From 77d38db00eeba0cfd9b4388f4be858c2f56909a9 Mon Sep 17 00:00:00 2001 From: zR <2448370773@qq.com> Date: Fri, 29 Sep 2023 20:17:29 +0800 Subject: [PATCH] =?UTF-8?q?=E4=BF=AE=E6=94=B9=E9=83=A8=E5=88=86=E9=94=99?= =?UTF-8?q?=E8=AF=AF=E7=9A=84=E6=96=87=E6=A1=A3=EF=BC=8C=E5=8A=A0=E9=95=BF?= =?UTF-8?q?=E4=BA=86=E5=8E=86=E5=8F=B2=E8=AE=B0=E5=BD=95=EF=BC=88for=2032k?= =?UTF-8?q?=20Models=EF=BC=89=20(#1629)?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit * 更新上agent提示词代码 * 更新部分文档，修复了issue中提到的bge匹配超过1 的bug * 按需修改 --- README.md | 23 ++++++--------- README_en.md | 39 +++++++++++++++----------- configs/model_config.py.example | 2 ++ server/knowledge_base/kb_cache/base.py | 2 +- webui_pages/dialogue/dialogue.py | 9 ++++-- 5 files changed, 41 insertions(+), 34 deletions(-) diff --git a/README.md b/README.md index 258069aed..4d5edf25d 100644 --- a/README.md +++ b/README.md @@ -167,6 +167,9 @@ docker run -d --gpus all -p 80:8501 registry.cn-beijing.aliyuncs.com/chatchat/ch - [BAAI/bge-small-zh](https://huggingface.co/BAAI/bge-small-zh) - [BAAI/bge-base-zh](https://huggingface.co/BAAI/bge-base-zh) - [BAAI/bge-large-zh](https://huggingface.co/BAAI/bge-large-zh) +- [BAAI/bge-base-zh-v1.5](https://huggingface.co/BAAI/bge-base-zh-v1.5) +- [BAAI/bge-large-zh-v1.5](https://huggingface.co/BAAI/bge-large-zh-v1.5)- [BAAI/bge-base-zh-v1.5](https://huggingface.co/BAAI/bge-base-zh-v1.5) +- [BAAI/bge-large-zh-v1.5](https://huggingface.co/BAAI/bge-large-zh-v1.5) - [BAAI/bge-large-zh-noinstruct](https://huggingface.co/BAAI/bge-large-zh-noinstruct) - [sensenova/piccolo-base-zh](https://huggingface.co/sensenova/piccolo-base-zh) - [sensenova/piccolo-large-zh](https://huggingface.co/sensenova/piccolo-large-zh) @@ -274,28 +277,20 @@ $ git clone https://huggingface.co/moka-ai/m3e-base 在开始执行 Web UI 或命令行交互前，请先检查 [configs/model_config.py](configs/model_config.py) 和 [configs/server_config.py](configs/server_config.py) 中的各项模型参数设计是否符合需求： - 请确认已下载至本地的 LLM 模型本地存储路径写在 `llm_model_dict` 对应模型的 `local_model_path` 属性中，如: +``` +"chatglm2-6b": "/Users/xxx/Downloads/chatglm2-6b", -```python -llm_model_dict={ - "chatglm2-6b": { - "local_model_path": "/Users/xxx/Downloads/chatglm2-6b", - "api_base_url": "http://localhost:8888/v1", # "name"修改为 FastChat 服务中的"api_base_url" - "api_key": "EMPTY" - }, - } ``` - 请确认已下载至本地的 Embedding 模型本地存储路径写在 `embedding_model_dict` 对应模型位置，如： -```python -embedding_model_dict = { - "m3e-base": "/Users/xxx/Downloads/m3e-base", - } +``` +"m3e-base": "/Users/xxx/Downloads/m3e-base", ``` - 请确认本地分词器路径是否已经填写，如： -```python +``` text_splitter_dict = { "ChineseRecursiveTextSplitter": { "source": "huggingface", ## 选择tiktoken则使用openai的方法,不填写则默认为字符长度切割方法。 @@ -358,7 +353,7 @@ $ python startup.py --all-webui --model-name Qwen-7B-Chat ```python gpus=None, -num_gpus=1, +num_gpus= 1, max_gpu_memory="20GiB" ``` diff --git a/README_en.md b/README_en.md index 8a209164d..72c3a4a8f 100644 --- a/README_en.md +++ b/README_en.md @@ -160,6 +160,8 @@ Following models are tested by developers with Embedding class of [HuggingFace]( - [moka-ai/m3e-large](https://huggingface.co/moka-ai/m3e-large) - [BAAI/bge-small-zh](https://huggingface.co/BAAI/bge-small-zh) - [BAAI/bge-base-zh](https://huggingface.co/BAAI/bge-base-zh) +- [BAAI/bge-base-zh-v1.5](https://huggingface.co/BAAI/bge-base-zh-v1.5) +- [BAAI/bge-large-zh-v1.5](https://huggingface.co/BAAI/bge-large-zh-v1.5) - [BAAI/bge-large-zh](https://huggingface.co/BAAI/bge-large-zh) - [BAAI/bge-large-zh-noinstruct](https://huggingface.co/BAAI/bge-large-zh-noinstruct) - [sensenova/piccolo-base-zh](https://huggingface.co/sensenova/piccolo-base-zh) @@ -228,30 +230,33 @@ $ git clone https://huggingface.co/moka-ai/m3e-base ``` ### 3. Setting Configuration +Copy the model-related parameter configuration template file [configs/model_config.py.example](configs/model_config.py.example) and store it under the project path `. /configs` path and rename it `model_config.py`. -Copy the model-related parameter configuration template file [configs/model_config.py.example](configs/model_config.py.example) and save it in the `./configs` path under the project path, and rename it to `model_config.py`. +Copy the service-related parameter configuration template file [configs/server_config.py.example](configs/server_config.py.example) and store it under the project path `. /configs` path and rename it `server_config.py`. -Copy the service-related parameter configuration template file [configs/server_config.py.example](configs/server_config.py.example) to save in the `./configs` path under the project path, and rename it to `server_config.py`. +Before you start executing the Web UI or command line interactions, check that each of the items in [configs/model_config.py](configs/model_config.py) and [configs/server_config.py](configs/server_config.py) The model parameters are designed to meet the requirements: -Before starting to execute Web UI or command line interaction, please check whether each model parameter in `configs/model_config.py` and `configs/server_config.py` meets the requirements. +- Please make sure that the local storage path of the downloaded LLM model is written in the `local_model_path` attribute of the corresponding model in `llm_model_dict`, e.g.. +``` +"chatglm2-6b":"/Users/xxx/Downloads/chatglm2-6b", + +``` -* Please confirm that the path to local LLM model and embedding model have been written in `llm_dict` of `configs/model_config.py`, here is an example: -* If you choose to use OpenAI's Embedding model, please write the model's ``key`` into `embedding_model_dict`. To use this model, you need to be able to access the OpenAI official API, or set up a proxy. +- Please make sure that the local storage path of the downloaded Embedding model is written in `embedding_model_dict` corresponding to the model location, e.g.: -```python -llm_model_dict={ - "chatglm2-6b": { - "local_model_path": "/Users/xxx/Downloads/chatglm2-6b", - "api_base_url": "http://localhost:8888/v1", # "name"修改为 FastChat 服务中的"api_base_url" - "api_key": "EMPTY" - }, - } +``` +"m3e-base":"/Users/xxx/Downloads/m3e-base", ``` Please make sure that the local storage path of the downloaded Embedding model is written in the location of the corresponding model, e.g. ``` -```python -embedding_model_dict = { - "m3e-base": "/Users/xxx/Downloads/m3e-base", - } +- Please make sure that the local participle path is filled in, e.g.: + +``` +text_splitter_dict = { + "ChineseRecursiveTextSplitter": { + "source": "huggingface", ## Select tiktoken to use openai's method, don't fill it in then it defaults to character length cutting method. + "tokenizer_name_or_path": "", ## Leave blank to use the big model of the tokeniser. + } +} ``` ### 4. Knowledge Base Migration diff --git a/configs/model_config.py.example b/configs/model_config.py.example index 59a1f7eb0..0111fc5d2 100644 --- a/configs/model_config.py.example +++ b/configs/model_config.py.example @@ -30,6 +30,8 @@ MODEL_PATH = { "bge-base-zh": "BAAI/bge-base-zh", "bge-large-zh": "BAAI/bge-large-zh", "bge-large-zh-noinstruct": "BAAI/bge-large-zh-noinstruct", + "bge-base-zh-v1.5": "BAAI/bge-base-zh-v1.5", + "bge-large-zh-v1.5": "BAAI/bge-large-zh-v1.5", "piccolo-base-zh": "sensenova/piccolo-base-zh", "piccolo-large-zh": "sensenova/piccolo-large-zh", "text-embedding-ada-002": "your OPENAI_API_KEY", diff --git a/server/knowledge_base/kb_cache/base.py b/server/knowledge_base/kb_cache/base.py index 59426fa6a..bb4942a79 100644 --- a/server/knowledge_base/kb_cache/base.py +++ b/server/knowledge_base/kb_cache/base.py @@ -133,7 +133,7 @@ def load_embeddings(self, model: str, device: str) -> Embeddings: else: # maybe ReRanker or else, just use empty string instead query_instruction = "" - embeddings = HuggingFaceBgeEmbeddings(model_name=embedding_model_dict[model], + embeddings = HuggingFaceBgeEmbeddings(model_name=get_model_path(model), model_kwargs={'device': device}, query_instruction=query_instruction) if model == "bge-large-zh-noinstruct": # bge large -noinstruct embedding diff --git a/webui_pages/dialogue/dialogue.py b/webui_pages/dialogue/dialogue.py index 05c2a4f7a..c89dbefcc 100644 --- a/webui_pages/dialogue/dialogue.py +++ b/webui_pages/dialogue/dialogue.py @@ -101,8 +101,10 @@ def llm_model_format_func(x): st.success(msg) st.session_state["prev_llm_model"] = llm_model - temperature = st.slider("Temperature：", 0.0, 1.0, TEMPERATURE, 0.05) - history_len = st.number_input("历史对话轮数：", 0, 10, HISTORY_LEN) + temperature = st.slider("Temperature：", 0.0, 1.0, TEMPERATURE, 0.01) + + ## 部分模型可以超过10抡对话 + history_len = st.number_input("历史对话轮数：", 0, 20, HISTORY_LEN) def on_kb_change(): st.toast(f"已加载知识库： {st.session_state.selected_kb}") @@ -117,7 +119,10 @@ def on_kb_change(): key="selected_kb", ) kb_top_k = st.number_input("匹配知识条数：", 1, 20, VECTOR_SEARCH_TOP_K) + + ## Bge 模型会超过1 score_threshold = st.slider("知识匹配分数阈值：", 0.0, 1.0, float(SCORE_THRESHOLD), 0.01) + # chunk_content = st.checkbox("关联上下文", False, disabled=True) # chunk_size = st.slider("关联长度：", 0, 500, 250, disabled=True) elif dialogue_mode == "搜索引擎问答":