community[minor]: migrate bigdl-llm to ipex-llm (#19518)

- **Description**: `bigdl-llm` library has been renamed to
[`ipex-llm`](https://github.com/intel-analytics/ipex-llm). This PR
migrates the `bigdl-llm` integration to `ipex-llm` .
- **Issue**: N/A. The original PR of `bigdl-llm` is
https://github.com/langchain-ai/langchain/pull/17953
- **Dependencies**: `ipex-llm` library
- **Contribution maintainer**: @shane-huang

Updated doc:   docs/docs/integrations/llms/ipex_llm.ipynb
Updated test:
libs/community/tests/integration_tests/llms/test_ipex_llm.py

docs/docs/integrations/llms/ipex_llm.ipynb

@@ -4,11 +4,11 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "# BigDL-LLM\n",
+    "# IPEX-LLM\n",
     "\n",
-    "> [BigDL-LLM](https://github.com/intel-analytics/BigDL/) is a low-bit LLM optimization library on Intel XPU (Xeon/Core/Flex/Arc/Max). It can make LLMs run extremely fast and consume much less memory on Intel platforms. It is open sourced under Apache 2.0 License.\n",
+    "> [IPEX-LLM](https://github.com/intel-analytics/ipex-llm/) is a low-bit LLM optimization library on Intel XPU (Xeon/Core/Flex/Arc/Max). It can make LLMs run extremely fast and consume much less memory on Intel platforms. It is open sourced under Apache 2.0 License.\n",
     "\n",
-    "This example goes over how to use LangChain to interact with BigDL-LLM for text generation. \n"
+    "This example goes over how to use LangChain to interact with IPEX-LLM for text generation. \n"
    ]
   },
   {
@@ -33,7 +33,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "Install BigDL-LLM for running LLMs locally on Intel CPU."
+    "Install IEPX-LLM for running LLMs locally on Intel CPU."
    ]
   },
   {
@@ -42,8 +42,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "# Install BigDL\n",
+    "%pip install --pre --upgrade ipex-llm[all]"
-    "%pip install --pre --upgrade bigdl-llm[all]"
    ]
   },
   {
@@ -60,7 +59,7 @@
    "outputs": [],
    "source": [
     "from langchain.chains import LLMChain\n",
-    "from langchain_community.llms.bigdl import BigdlLLM\n",
+    "from langchain_community.llms import IpexLLM\n",
     "from langchain_core.prompts import PromptTemplate"
    ]
   },
@@ -89,7 +88,7 @@
     {
      "data": {
       "application/vnd.jupyter.widget-view+json": {
-       "model_id": "69e018750ffb4de1af22ce49cd6957f4",
+       "model_id": "27c08180714a44c7ab766624d5054163",
        "version_major": 2,
        "version_minor": 0
       },
@@ -104,13 +103,12 @@
      "name": "stderr",
      "output_type": "stream",
      "text": [
-      "2024-02-23 18:10:22,896 - INFO - Converting the current model to sym_int4 format......\n",
+      "2024-03-27 00:58:43,670 - INFO - Converting the current model to sym_int4 format......\n"
-      "2024-02-23 18:10:25,415 - INFO - BIGDL_OPT_IPEX: False\n"
      ]
     }
    ],
    "source": [
-    "llm = BigdlLLM.from_model_id(\n",
+    "llm = IpexLLM.from_model_id(\n",
     "    model_id=\"lmsys/vicuna-7b-v1.5\",\n",
     "    model_kwargs={\"temperature\": 0, \"max_length\": 64, \"trust_remote_code\": True},\n",
     ")"
@@ -135,6 +133,10 @@
       "/opt/anaconda3/envs/shane-langchain2/lib/python3.9/site-packages/langchain_core/_api/deprecation.py:117: LangChainDeprecationWarning: The function `run` was deprecated in LangChain 0.1.0 and will be removed in 0.2.0. Use invoke instead.\n",
       "  warn_deprecated(\n",
       "/opt/anaconda3/envs/shane-langchain2/lib/python3.9/site-packages/transformers/generation/utils.py:1369: UserWarning: Using `max_length`'s default (4096) to control the generation length. This behaviour is deprecated and will be removed from the config in v5 of Transformers -- we recommend using `max_new_tokens` to control the maximum length of the generation.\n",
+      "  warnings.warn(\n",
+      "/opt/anaconda3/envs/shane-langchain2/lib/python3.9/site-packages/ipex_llm/transformers/models/llama.py:218: UserWarning: Passing `padding_mask` is deprecated and will be removed in v4.37.Please make sure use `attention_mask` instead.`\n",
+      "  warnings.warn(\n",
+      "/opt/anaconda3/envs/shane-langchain2/lib/python3.9/site-packages/ipex_llm/transformers/models/llama.py:218: UserWarning: Passing `padding_mask` is deprecated and will be removed in v4.37.Please make sure use `attention_mask` instead.`\n",
       "  warnings.warn(\n"
      ]
     },
@@ -156,6 +158,13 @@
     "question = \"What is AI?\"\n",
     "output = llm_chain.run(question)"
    ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
   }
  ],
  "metadata": {

docs/vercel.json

@@ -92,6 +92,10 @@
       "source": "/docs/integrations/llms/huggingface_hub",
       "destination": "/docs/integrations/llms/huggingface_endpoint"
     },
+    {
+      "source": "/docs/integrations/llms/bigdl",
+      "destination": "/docs/integrations/llms/ipex_llm"
+    },
     {
       "source": "/docs/integrations/llms/watsonxllm",
       "destination": "/docs/integrations/llms/ibm_watsonx"

libs/community/langchain_community/llms/__init__.py

@@ -114,6 +114,12 @@ def _import_bedrock() -> Type[BaseLLM]:
     return Bedrock
 
 
+def _import_bigdlllm() -> Type[BaseLLM]:
+    from langchain_community.llms.bigdl_llm import BigdlLLM
+
+    return BigdlLLM
+
+
 def _import_bittensor() -> Type[BaseLLM]:
     from langchain_community.llms.bittensor import NIBittensorLLM
 
@@ -278,6 +284,12 @@ def _import_human() -> Type[BaseLLM]:
     return HumanInputLLM
 
 
+def _import_ipex_llm() -> Type[BaseLLM]:
+    from langchain_community.llms.ipex_llm import IpexLLM
+
+    return IpexLLM
+
+
 def _import_javelin_ai_gateway() -> Type[BaseLLM]:
     from langchain_community.llms.javelin_ai_gateway import JavelinAIGateway
 
@@ -645,6 +657,8 @@ def __getattr__(name: str) -> Any:
         return _import_beam()
     elif name == "Bedrock":
         return _import_bedrock()
+    elif name == "BigdlLLM":
+        return _import_bigdlllm()
     elif name == "NIBittensorLLM":
         return _import_bittensor()
     elif name == "CerebriumAI":
@@ -695,6 +709,8 @@ def __getattr__(name: str) -> Any:
         return _import_huggingface_text_gen_inference()
     elif name == "HumanInputLLM":
         return _import_human()
+    elif name == "IpexLLM":
+        return _import_ipex_llm()
     elif name == "JavelinAIGateway":
         return _import_javelin_ai_gateway()
     elif name == "KoboldApiLLM":
@@ -851,6 +867,7 @@ __all__ = [
     "HuggingFacePipeline",
     "HuggingFaceTextGenInference",
     "HumanInputLLM",
+    "IpexLLM",
     "JavelinAIGateway",
     "KoboldApiLLM",
     "Konko",

libs/community/langchain_community/llms/bigdl_llm.py

@@ -0,0 +1,145 @@
+import logging
+from typing import Any, Optional
+
+from langchain_core.language_models.llms import LLM
+
+from langchain_community.llms.ipex_llm import IpexLLM
+
+logger = logging.getLogger(__name__)
+
+
+class BigdlLLM(IpexLLM):
+    """Wrapper around the BigdlLLM model
+
+    Example:
+        .. code-block:: python
+
+            from langchain_community.llms import BigdlLLM
+            llm = BigdlLLM.from_model_id(model_id="THUDM/chatglm-6b")
+    """
+
+    @classmethod
+    def from_model_id(
+        cls,
+        model_id: str,
+        model_kwargs: Optional[dict] = None,
+        **kwargs: Any,
+    ) -> LLM:
+        """
+        Construct object from model_id
+
+        Args:
+            model_id: Path for the huggingface repo id to be downloaded or
+                      the huggingface checkpoint folder.
+            model_kwargs: Keyword arguments to pass to the model and tokenizer.
+            kwargs: Extra arguments to pass to the model and tokenizer.
+
+        Returns:
+            An object of BigdlLLM.
+        """
+        logger.warning("BigdlLLM was deprecated. Please use IpexLLM instead.")
+
+        try:
+            from bigdl.llm.transformers import (
+                AutoModel,
+                AutoModelForCausalLM,
+            )
+            from transformers import AutoTokenizer, LlamaTokenizer
+
+        except ImportError:
+            raise ValueError(
+                "Could not import bigdl-llm or transformers. "
+                "Please install it with `pip install --pre --upgrade bigdl-llm[all]`."
+            )
+
+        _model_kwargs = model_kwargs or {}
+
+        try:
+            tokenizer = AutoTokenizer.from_pretrained(model_id, **_model_kwargs)
+        except Exception:
+            tokenizer = LlamaTokenizer.from_pretrained(model_id, **_model_kwargs)
+
+        try:
+            model = AutoModelForCausalLM.from_pretrained(
+                model_id, load_in_4bit=True, **_model_kwargs
+            )
+        except Exception:
+            model = AutoModel.from_pretrained(
+                model_id, load_in_4bit=True, **_model_kwargs
+            )
+
+        if "trust_remote_code" in _model_kwargs:
+            _model_kwargs = {
+                k: v for k, v in _model_kwargs.items() if k != "trust_remote_code"
+            }
+
+        return cls(
+            model_id=model_id,
+            model=model,
+            tokenizer=tokenizer,
+            model_kwargs=_model_kwargs,
+            **kwargs,
+        )
+
+    @classmethod
+    def from_model_id_low_bit(
+        cls,
+        model_id: str,
+        model_kwargs: Optional[dict] = None,
+        **kwargs: Any,
+    ) -> LLM:
+        """
+        Construct low_bit object from model_id
+
+        Args:
+
+            model_id: Path for the bigdl-llm transformers low-bit model folder.
+            model_kwargs: Keyword arguments to pass to the model and tokenizer.
+            kwargs: Extra arguments to pass to the model and tokenizer.
+
+        Returns:
+            An object of BigdlLLM.
+        """
+
+        logger.warning("BigdlLLM was deprecated. Please use IpexLLM instead.")
+
+        try:
+            from bigdl.llm.transformers import (
+                AutoModel,
+                AutoModelForCausalLM,
+            )
+            from transformers import AutoTokenizer, LlamaTokenizer
+
+        except ImportError:
+            raise ValueError(
+                "Could not import bigdl-llm or transformers. "
+                "Please install it with `pip install --pre --upgrade bigdl-llm[all]`."
+            )
+
+        _model_kwargs = model_kwargs or {}
+        try:
+            tokenizer = AutoTokenizer.from_pretrained(model_id, **_model_kwargs)
+        except Exception:
+            tokenizer = LlamaTokenizer.from_pretrained(model_id, **_model_kwargs)
+
+        try:
+            model = AutoModelForCausalLM.load_low_bit(model_id, **_model_kwargs)
+        except Exception:
+            model = AutoModel.load_low_bit(model_id, **_model_kwargs)
+
+        if "trust_remote_code" in _model_kwargs:
+            _model_kwargs = {
+                k: v for k, v in _model_kwargs.items() if k != "trust_remote_code"
+            }
+
+        return cls(
+            model_id=model_id,
+            model=model,
+            tokenizer=tokenizer,
+            model_kwargs=_model_kwargs,
+            **kwargs,
+        )
+
+    @property
+    def _llm_type(self) -> str:
+        return "bigdl-llm"

libs/community/langchain_community/llms/ipex_llm.py

@@ -7,17 +7,18 @@ from langchain_core.pydantic_v1 import Extra
 
 DEFAULT_MODEL_ID = "gpt2"
 
+
 logger = logging.getLogger(__name__)
 
 
-class BigdlLLM(LLM):
+class IpexLLM(LLM):
-    """Wrapper around the BigDL-LLM Transformer-INT4 model
+    """Wrapper around the IpexLLM model
 
     Example:
         .. code-block:: python
 
-            from langchain.llms import TransformersLLM
+            from langchain_community.llms import IpexLLM
-            llm = TransformersLLM.from_model_id(model_id="THUDM/chatglm-6b")
+            llm = IpexLLM.from_model_id(model_id="THUDM/chatglm-6b")
     """
 
     model_id: str = DEFAULT_MODEL_ID
@@ -25,7 +26,7 @@ class BigdlLLM(LLM):
     model_kwargs: Optional[dict] = None
     """Keyword arguments passed to the model."""
     model: Any  #: :meta private:
-    """BigDL-LLM Transformers-INT4 model."""
+    """IpexLLM model."""
     tokenizer: Any  #: :meta private:
     """Huggingface tokenizer model."""
     streaming: bool = True
@@ -53,10 +54,10 @@ class BigdlLLM(LLM):
             kwargs: Extra arguments to pass to the model and tokenizer.
 
         Returns:
-            An object of TransformersLLM.
+            An object of IpexLLM.
         """
         try:
-            from bigdl.llm.transformers import (
+            from ipex_llm.transformers import (
                 AutoModel,
                 AutoModelForCausalLM,
             )
@@ -64,8 +65,8 @@ class BigdlLLM(LLM):
 
         except ImportError:
             raise ValueError(
-                "Could not import bigdl-llm or transformers. "
+                "Could not import ipex-llm or transformers. "
-                "Please install it with `pip install --pre --upgrade bigdl-llm[all]`."
+                "Please install it with `pip install --pre --upgrade ipex-llm[all]`."
             )
 
         _model_kwargs = model_kwargs or {}
@@ -109,15 +110,15 @@ class BigdlLLM(LLM):
 
         Args:
 
-            model_id: Path for the bigdl transformers low-bit model checkpoint folder.
+            model_id: Path for the ipex-llm transformers low-bit model folder.
             model_kwargs: Keyword arguments to pass to the model and tokenizer.
             kwargs: Extra arguments to pass to the model and tokenizer.
 
         Returns:
-            An object of TransformersLLM.
+            An object of IpexLLM.
         """
         try:
-            from bigdl.llm.transformers import (
+            from ipex_llm.transformers import (
                 AutoModel,
                 AutoModelForCausalLM,
             )
@@ -125,8 +126,8 @@ class BigdlLLM(LLM):
 
         except ImportError:
             raise ValueError(
-                "Could not import bigdl-llm or transformers. "
+                "Could not import ipex-llm or transformers. "
-                "Please install it with `pip install --pre --upgrade bigdl-llm[all]`"
+                "Please install it with `pip install --pre --upgrade ipex-llm[all]`."
             )
 
         _model_kwargs = model_kwargs or {}
@@ -163,7 +164,7 @@ class BigdlLLM(LLM):
 
     @property
     def _llm_type(self) -> str:
-        return "BigDL-llm"
+        return "ipex-llm"
 
     def _call(
         self,

libs/community/tests/integration_tests/llms/test_bigdl_llm.py

@@ -1,11 +1,11 @@
-"""Test BigDL LLM"""
+"""Test BigdlLLM"""
 from langchain_core.outputs import LLMResult
 
-from langchain_community.llms.bigdl import BigdlLLM
+from langchain_community.llms.bigdl_llm import BigdlLLM
 
 
 def test_call() -> None:
-    """Test valid call to baichuan."""
+    """Test valid call to bigdl-llm."""
     llm = BigdlLLM.from_model_id(
         model_id="lmsys/vicuna-7b-v1.5",
         model_kwargs={"temperature": 0, "max_length": 16, "trust_remote_code": True},
@@ -15,7 +15,7 @@ def test_call() -> None:
 
 
 def test_generate() -> None:
-    """Test valid call to baichuan."""
+    """Test valid call to bigdl-llm."""
     llm = BigdlLLM.from_model_id(
         model_id="lmsys/vicuna-7b-v1.5",
         model_kwargs={"temperature": 0, "max_length": 16, "trust_remote_code": True},

libs/community/tests/integration_tests/llms/test_ipex_llm.py

@@ -0,0 +1,25 @@
+"""Test IPEX LLM"""
+from langchain_core.outputs import LLMResult
+
+from langchain_community.llms.ipex_llm import IpexLLM
+
+
+def test_call() -> None:
+    """Test valid call to ipex-llm."""
+    llm = IpexLLM.from_model_id(
+        model_id="lmsys/vicuna-7b-v1.5",
+        model_kwargs={"temperature": 0, "max_length": 16, "trust_remote_code": True},
+    )
+    output = llm("Hello!")
+    assert isinstance(output, str)
+
+
+def test_generate() -> None:
+    """Test valid call to ipex-llm."""
+    llm = IpexLLM.from_model_id(
+        model_id="lmsys/vicuna-7b-v1.5",
+        model_kwargs={"temperature": 0, "max_length": 16, "trust_remote_code": True},
+    )
+    output = llm.generate(["Hello!"])
+    assert isinstance(output, LLMResult)
+    assert isinstance(output.generations, list)

libs/community/tests/unit_tests/llms/test_imports.py

@@ -42,6 +42,7 @@ EXPECT_ALL = [
     "HuggingFacePipeline",
     "HuggingFaceTextGenInference",
     "HumanInputLLM",
+    "IpexLLM",
     "KoboldApiLLM",
     "Konko",
     "LlamaCpp",