mirror of
https://github.com/modelscope/modelscope.git
synced 2025-12-16 16:27:45 +01:00
Add a QA example based on llamaindex only (#759)
This commit is contained in:
williamcc
@@ -0,0 +1,223 @@
|
||||
{
|
||||
"cells": [
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"source": [
|
||||
"# Usage\n",
|
||||
"\n",
|
||||
"## 1. Install necessary libs\n",
|
||||
"```shell\n",
|
||||
"!pip install modelscope\n",
|
||||
"!pip install transformers -U\n",
|
||||
"!pip install llama-index llama-index-llms-huggingface ipywidgets \n",
|
||||
"```\n",
|
||||
"\n",
|
||||
"## 2. Download data files we need in this example\n",
|
||||
"```shell\n",
|
||||
"!wget https://modelscope.oss-cn-beijing.aliyuncs.com/resource/rag/punkt.zip\n",
|
||||
"!wget https://modelscope.oss-cn-beijing.aliyuncs.com/resource/rag/stopwords.zip\n",
|
||||
"!wget https://modelscope.oss-cn-beijing.aliyuncs.com/resource/rag/xianjiaoda.md\n",
|
||||
"\n",
|
||||
"!mkdir -p /opt/conda/lib/python3.10/site-packages/llama_index/core/_static/nltk_cache/tokenizers\n",
|
||||
"!mkdir -p /opt/conda/lib/python3.10/site-packages/llama_index/core/_static/nltk_cache/corpora\n",
|
||||
"\n",
|
||||
"!cp /mnt/workspace/punkt.zip /opt/conda/lib/python3.10/site-packages/llama_index/core/_static/nltk_cache/tokenizers\n",
|
||||
"!cp /mnt/workspace/stopwords.zip /opt/conda/lib/python3.10/site-packages/llama_index/core/_static/nltk_cache/corpora\n",
|
||||
"!cd /opt/conda/lib/python3.10/site-packages/llama_index/core/_static/nltk_cache/tokenizers; unzip punkt.zip;\n",
|
||||
"!cd /opt/conda/lib/python3.10/site-packages/llama_index/core/_static/nltk_cache/corpora; unzip stopwords.zip;\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"!mkdir -p /mnt/workspace/custom_data\n",
|
||||
"!mv /mnt/workspace/xianjiaoda.md /mnt/workspace/custom_data\n",
|
||||
"\n",
|
||||
"!cd /mnt/workspace\n",
|
||||
"```\n",
|
||||
"\n",
|
||||
"## 3. Go!"
|
||||
],
|
||||
"metadata": {
|
||||
"collapsed": false
|
||||
},
|
||||
"id": "f4abc589d9bfffca"
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"!pip install modelscope\n",
|
||||
"!pip install transformers -U\n",
|
||||
"!pip install llama-index llama-index-llms-huggingface ipywidgets "
|
||||
],
|
||||
"metadata": {
|
||||
"collapsed": false
|
||||
},
|
||||
"id": "c32122833dd7b8c8"
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"!wget https://modelscope.oss-cn-beijing.aliyuncs.com/resource/rag/punkt.zip\n",
|
||||
"!wget https://modelscope.oss-cn-beijing.aliyuncs.com/resource/rag/stopwords.zip\n",
|
||||
"!wget https://modelscope.oss-cn-beijing.aliyuncs.com/resource/rag/xianjiaoda.md\n",
|
||||
"\n",
|
||||
"!mkdir -p /opt/conda/lib/python3.10/site-packages/llama_index/core/_static/nltk_cache/tokenizers\n",
|
||||
"!mkdir -p /opt/conda/lib/python3.10/site-packages/llama_index/core/_static/nltk_cache/corpora\n",
|
||||
"\n",
|
||||
"!cp /mnt/workspace/punkt.zip /opt/conda/lib/python3.10/site-packages/llama_index/core/_static/nltk_cache/tokenizers\n",
|
||||
"!cp /mnt/workspace/stopwords.zip /opt/conda/lib/python3.10/site-packages/llama_index/core/_static/nltk_cache/corpora\n",
|
||||
"!cd /opt/conda/lib/python3.10/site-packages/llama_index/core/_static/nltk_cache/tokenizers; unzip punkt.zip;\n",
|
||||
"!cd /opt/conda/lib/python3.10/site-packages/llama_index/core/_static/nltk_cache/corpora; unzip stopwords.zip;\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"!mkdir -p /mnt/workspace/custom_data\n",
|
||||
"!mv /mnt/workspace/xianjiaoda.md /mnt/workspace/custom_data\n",
|
||||
"\n",
|
||||
"!cd /mnt/workspace"
|
||||
],
|
||||
"metadata": {
|
||||
"collapsed": false
|
||||
},
|
||||
"id": "63704e2b21a9ba52"
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"import logging\n",
|
||||
"import sys\n",
|
||||
"from abc import ABC\n",
|
||||
"from typing import Any, List\n",
|
||||
"\n",
|
||||
"import torch\n",
|
||||
"from llama_index.core import (\n",
|
||||
" SimpleDirectoryReader,\n",
|
||||
" VectorStoreIndex,\n",
|
||||
" Settings,\n",
|
||||
" ServiceContext,\n",
|
||||
" set_global_service_context,\n",
|
||||
")\n",
|
||||
"from llama_index.core.base.embeddings.base import BaseEmbedding, Embedding\n",
|
||||
"from llama_index.core.prompts import PromptTemplate\n",
|
||||
"from llama_index.llms.huggingface import HuggingFaceLLM\n",
|
||||
"\n",
|
||||
"from modelscope import snapshot_download\n",
|
||||
"\n",
|
||||
"logging.basicConfig(stream=sys.stdout, level=logging.INFO)\n",
|
||||
"logging.getLogger().addHandler(logging.StreamHandler(stream=sys.stdout))\n",
|
||||
"\n",
|
||||
"# download QWEN model from modelscope\n",
|
||||
"qwen15_4B_CHAT = \"qwen/Qwen1.5-4B-Chat\"\n",
|
||||
"selected_model = snapshot_download(qwen15_4B_CHAT)\n",
|
||||
"\n",
|
||||
"# define sys prompt\n",
|
||||
"SYSTEM_PROMPT = \"\"\"You are a helpful AI assistant.\"\"\"\n",
|
||||
"query_wrapper_prompt = PromptTemplate(\n",
|
||||
" \"[INST]<<SYS>>\\n\" + SYSTEM_PROMPT + \"<</SYS>>\\n\\n{query_str}[/INST] \"\n",
|
||||
")\n",
|
||||
"\n",
|
||||
"# create HuggingFaceLLM with qwen1.5 \n",
|
||||
"llm = HuggingFaceLLM(\n",
|
||||
" context_window=4096,\n",
|
||||
" max_new_tokens=2048,\n",
|
||||
" generate_kwargs={\"temperature\": 0.0, \"do_sample\": False},\n",
|
||||
" query_wrapper_prompt=query_wrapper_prompt,\n",
|
||||
" tokenizer_name=selected_model,\n",
|
||||
" model_name=selected_model,\n",
|
||||
" device_map=\"auto\",\n",
|
||||
" # change these settings below depending on your GPU\n",
|
||||
" model_kwargs={\"torch_dtype\": torch.float16},\n",
|
||||
")\n",
|
||||
"print(\"llm created\")\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"# wrap modelscope embedding for llama-index (based on BaseEmbedding)\n",
|
||||
"class ModelScopeEmbeddings4LlamaIndex(BaseEmbedding, ABC):\n",
|
||||
" embed: Any = None\n",
|
||||
" model_id: str = \"damo/nlp_gte_sentence-embedding_chinese-base\"\n",
|
||||
"\n",
|
||||
" def __init__(\n",
|
||||
" self,\n",
|
||||
" model_id: str,\n",
|
||||
" **kwargs: Any,\n",
|
||||
" ) -> None:\n",
|
||||
" super().__init__(**kwargs)\n",
|
||||
" try:\n",
|
||||
" from modelscope.models import Model\n",
|
||||
" from modelscope.pipelines import pipeline\n",
|
||||
" from modelscope.utils.constant import Tasks\n",
|
||||
" # 使用modelscope的embedding模型（包含下载）\n",
|
||||
" self.embed = pipeline(Tasks.sentence_embedding, model=self.model_id)\n",
|
||||
"\n",
|
||||
" except ImportError as e:\n",
|
||||
" raise ValueError(\n",
|
||||
" \"Could not import some python packages.\" \"Please install it with `pip install modelscope`.\"\n",
|
||||
" ) from e\n",
|
||||
"\n",
|
||||
" def _get_query_embedding(self, query: str) -> Embedding:\n",
|
||||
" text = query.replace(\"\\n\", \" \")\n",
|
||||
" inputs = {\"source_sentence\": [text]}\n",
|
||||
" # note that we have to call tolist() to change numpy.ndarray into python list\n",
|
||||
" return self.embed(input=inputs)['text_embedding'][0].tolist()\n",
|
||||
"\n",
|
||||
" def _get_text_embedding(self, text: str) -> Embedding:\n",
|
||||
" text = text.replace(\"\\n\", \" \")\n",
|
||||
" inputs = {\"source_sentence\": [text]}\n",
|
||||
" return self.embed(input=inputs)['text_embedding'][0].tolist()\n",
|
||||
"\n",
|
||||
" def _get_text_embeddings(self, texts: List[str]) -> List[Embedding]:\n",
|
||||
" texts = list(map(lambda x: x.replace(\"\\n\", \" \"), texts))\n",
|
||||
" inputs = {\"source_sentence\": texts}\n",
|
||||
" return self.embed(input=inputs)['text_embedding'].tolist()\n",
|
||||
"\n",
|
||||
" async def _aget_query_embedding(self, query: str) -> Embedding:\n",
|
||||
" return self._get_query_embedding(query)\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"embedding_model = \"damo/nlp_gte_sentence-embedding_chinese-base\"\n",
|
||||
"embeddings = ModelScopeEmbeddings4LlamaIndex(model_id=embedding_model)\n",
|
||||
"service_context = ServiceContext.from_defaults(embed_model=embeddings, llm=llm)\n",
|
||||
"set_global_service_context(service_context)\n",
|
||||
"Settings.embed_model = embeddings\n",
|
||||
"\n",
|
||||
"# load example documents\n",
|
||||
"documents = SimpleDirectoryReader(\"/mnt/workspace/custom_data/\").load_data()\n",
|
||||
"\n",
|
||||
"# create Vector DB\n",
|
||||
"index = VectorStoreIndex.from_documents(documents)\n",
|
||||
"\n",
|
||||
"# set Logging to DEBUG for more detailed outputs\n",
|
||||
"query_engine = index.as_query_engine()\n",
|
||||
"\n",
|
||||
"# do query\n",
|
||||
"response = query_engine.query(\"西安较大的校训是什么\")\n",
|
||||
"print(response)\n"
|
||||
],
|
||||
"metadata": {
|
||||
"collapsed": false
|
||||
},
|
||||
"id": "eef67659e94045c5"
|
||||
}
|
||||
],
|
||||
"metadata": {
|
||||
"kernelspec": {
|
||||
"display_name": "Python 3",
|
||||
"language": "python",
|
||||
"name": "python3"
|
||||
},
|
||||
"language_info": {
|
||||
"codemirror_mode": {
|
||||
"name": "ipython",
|
||||
"version": 2
|
||||
},
|
||||
"file_extension": ".py",
|
||||
"mimetype": "text/x-python",
|
||||
"name": "python",
|
||||
"nbconvert_exporter": "python",
|
||||
"pygments_lexer": "ipython2",
|
||||
"version": "2.7.6"
|
||||
}
|
||||
},
|
||||
"nbformat": 4,
|
||||
"nbformat_minor": 5
|
||||
}
|
||||
Reference in New Issue
Block a user