xpu ; extra == 'xpu-2-6'
 Requires-Dist: torchvision ==0.21.0+xpu ; extra == 'xpu-2-6'
 Requires-Dist: torchaudio ==2.6.0+xpu ; extra == 'xpu-2-6'
-Requires-Dist: bigdl-core-xe-all ==2.6.0b20250218 ; extra == 'xpu-2-6'
+Requires-Dist: bigdl-core-xe-all ==2.6.0b20250219 ; extra == 'xpu-2-6'
 Requires-Dist: onednn-devel ==2025.0.1 ; extra == 'xpu-2-6'
 Requires-Dist: onednn ==2025.0.1 ; extra == 'xpu-2-6'
 Requires-Dist: dpcpp-cpp-rt ==2025.0.2 ; extra == 'xpu-2-6'
@@ -133,9 +133,9 @@ Requires-Dist: tokenizers ==0.15.2 ; extra == 'xpu-arc'
 Requires-Dist: accelerate ==0.23.0 ; extra == 'xpu-arc'
 Requires-Dist: tabulate ; extra == 'xpu-arc'
 Requires-Dist: setuptools ; extra == 'xpu-arc'
-Requires-Dist: bigdl-core-xe-23 ==2.6.0b20250218 ; extra == 'xpu-arc'
-Requires-Dist: bigdl-core-xe-batch-23 ==2.6.0b20250218 ; extra == 'xpu-arc'
-Requires-Dist: bigdl-core-xe-addons-23 ==2.6.0b20250218 ; extra == 'xpu-arc'
+Requires-Dist: bigdl-core-xe-23 ==2.6.0b20250219 ; extra == 'xpu-arc'
+Requires-Dist: bigdl-core-xe-batch-23 ==2.6.0b20250219 ; extra == 'xpu-arc'
+Requires-Dist: bigdl-core-xe-addons-23 ==2.6.0b20250219 ; extra == 'xpu-arc'
 Requires-Dist: intel-openmp ; (platform_machine == "x86_64" or platform_machine == "AMD64") and extra == 'xpu-arc'
 Requires-Dist: torch ==2.3.1+cxx11.abi ; (platform_system == "Linux") and extra == 'xpu-arc'
 Requires-Dist: torchvision ==0.18.1+cxx11.abi ; (platform_system == "Linux") and extra == 'xpu-arc'
@@ -156,9 +156,9 @@ Requires-Dist: tokenizers ==0.15.2 ; extra == 'xpu-arl'
 Requires-Dist: accelerate ==0.23.0 ; extra == 'xpu-arl'
 Requires-Dist: tabulate ; extra == 'xpu-arl'
 Requires-Dist: setuptools ; extra == 'xpu-arl'
-Requires-Dist: bigdl-core-xe-23 ==2.6.0b20250218 ; extra == 'xpu-arl'
-Requires-Dist: bigdl-core-xe-batch-23 ==2.6.0b20250218 ; extra == 'xpu-arl'
-Requires-Dist: bigdl-core-xe-addons-23 ==2.6.0b20250218 ; extra == 'xpu-arl'
+Requires-Dist: bigdl-core-xe-23 ==2.6.0b20250219 ; extra == 'xpu-arl'
+Requires-Dist: bigdl-core-xe-batch-23 ==2.6.0b20250219 ; extra == 'xpu-arl'
+Requires-Dist: bigdl-core-xe-addons-23 ==2.6.0b20250219 ; extra == 'xpu-arl'
 Requires-Dist: intel-openmp ; (platform_machine == "x86_64" or platform_machine == "AMD64") and extra == 'xpu-arl'
 Requires-Dist: torch ==2.3.1+cxx11.abi ; (platform_system == "Linux") and extra == 'xpu-arl'
 Requires-Dist: torchvision ==0.18.1+cxx11.abi ; (platform_system == "Linux") and extra == 'xpu-arl'
@@ -179,9 +179,9 @@ Requires-Dist: tokenizers ==0.15.2 ; extra == 'xpu-lnl'
 Requires-Dist: accelerate ==0.23.0 ; extra == 'xpu-lnl'
 Requires-Dist: tabulate ; extra == 'xpu-lnl'
 Requires-Dist: setuptools ; extra == 'xpu-lnl'
-Requires-Dist: bigdl-core-xe-23 ==2.6.0b20250218 ; extra == 'xpu-lnl'
-Requires-Dist: bigdl-core-xe-batch-23 ==2.6.0b20250218 ; extra == 'xpu-lnl'
-Requires-Dist: bigdl-core-xe-addons-23 ==2.6.0b20250218 ; extra == 'xpu-lnl'
+Requires-Dist: bigdl-core-xe-23 ==2.6.0b20250219 ; extra == 'xpu-lnl'
+Requires-Dist: bigdl-core-xe-batch-23 ==2.6.0b20250219 ; extra == 'xpu-lnl'
+Requires-Dist: bigdl-core-xe-addons-23 ==2.6.0b20250219 ; extra == 'xpu-lnl'
 Requires-Dist: intel-openmp ; (platform_machine == "x86_64" or platform_machine == "AMD64") and extra == 'xpu-lnl'
 Requires-Dist: torch ==2.3.1+cxx11.abi ; (platform_system == "Linux") and extra == 'xpu-lnl'
 Requires-Dist: torchvision ==0.18.1+cxx11.abi ; (platform_system == "Linux") and extra == 'xpu-lnl'

{ipex_llm-2.2.0b20250218.dist-info → ipex_llm-2.2.0b20250219.dist-info}/RECORD

@@ -2,7 +2,7 @@ ipex_llm/__init__.py,sha256=kSA9JjVLPlpN4YchWtfOybRh4XiP6d_VTYvzbAouPSU,2118
 ipex_llm/convert_model.py,sha256=jopEe6wu88ZPZfNFhgnQUu7807iciiWW_EMyTsVni5A,6816
 ipex_llm/llm_patching.py,sha256=becMYcawtR8lgl2yeRQhvvX6CLaq09WZGm9dDmLJWL0,3232
 ipex_llm/models.py,sha256=XROP6GLLrGQDlogGXpXZENbV143YNi6j0VPJeOdQ3Cg,1063
-ipex_llm/optimize.py,sha256=ml-qEpzsrWGcd-Wia6IxPBC1PhqT5pi_lp3VTOH_ns0,12415
+ipex_llm/optimize.py,sha256=wzh2oSYf8gHlGXZ0gI64LphwG2x443mf2E54atd1ikI,12512
 ipex_llm/cli/llm-chat.ps1,sha256=6qrs-hGVAV8IKh7Jx8nq_XrnZcjd7qGU5wndArM7Yag,2769
 ipex_llm/cli/llm-cli.ps1,sha256=3qBtTLs_EjYDnM8YyCpJhzLnGCKTEGssu9UNqfkjVXs,3009
 ipex_llm/cli/prompts/chat-with-llm.txt,sha256=PpSyd4FQQd-T7ptfXL9jZp7dgstevu1fsxWFa0IQ5Oc,216
@@ -41,35 +41,35 @@ ipex_llm/langchain/llms/transformerspipelinellm.py,sha256=vm522YPPwWxxAPVvQBtxRf
 ipex_llm/langchain/vllm/__init__.py,sha256=T-EbRT6GJ_8RCu-iLmSzcftOimXSPQf2d5X72AUAy2Y,874
 ipex_llm/langchain/vllm/vllm.py,sha256=6dxc-ZISZQrJilEa_HA827l75Dv9rcHpY_G6FdJ8BVs,7793
 ipex_llm/libs/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-ipex_llm/libs/bloom-api.dll,sha256=d8aCw_Zs6BKCh5zpmUwjBZGOs46bcfBXnMIy_0an3ow,36352
-ipex_llm/libs/bloom.dll,sha256=TP6GVjWYijUlkQnWBjDlW7kvKpcJzJxHEDGtBvT7olw,507904
-ipex_llm/libs/gptneox-api.dll,sha256=2fS-oDM2bmOBipq-95m2RtVv51ze4PdTfZ28UXSIPDQ,24576
-ipex_llm/libs/gptneox.dll,sha256=sKm5TgnHY3BJIIjiFe6cQt4lBphpMFU6BRr4QZmDci4,568320
-ipex_llm/libs/libbloom_avx.dll,sha256=qLmBoimp4aX7KhPY7KGohyCu0Mh9SBI8VxJY44ac0gQ,536576
-ipex_llm/libs/libbloom_vnni.dll,sha256=RXSbcTRhzH2AnVQBq42xx67MFACA6CuTIR2LKTtRlaI,508416
-ipex_llm/libs/libgptneox_avx.dll,sha256=21mvCjWoi1fEBP5NnizQyphOk1xA7yJdcOR7qcBDA5s,596992
-ipex_llm/libs/libgptneox_vnni.dll,sha256=_qSoRKMsJ88eQaH4jIvrvPrx_0oxivaUGuFVy411y10,568832
-ipex_llm/libs/libllama_avx.dll,sha256=n4xs2B67ZyLJtlnjXTJZThSdsniGB06rFJeagMAxvzs,591360
-ipex_llm/libs/libllama_vnni.dll,sha256=bUwdPvNqoEBTROz9TvucoDh0xxWiSi1U-FdC8WNPfSQ,563200
-ipex_llm/libs/libstarcoder_avx.dll,sha256=5tuG5sVflMpdHVoZsTvtR-I5JOq0S1qyf2aivXXXObc,627712
-ipex_llm/libs/libstarcoder_vnni.dll,sha256=_50ImpAtJKC4hthiTJMXUYKFkDulmNdINcVlJNrxUpw,599552
-ipex_llm/libs/llama-api.dll,sha256=38Jai99Lw1mzt4PtxDhsCb5W0vWaD64_o33YxdMzsE8,25600
-ipex_llm/libs/llama.dll,sha256=zk5E0PDRgIt-iNCKVRW_U_q3mympN_2YTakjTrFbn54,562688
-ipex_llm/libs/main-bloom.exe,sha256=AMDXb4-uGkIv3eFuSWNDnEJ-dVxoSG2xd4bBS1acpno,103424
-ipex_llm/libs/main-gptneox.exe,sha256=Tv7QqocEixaphKkPstbgfTZru-XapM_DsiDEY_XPkJw,98816
-ipex_llm/libs/main-llama.exe,sha256=FpdNWYr9p9r2GUpHbjZ1WhtKL682A0rJ1I0DozKJEIM,99840
-ipex_llm/libs/main-starcoder.exe,sha256=uESuQN4jFspABcUl_EveacbB5qWvunB8RWikqBHD0mA,157696
-ipex_llm/libs/pipeline.dll,sha256=lAcAPf96GEelXDl4yzXsHoVg4G_Wz1SnVWMlzvSteCw,73216
-ipex_llm/libs/quantize-bloom.exe,sha256=wPQiIfQpaq3dlICHe9hnLIf0C5PK6L05jpVGJf3VUa4,126464
-ipex_llm/libs/quantize-bloom_vnni.exe,sha256=bqjOC9x7wM5pC81w7UVzp0mycojXnd96HrqFOiq0pOE,128000
-ipex_llm/libs/quantize-gptneox.exe,sha256=icCLsKxtxRm5NZ0sgDGQnJySKJyf-6SKox3ljG0OTyI,104448
-ipex_llm/libs/quantize-gptneox_vnni.exe,sha256=DUdtAcnyoMJtiPagmMgARZfa7Qlo1s6SUQivGHbxa10,104960
-ipex_llm/libs/quantize-llama.exe,sha256=kFtKNTjsB6P2UOQDi6Z0kiVAt960xgMO4tUGC6k6Y38,110080
-ipex_llm/libs/quantize-llama_vnni.exe,sha256=j7G7L1G6bJ5RRgYpPP-u69YyfU9g33kaRzf9sCFDtuk,110592
-ipex_llm/libs/quantize-starcoder.exe,sha256=jFmpXkT5b-EyLUuAO0UUarGfAQknBMOwbg5sK9kAehU,127488
-ipex_llm/libs/quantize-starcoder_vnni.exe,sha256=AgVdI68TRJV_QFaibI0uWbnoufqqt99jzOyvKlmVTEM,128512
-ipex_llm/libs/starcoder-api.dll,sha256=lWSurNBL4YsDICkUGW3SP-u1C1h5vcl8DZO4MfLdZKg,21504
-ipex_llm/libs/starcoder.dll,sha256=z95r5h1HSr7gKdQMxqRbiZje09JVqpOSH7xc-g_f1nE,599040
+ipex_llm/libs/bloom-api.dll,sha256=sB4zGLhQFAKWtLDXc6XAYWz8x_8Qm5v0bv5aU2LzSDw,36352
+ipex_llm/libs/bloom.dll,sha256=7y5cEjNL-PkswkFtKhO-lzn8-sSbJF1kHOcGq_eNe8I,507904
+ipex_llm/libs/gptneox-api.dll,sha256=nthBzorbkLzW2r7amDjKTdRqXBrH2F84t9HZTCwvFhc,24576
+ipex_llm/libs/gptneox.dll,sha256=2gh5kR-F22kN-IQyFU9q2bD5ziigOorKKqK48-iFC0U,568320
+ipex_llm/libs/libbloom_avx.dll,sha256=1UUBqLktjzlpTl75ySH7ShtlFsj8nFgEkYIu3DPaVP0,536576
+ipex_llm/libs/libbloom_vnni.dll,sha256=l83Scy5SYjRRekhpFy_yVa7MZ2QTKspJVZzA74BWBiU,508416
+ipex_llm/libs/libgptneox_avx.dll,sha256=ONlYccybul-zilyMIaou-01adyUCRMXZGcq4RRdYvF4,596992
+ipex_llm/libs/libgptneox_vnni.dll,sha256=BE_wB1VA7cTtEpUL7Nv3rGeOPSIQ-ObM7x1O7kEuqsk,568832
+ipex_llm/libs/libllama_avx.dll,sha256=GTYP0R_spavOan6-Qm5SC-bVp4eE69NKPyJBkaSULGw,591360
+ipex_llm/libs/libllama_vnni.dll,sha256=7OVS1VxkJNQQL7eCLZ3nefROQpD7_gm_xidP-7I409o,563200
+ipex_llm/libs/libstarcoder_avx.dll,sha256=Ml7xNu3eM76iVaMgVPwB_VEBPrxZTpNYalQsubhWET0,627712
+ipex_llm/libs/libstarcoder_vnni.dll,sha256=co8mE-IQfWJD9ALvzso3wyoH-dOoa_1bbeJLCkpq858,599552
+ipex_llm/libs/llama-api.dll,sha256=kEc3f8UKCy4PgRUdeMhOdIQEztPd801axX1vTai1gjc,25600
+ipex_llm/libs/llama.dll,sha256=mv7fwZFpy7szdjmI101TMbg2a_MHjcQZxOeF7jogRVI,562688
+ipex_llm/libs/main-bloom.exe,sha256=mezuJBuPVuGlC6VkBEqqKnpJhxWyqbz7FbUd32ToLNk,103424
+ipex_llm/libs/main-gptneox.exe,sha256=gWDigNVAHVRPoWvi0X7dV_Rwt_KeNSjw1ZJhOqKH-Ek,98816
+ipex_llm/libs/main-llama.exe,sha256=X72favOkzbXleVlAiBKKSMHpmQTQLzwkkm6ltqSzjrI,99840
+ipex_llm/libs/main-starcoder.exe,sha256=DpG5TGk7-G4Pv70Dqfwv7LQ8AlVDlAB-LYG9ezTMLSo,157696
+ipex_llm/libs/pipeline.dll,sha256=PFoGpdWfQy3eZF43--pQbQMuZVHC1fhDN6PoNWkSgtA,73216
+ipex_llm/libs/quantize-bloom.exe,sha256=xDQhEPew1hX5qF6rZ0qTILAs1A5n45sCBc0Q3VWtrP0,126464
+ipex_llm/libs/quantize-bloom_vnni.exe,sha256=73PI5zcKGsSpoE1TctdmURY7o5OnmzmHG5mBRyXCxeM,128000
+ipex_llm/libs/quantize-gptneox.exe,sha256=J9qck5JKSgR5QXzjVNqEh79rERaVSeSsk2ww13FNvtA,104448
+ipex_llm/libs/quantize-gptneox_vnni.exe,sha256=g0y3bd8nbQlLBbc5MS9o811XSb4TnNrI5MXug2J6iK4,104960
+ipex_llm/libs/quantize-llama.exe,sha256=UDLBq6C4xAx6RkXEDvJYU140PEaNKnrkPvb2SFwBMo0,110080
+ipex_llm/libs/quantize-llama_vnni.exe,sha256=vGEqClsNa0G1BbSEIH4DmhDkiqTo-NBgLAd9ilTI8CM,110592
+ipex_llm/libs/quantize-starcoder.exe,sha256=6lMuSEcGCHg9P4DUVtaGjzN5ItI0pkc_JoJZsgI0nvA,127488
+ipex_llm/libs/quantize-starcoder_vnni.exe,sha256=pJoxnrIQs1soH3tJcR_AA7skOX1du7n5kIc3pmQr0AY,128512
+ipex_llm/libs/starcoder-api.dll,sha256=1cs3o08RBH1X79sYCeRLQmBQgGmJnfS4xYU1MnuICAo,21504
+ipex_llm/libs/starcoder.dll,sha256=TkpC8rNFgsh16_jeKtpMaIxpSZ2UTB9bvDh-lElLjNg,599040
 ipex_llm/llamaindex/__init__.py,sha256=T-EbRT6GJ_8RCu-iLmSzcftOimXSPQf2d5X72AUAy2Y,874
 ipex_llm/llamaindex/llms/__init__.py,sha256=KP1lEdGqDuxPoxL1ZSH25Pm2kKMPJBWUTLR0ckSLMIU,1139
 ipex_llm/llamaindex/llms/bigdlllm.py,sha256=FQBzq1KOjfc6uofTXAha3O7TqpJkNfOFepXQmOVlbnI,26314
@@ -86,8 +86,8 @@ ipex_llm/serving/fastchat/model_worker.py,sha256=qJSLyWNkP6z70ysq4AV5SqHzXPJJiX2
 ipex_llm/serving/fastchat/tgi_api_protocol.py,sha256=brT3k3-V0NJrU4fRqUwWjC0O3iOitdttDfduXXEefh0,5918
 ipex_llm/serving/fastchat/tgi_api_server.py,sha256=agNTAEiZPSuj3dEdIdYKwkoY0cXOUDX06DiM9VP2knQ,24418
 ipex_llm/serving/fastchat/vllm_worker.py,sha256=ZLz2Q9GxJO6r_LOiP6epgCRjBGk-K4EB1SNEWSJp5DA,11091
-ipex_llm/transformers/__init__.py,sha256=pJHs6GZXHIObVE4BUCuej-6BKBZZg9pYWKPrkhWSfB4,1192
-ipex_llm/transformers/convert.py,sha256=QhiAVKavNCbrYUAtFiFnQAYaUkG-rB7l0DRj_fDzOFY,101108
+ipex_llm/transformers/__init__.py,sha256=BreA3EY6hkNq0rVixb_sUuTLzMrcWXTt3yRsshCPHHQ,1214
+ipex_llm/transformers/convert.py,sha256=--X5moNSvtV8Mtzg2fh9v1Ej1iyYyocQwo5pHlhkPqo,102230
 ipex_llm/transformers/convert_ipex.py,sha256=_nSnUTQy-yfkKaqGdqnBdWztZf3NGmnbZ0TKaDrF4X4,14617
 ipex_llm/transformers/embedding.py,sha256=bdgk59DvD4ZZyxRzewXOR7g56nThgO6uhIwk8QL7f-s,9299
 ipex_llm/transformers/kv.py,sha256=k4TU18LlA-Sbq9WNNQnfuzu3RSFBwFhmaV3BcGN5bAo,19191
@@ -145,8 +145,9 @@ ipex_llm/transformers/models/chatglm.py,sha256=UHai1t2AUtGmF765_eHF8LUMVQzp_oCBx
 ipex_llm/transformers/models/chatglm2.py,sha256=KyAIX7zGVQDQuwwM3QMBNWZbTeMHEzKUIgAryT0voHc,14933
 ipex_llm/transformers/models/chatglm4.py,sha256=QvUehdaCePB3MNHyWg3dneDxmjtBdxYeKUyQUVcsgfM,16886
 ipex_llm/transformers/models/chatglm4v.py,sha256=L6y45M_wjS2_HqchmCUxRlQZUNuSNCGOiynAQrGh918,14124
-ipex_llm/transformers/models/common.py,sha256=Q3IEfGqvxoHyfIIF5s8qHmOJBBP3b2jyVAXk8C3b1Pg,11636
+ipex_llm/transformers/models/common.py,sha256=VKouwfP3q7nQkYnDeDYfjLwKgr_Qk9uc3ZCISaxbfn4,11646
 ipex_llm/transformers/models/decilm.py,sha256=P-PBuDPf07GvKggLwJx_wPwIn6esN3rX8ai2JxRuZmE,5246
+ipex_llm/transformers/models/deepseek_v3.py,sha256=CTgwIKQlUPlUCbOxc9Id5GapWkXOP6pMtkguYrWpCio,10003
 ipex_llm/transformers/models/gemma.py,sha256=_E3Yw8Y45xyNVeLqyVKcpr8kjuICtETeL82cJ-bWJuU,9424
 ipex_llm/transformers/models/gemma2.py,sha256=2WZuv-FLzJyTJFaYxOuzJt47QE64M0lHnzAiO5T6ozI,8049
 ipex_llm/transformers/models/glm.py,sha256=lmeEWd_W2O638VzVW4Gm6cJre5XZcg_QBmPs8NWqXsM,7202
@@ -222,6 +223,7 @@ ipex_llm/utils/benchmark_util_4_43.py,sha256=VnnLkQCsHe80_suDAAvNUr07OB6ZfOmL1ab
 ipex_llm/utils/benchmark_util_4_44.py,sha256=vN3OKJxln9PGN68yz1W7wS3LKm-x5btLJhhe13TZ8Bw,247085
 ipex_llm/utils/benchmark_util_4_45.py,sha256=r_iwugGhW1GzCkItyniWtBJkZ-xBDZsIcpIj-eNIvuY,247317
 ipex_llm/utils/benchmark_util_4_47.py,sha256=zCBsERwmBKKDPrIG-G8vYP9aEutqkIOpcZhe1ZnMHM4,261335
+ipex_llm/utils/benchmark_util_deepseek.py,sha256=MhhrXb46FBZXCPVaiUZ2hKHJNXmJUoJDk5HVlmLR76E,261577
 ipex_llm/utils/convert_chatglm.py,sha256=YLbpJIbqnQOd8aJSijG5q4xBudNi_QR5sYVjCtYoVkY,18074
 ipex_llm/utils/convert_util.py,sha256=X1eLTdF9crlca6jPEQKymEITRx_57i_87dhl1htPEhc,72612
 ipex_llm/utils/glibc_checker.py,sha256=bm6kN6gbpA7GKtnUgsCE6K16iZZyil-Ylp55SoRvHG8,2093
@@ -246,16 +248,16 @@ ipex_llm/vllm/cpu/entrypoints/openai/cli_args.py,sha256=hB398yYtKauASRzevctScdbF
 ipex_llm/vllm/xpu/__init__.py,sha256=zBSG6nzrVF5QnpR6_f7kPhBFeowTE9gaZ7D5m98E7_w,585
 ipex_llm/vllm/xpu/ipex_llm_v1_wrapper.py,sha256=pd939vFomKIg9Qn2NO4u0OF6hPgvQpqcfJSxqBzcqhA,825
 ipex_llm/vllm/xpu/ipex_llm_wrapper.py,sha256=_CbhvBuf_KPnmLfngYKtJl5gPAHVsG2mWth3wSeaH3M,892
-ipex_llm/vllm/xpu/model_convert.py,sha256=0lfzHhCtTVSZZdlqXAOAEtK3FsouqEheSGr9MjpWz38,8137
+ipex_llm/vllm/xpu/model_convert.py,sha256=4s7OxAUhxC2QpIf_yznrd-AL_Eu-ozaEr3tPTyCgS-c,8882
 ipex_llm/vllm/xpu/engine/__init__.py,sha256=pY_CpyuZd72fr6s32ejeKHKFW0K4vUU2rzZjWqm5YiA,783
-ipex_llm/vllm/xpu/engine/engine.py,sha256=ClvTHz_w52IU_vUw-LhotLduCMu0zPl9kebnQEKQq4A,9494
+ipex_llm/vllm/xpu/engine/engine.py,sha256=NvCMbp0X8NVrOqbwm4FTvXOptTRLzu9jQsy37ZHnTk8,9493
 ipex_llm/vllm/xpu/entrypoints/openai/api_server.py,sha256=OScQFmAdrd-PyEyQ2Mr8fJrk_lC6MHqPraphVk-NDzw,33140
 ipex_llm/vllm/xpu/entrypoints/openai/cli_args.py,sha256=hB398yYtKauASRzevctScdbFIjiiSGMAe1bwEuIHrhY,10893
-ipex_llm-2.2.0b20250218.data/scripts/ipex-llm-init.bat,sha256=HPtCYuDYwEatq7dAwOvdfVcHYCpAVdbj75K1qh0vQek,2578
-ipex_llm-2.2.0b20250218.data/scripts/llm-chat.ps1,sha256=6qrs-hGVAV8IKh7Jx8nq_XrnZcjd7qGU5wndArM7Yag,2769
-ipex_llm-2.2.0b20250218.data/scripts/llm-cli.ps1,sha256=3qBtTLs_EjYDnM8YyCpJhzLnGCKTEGssu9UNqfkjVXs,3009
-ipex_llm-2.2.0b20250218.dist-info/METADATA,sha256=_o3nEy4q2t4hAcjCrxGyZxuVHFYnLbIui9DJg5JTDIQ,12369
-ipex_llm-2.2.0b20250218.dist-info/WHEEL,sha256=6iYPr8vTHsyDK75jr9X0V3I9wPSVmtwr_8fdATBciGk,98
-ipex_llm-2.2.0b20250218.dist-info/entry_points.txt,sha256=TiUyBB2MRmfF3ko-pyAEzqeBCRnyhu27bNOAsWPp3e8,61
-ipex_llm-2.2.0b20250218.dist-info/top_level.txt,sha256=CGCMHM-SyqUabU4h8RqJ2KTYckQUO3LvIWwmUQ6Qbzw,9
-ipex_llm-2.2.0b20250218.dist-info/RECORD,,
+ipex_llm-2.2.0b20250219.data/scripts/ipex-llm-init.bat,sha256=HPtCYuDYwEatq7dAwOvdfVcHYCpAVdbj75K1qh0vQek,2578
+ipex_llm-2.2.0b20250219.data/scripts/llm-chat.ps1,sha256=6qrs-hGVAV8IKh7Jx8nq_XrnZcjd7qGU5wndArM7Yag,2769
+ipex_llm-2.2.0b20250219.data/scripts/llm-cli.ps1,sha256=3qBtTLs_EjYDnM8YyCpJhzLnGCKTEGssu9UNqfkjVXs,3009
+ipex_llm-2.2.0b20250219.dist-info/METADATA,sha256=hCcROHeDK5347fIyeDN0YbL7WQAA3Ql7Ue9LFxtu2Q0,12369
+ipex_llm-2.2.0b20250219.dist-info/WHEEL,sha256=6iYPr8vTHsyDK75jr9X0V3I9wPSVmtwr_8fdATBciGk,98
+ipex_llm-2.2.0b20250219.dist-info/entry_points.txt,sha256=TiUyBB2MRmfF3ko-pyAEzqeBCRnyhu27bNOAsWPp3e8,61
+ipex_llm-2.2.0b20250219.dist-info/top_level.txt,sha256=CGCMHM-SyqUabU4h8RqJ2KTYckQUO3LvIWwmUQ6Qbzw,9
+ipex_llm-2.2.0b20250219.dist-info/RECORD,,