{ "cells": [ { "cell_type": "code", "execution_count": 2, "id": "d2ff171c-f5f8-4590-9ce0-21c87e3d5b39", "metadata": {}, "outputs": [], "source": [ "import sys\n", "sys.path.append('/media/gpt4-pdf-chatbot-langchain/langchain-ChatGLM/')\n", "from langchain.llms.base import LLM\n", "import torch\n", "import transformers \n", "import models.shared as shared \n", "from abc import ABC\n", "\n", "from langchain.llms.base import LLM\n", "import random\n", "from transformers.generation.logits_process import LogitsProcessor\n", "from transformers.generation.utils import LogitsProcessorList, StoppingCriteriaList\n", "from typing import Optional, List, Dict, Any\n", "from models.loader import LoaderCheckPoint\n", "from models.extensions.callback import (Iteratorize, Stream, FixedLengthQueue) \n", "from models.base import (BaseAnswer,\n", " AnswerResult,\n", " AnswerResultStream,\n", " AnswerResultQueueSentinelTokenListenerQueue)\n", "from langchain.callbacks.manager import (\n", " CallbackManagerForLLMRun\n", ")\n", "\n" ] }, { "cell_type": "code", "execution_count": 3, "id": "68978c38-c0e9-4ae9-ba90-9c02aca335be", "metadata": {}, "outputs": [ { "name": "stderr", "output_type": "stream", "text": [ "Overriding torch_dtype=None with `torch_dtype=torch.float16` due to requirements of `bitsandbytes` to enable model loading in mixed int8. Either pass torch_dtype=torch.float16 or don't pass this argument at all to remove this warning.\n" ] }, { "name": "stdout", "output_type": "stream", "text": [ "Loading vicuna-13b-hf...\n" ] }, { "data": { "application/vnd.jupyter.widget-view+json": { "model_id": "999ea6baab394a6f9b2b4a815ead7ef4", "version_major": 2, "version_minor": 0 }, "text/plain": [ "Loading checkpoint shards: 0%| | 0/3 [00:00 Dict[str, Any]:\n", " \"\"\"\n", " :return: 响应结构\n", " \"\"\"\n", " return {\n", " \"text\": \"\"\n", " }\n", "\n", "\n", "def _update_response(response: Dict[str, Any], stream_response: str) -> None:\n", " \"\"\"Update response from the stream response.\"\"\"\n", " response[\"text\"] += stream_response\n", "\n", "\n", "class InvalidScoreLogitsProcessor(LogitsProcessor):\n", " def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor) -> torch.FloatTensor:\n", " if torch.isnan(scores).any() or torch.isinf(scores).any():\n", " scores.zero_()\n", " scores[..., 5] = 5e4\n", " return scores\n", "\n", "\n", "class LLamaLLM(BaseAnswer, LLM, ABC):\n", " checkPoint: LoaderCheckPoint = None\n", " history = []\n", " history_len: int = 3\n", " max_new_tokens: int = 500\n", " num_beams: int = 1\n", " temperature: float = 0.5\n", " top_p: float = 0.4\n", " top_k: int = 10\n", " repetition_penalty: float = 1.12\n", " encoder_repetition_penalty: int = 1\n", " min_length: int = 0\n", " logits_processor: LogitsProcessorList = None\n", " stopping_criteria: Optional[StoppingCriteriaList] = None\n", "\n", " state: object = {'max_new_tokens': 50,\n", " 'seed': 1,\n", " 'temperature': 0, 'top_p': 0.1,\n", " 'top_k': 40, 'typical_p': 1,\n", " 'repetition_penalty': 1.18,\n", " 'encoder_repetition_penalty': 1,\n", " 'no_repeat_ngram_size': 0,\n", " 'min_length': 0,\n", " 'penalty_alpha': 0,\n", " 'num_beams': 1,\n", " 'length_penalty': 1,\n", " 'early_stopping': False, 'add_bos_token': True, 'ban_eos_token': False,\n", " 'truncation_length': 2048, 'custom_stopping_strings': '',\n", " 'cpu_memory': 0, 'auto_devices': False, 'disk': False, 'cpu': False, 'bf16': False,\n", " 'load_in_8bit': False, 'wbits': 'None', 'groupsize': 'None', 'model_type': 'None',\n", " 'pre_layer': 0, 'gpu_memory_0': 0}\n", "\n", " def __init__(self, checkPoint: LoaderCheckPoint = None):\n", " super().__init__()\n", " self.checkPoint = checkPoint\n", "\n", " @property\n", " def _llm_type(self) -> str:\n", " return \"LLamaLLM\"\n", "\n", " @property\n", " def _check_point(self) -> LoaderCheckPoint:\n", " return self.checkPoint\n", "\n", " def encode(self, prompt, add_special_tokens=True, add_bos_token=True, truncation_length=None):\n", " input_ids = self.checkPoint.tokenizer.encode(str(prompt), return_tensors='pt',\n", " add_special_tokens=add_special_tokens)\n", " # This is a hack for making replies more creative.\n", " if not add_bos_token and input_ids[0][0] == self.checkPoint.tokenizer.bos_token_id:\n", " input_ids = input_ids[:, 1:]\n", "\n", " # Llama adds this extra token when the first character is '\\n', and this\n", " # compromises the stopping criteria, so we just remove it\n", " if type(self.checkPoint.tokenizer) is transformers.LlamaTokenizer and input_ids[0][0] == 29871:\n", " input_ids = input_ids[:, 1:]\n", "\n", " # Handling truncation\n", " if truncation_length is not None:\n", " input_ids = input_ids[:, -truncation_length:]\n", "\n", " return input_ids.cuda()\n", "\n", " def decode(self, output_ids):\n", " reply = self.checkPoint.tokenizer.decode(output_ids, skip_special_tokens=True)\n", " return reply\n", "\n", " def generate_with_callback(self, callback=None, **kwargs):\n", " self.checkPoint.clear_torch_cache()\n", " kwargs['stopping_criteria'].append(Stream(callback_func=callback))\n", " with torch.no_grad():\n", " self.checkPoint.model.generate(**kwargs)\n", " print(\"方法结束\")\n", "\n", " def generate_with_streaming(self, **kwargs):\n", " return Iteratorize(self.generate_with_callback, kwargs)\n", "\n", " # 将历史对话数组转换为文本格式\n", " def history_to_text(self, query):\n", " formatted_history = ''\n", " history = self.history[-self.history_len:] if self.history_len > 0 else []\n", " for i, (old_query, response) in enumerate(history):\n", " formatted_history += \"[Round {}]\\n问：{}\\n答：{}\\n\".format(i, old_query, response)\n", " formatted_history += \"[Round {}]\\n问：{}\\n答：\".format(len(history), query)\n", " return formatted_history\n", "\n", " def prepare_inputs_for_generation(self,\n", " input_ids: torch.LongTensor):\n", " \"\"\"\n", " 预生成注意力掩码和 输入序列中每个位置的索引的张量\n", " # TODO 没有思路\n", " :return:\n", " \"\"\"\n", "\n", " mask_positions = torch.zeros((1, input_ids.shape[1]), dtype=input_ids.dtype).to(self.checkPoint.model.device)\n", "\n", " attention_mask = self.get_masks(input_ids, input_ids.device)\n", "\n", " position_ids = self.get_position_ids(\n", " input_ids,\n", " device=input_ids.device,\n", " mask_positions=mask_positions\n", " )\n", "\n", " return input_ids, position_ids, attention_mask\n", "\n", " def get_position_ids(self, input_ids: torch.LongTensor, mask_positions, device):\n", " \"\"\"\n", " 注意力偏移量\n", " :param input_ids:\n", " :param mask_positions:\n", " :param device:\n", " :param use_gmasks:\n", " :return:\n", " \"\"\"\n", " batch_size, seq_length = input_ids.shape\n", " context_lengths = [seq.tolist().index(self.checkPoint.model_config.bos_token_id) for seq in input_ids]\n", " position_ids = torch.arange(seq_length, dtype=torch.long, device=device).unsqueeze(0).repeat(batch_size, 1)\n", " for i, context_length in enumerate(context_lengths):\n", " position_ids[i, context_length:] = mask_positions[i]\n", " block_position_ids = [torch.cat((\n", " torch.zeros(context_length, dtype=torch.long, device=device),\n", " torch.arange(seq_length - context_length, dtype=torch.long, device=device) + 1\n", " )) for context_length in context_lengths]\n", " block_position_ids = torch.stack(block_position_ids, dim=0)\n", " position_ids = torch.stack((position_ids, block_position_ids), dim=1)\n", " return position_ids\n", "\n", " def get_masks(self, input_ids, device):\n", " \"\"\"\n", " 获取注意力掩码\n", " :param input_ids:\n", " :param device:\n", " :return:\n", " \"\"\"\n", " batch_size, seq_length = input_ids.shape\n", " context_lengths = [seq.tolist().index(self.checkPoint.model_config.bos_token_id) for seq in input_ids]\n", " attention_mask = torch.ones((batch_size, seq_length, seq_length), device=device)\n", " attention_mask.tril_()\n", " for i, context_length in enumerate(context_lengths):\n", " attention_mask[i, :, :context_length] = 1\n", " attention_mask.unsqueeze_(1)\n", " attention_mask = (attention_mask < 0.5).bool()\n", " return attention_mask\n", "\n", " def generate_softprompt_history_tensors(self, query):\n", " \"\"\"\n", " 历史对话软提示\n", " 这段代码首先定义了一个名为 history_to_text 的函数，用于将 self.history\n", " 数组转换为所需的文本格式。然后，我们将格式化后的历史文本\n", " 再用 self.encode 将其转换为向量表示。最后，将历史对话向量与当前输入的对话向量拼接在一起。\n", " :return:\n", " \"\"\"\n", "\n", " # 对话内容\n", " # 处理历史对话\n", " formatted_history = self.history_to_text(query)\n", " return formatted_history\n", "\n", " def _call(self,\n", " prompt: str,\n", " stop: Optional[List[str]] = None,\n", " run_manager: Optional[CallbackManagerForLLMRun] = None) -> str:\n", " print(f\"__call:{prompt}\")\n", " if self.logits_processor is None:\n", " self.logits_processor = LogitsProcessorList()\n", " self.logits_processor.append(InvalidScoreLogitsProcessor())\n", "\n", " gen_kwargs = {\n", " \"max_new_tokens\": self.max_new_tokens,\n", " \"num_beams\": self.num_beams,\n", " \"top_p\": self.top_p,\n", " \"top_k\": self.top_k,\n", " \"repetition_penalty\": self.repetition_penalty,\n", " \"encoder_repetition_penalty\": self.encoder_repetition_penalty,\n", " \"min_length\": self.min_length,\n", " \"temperature\": self.temperature,\n", " \"logits_processor\": self.logits_processor}\n", "\n", " # 向量拼接\n", " input_ids = self.encode(prompt, add_bos_token=self.state['add_bos_token'], truncation_length=self.max_new_tokens)\n", " # input_ids, position_ids, attention_mask = self.prepare_inputs_for_generation(input_ids=filler_input_ids)\n", "\n", " # 对话模型prompt\n", " gen_kwargs.update({'inputs': input_ids})\n", " # 注意力掩码\n", " # gen_kwargs.update({'attention_mask': attention_mask})\n", " # gen_kwargs.update({'position_ids': position_ids})\n", " if self.stopping_criteria is None:\n", " self.stopping_criteria = transformers.StoppingCriteriaList()\n", " # 观测输出\n", " gen_kwargs.update({'stopping_criteria': self.stopping_criteria})\n", " shared.stop_everything = False\n", " stopped = False\n", " response_template = _streaming_response_template()\n", "\n", " # TODO 此流输出方法需要重写！！！！！！！！！！！！！！！！！！！！！！！！！！！！！！！！！\n", " # stopping_criteria方法不可控制 迭代器的变量无法共享\n", " with self.generate_with_streaming(**gen_kwargs) as generator:\n", " last_reply_len = 0\n", " reply_index = 0\n", " # Create a FixedLengthQueue with the desired stop sequence and a maximum length.\n", " queue = FixedLengthQueue(stop)\n", " for output in generator:\n", " new_tokens = len(output) - len(input_ids[0])\n", " reply = self.decode(output[-new_tokens:])\n", "\n", " new_reply = len(reply) - last_reply_len\n", " output_reply = reply[-new_reply:]\n", " queue.add(reply_index, output_reply)\n", " queue.contains_replace_sequence()\n", " if stop:\n", " pos = queue.contains_stop_sequence()\n", " if pos != -1:\n", " shared.stop_everything = True\n", " stopped = True\n", "\n", " #print(f\"{reply_index}：reply {output_reply}\")\n", " english_reply = queue.put_replace_out(reply_index)\n", " #print(f\"{reply_index}：english_reply {english_reply}\")\n", " _update_response(response_template, english_reply)\n", " last_reply_len = len(reply)\n", "\n", " reply_index += 1\n", " if new_tokens == self.max_new_tokens - 1 or stopped:\n", " break\n", "\n", " response = response_template['text']\n", " print(f\"response:{response}\")\n", " self.history = self.history + [[None, response]]\n", " return response\n", "\n", " def _generate_answer(self, prompt: str,\n", " history: List[List[str]] = [],\n", " streaming: bool = False,\n", " generate_with_callback: AnswerResultStream = None) -> None:\n", " if history:\n", " self.history = history\n", " # Create the StoppingCriteriaList with the stopping strings\n", " self.stopping_criteria = transformers.StoppingCriteriaList()\n", " # 定义模型stopping_criteria 队列，在每次响应时将 torch.LongTensor, torch.FloatTensor同步到AnswerResult\n", " listenerQueue = AnswerResultQueueSentinelTokenListenerQueue()\n", " self.stopping_criteria.append(listenerQueue)\n", " # TODO 需要实现chat对话模块和注意力模型，目前_call为langchain的LLM拓展的api，默认为无提示词模式，如果需要操作注意力模型，可以参考chat_glm的实现\n", " softprompt = self.generate_softprompt_history_tensors(prompt)\n", " response = self._call(prompt=softprompt, stop=['\\n###'])\n", " answer_result = AnswerResult()\n", " answer_result.history = self.history\n", " if listenerQueue.listenerQueue.__len__() > 0:\n", " answer_result.listenerToken = listenerQueue.listenerQueue.pop()\n", " answer_result.llm_output = {\"answer\": response}\n", " generate_with_callback(answer_result)\n" ] }, { "cell_type": "code", "execution_count": 5, "id": "306450e6-b1fd-4c9c-9083-7419094a14f5", "metadata": {}, "outputs": [], "source": [ "\n", "llm_model_ins = LLamaLLM(checkPoint=shared.loaderCheckPoint) " ] }, { "cell_type": "code", "execution_count": 6, "id": "7180ae6a-7050-48a0-b89a-82e51bec7db1", "metadata": {}, "outputs": [], "source": [ "\n", "from langchain.chains import LLMChain\n", "from langchain.memory import ConversationBufferMemory, ReadOnlySharedMemory\n", "from langchain.prompts import PromptTemplate\n", "\n", "template = \"\"\"This is a conversation between a human and a bot:\n", "\n", "{chat_history}\n", "\n", "Write a summary of the conversation for {input}:\n", "\"\"\"\n", "\n", "prompt = PromptTemplate(\n", " input_variables=[\"input\", \"chat_history\"],\n", " template=template\n", ")\n", "memory = ConversationBufferMemory(memory_key=\"chat_history\")\n", "readonlymemory = ReadOnlySharedMemory(memory=memory)\n", "summry_chain = LLMChain(\n", " llm=llm_model_ins,\n", " prompt=prompt,\n", " verbose=True,\n", " memory=readonlymemory, # use the read-only memory to prevent the tool from modifying the memory\n", ")" ] }, { "cell_type": "code", "execution_count": 7, "id": "8516d438-8f9c-4cb7-b921-a7ee216648d5", "metadata": {}, "outputs": [], "source": [ "from langchain.agents import ZeroShotAgent, Tool, AgentExecutor\n", "from typing import List, Set\n", "\n", "\n", "class CustomLLMSingleActionAgent(ZeroShotAgent):\n", " allowed_tools: List[str]\n", "\n", " def __init__(self, *args, **kwargs):\n", " super(CustomLLMSingleActionAgent, self).__init__(*args, **kwargs)\n", " self.allowed_tools = kwargs['allowed_tools']\n", "\n", " def get_allowed_tools(self) -> Set[str]:\n", " return set(self.allowed_tools)\n", " \n", " \n", "tools = [ \n", " Tool(\n", " name=\"Summary\",\n", " func=summry_chain.run,\n", " description=\"useful for when you summarize a conversation. The input to this tool should be a string, representing who will read this summary.\"\n", " )\n", " ]\n", "\n", "prefix = \"\"\"Have a conversation with a human, answering the following questions as best you can. You have access to the following tools:\"\"\"\n", "suffix = \"\"\"Begin!\n", " \n", "Question: {input}\n", "{agent_scratchpad}\"\"\"\n", "\n", "\n", "prompt = ZeroShotAgent.create_prompt(\n", " tools,\n", " prefix=prefix,\n", " suffix=suffix,\n", " input_variables=[\"input\", \"agent_scratchpad\"]\n", ")\n", "tool_names = [tool.name for tool in tools]\n", "llm_chain = LLMChain(llm=llm_model_ins, prompt=prompt)\n", "agent = CustomLLMSingleActionAgent(llm_chain=llm_chain, tools=tools, allowed_tools=tool_names)\n", "agent_chain = AgentExecutor.from_agent_and_tools(agent=agent, tools=tools)" ] }, { "cell_type": "code", "execution_count": 8, "id": "233c3097-66cb-48b0-8de7-d34177e60bd6", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "__call:Have a conversation with a human, answering the following questions as best you can. You have access to the following tools:\n", "\n", "Summary: useful for when you summarize a conversation. The input to this tool should be a string, representing who will read this summary.\n", "\n", "Use the following format:\n", "\n", "Question: the input question you must answer\n", "Thought: you should always think about what to do\n", "Action: the action to take, should be one of [Summary]\n", "Action Input: the input to the action\n", "Observation: the result of the action\n", "... (this Thought/Action/Action Input/Observation can repeat N times)\n", "Thought: I now know the final answer\n", "Final Answer: the final answer to the original input question\n", "\n", "Begin!\n", " \n", "Question: 你好\n", "\n", "response:Thought: 我必��hought: 我必须回��hought: 我必须回答这个问题。\n", "Action: 问��hought: 我必须回答这个问题。\n", "Action: 问她想知道��hought: 我必须回答这个问题。\n", "Action: 问她想知道什么。\n", "Action Input: 没有需要的输入。\n", "Observation:\n", "__call:Have a conversation with a human, answering the following questions as best you can. You have access to the following tools:\n", "\n", "Summary: useful for when you summarize a conversation. The input to this tool should be a string, representing who will read this summary.\n", "\n", "Use the following format:\n", "\n", "Question: the input question you must answer\n", "Thought: you should always think about what to do\n", "Action: the action to take, should be one of [Summary]\n", "Action Input: the input to the action\n", "Observation: the result of the action\n", "... (this Thought/Action/Action Input/Observation can repeat N times)\n", "Thought: I now know the final answer\n", "Final Answer: the final answer to the original input question\n", "\n", "Begin!\n", " \n", "Question: 你好\n", "Thought: 我必��hought: 我必须回��hought: 我必须回答这个问题。\n", "Action: 问��hought: 我必须回答这个问题。\n", "Action: 问她想知道��hought: 我必须回答这个问题。\n", "Action: 问她想知道什么。\n", "Action Input: 没有需要的输入。\n", "Observation:\n", "Observation: 问��hought: 我必须回答这个问题。\n", "Action: 问她想知道��hought: 我必须回答这个问题。\n", "Action: 问她想知道什么。 is not a valid tool, try another one.\n", "Thought:\n", "traceback.print_exc()\n" ] }, { "name": "stderr", "output_type": "stream", "text": [ "Traceback (most recent call last):\n", " File \"/media/gpt4-pdf-chatbot-langchain/langchain-ChatGLM/models/extensions/callback.py\", line 188, in gen\n", " ret = self.mfunc(callback=_callback, **self.kwargs)\n", " File \"/tmp/ipykernel_10879/1809555145.py\", line 114, in generate_with_callback\n", " self.checkPoint.model.generate(**kwargs)\n", " File \"/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/torch/utils/_contextlib.py\", line 115, in decorate_context\n", " return func(*args, **kwargs)\n", " File \"/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/transformers/generation/utils.py\", line 1437, in generate\n", " return self.greedy_search(\n", " File \"/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/transformers/generation/utils.py\", line 2245, in greedy_search\n", " model_inputs = self.prepare_inputs_for_generation(input_ids, **model_kwargs)\n", " File \"/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/transformers/models/llama/modeling_llama.py\", line 737, in prepare_inputs_for_generation\n", " position_ids.masked_fill_(attention_mask == 0, 1)\n", "SystemExit\n" ] }, { "name": "stdout", "output_type": "stream", "text": [ "方法结束\n", "invalid thread id\n", "response:我不能回��不能回答这个问题，因为我是一个人工智能程序。\n", "Final Answer: 我不能回��不能回答这个问题，因为我是一个人工智能程序。\n", "Final Answer: 我不能回答这个问题，因为我是一个人工智能程序。\n" ] }, { "data": { "text/plain": [ "'我不能回答这个问题，因为我是一个人工智能程序。'" ] }, "execution_count": 8, "metadata": {}, "output_type": "execute_result" } ], "source": [ "\n", "agent_chain.run(input=\"你好\")" ] }, { "cell_type": "code", "execution_count": 8, "id": "71a4259f-f589-4aa6-b868-3e10b4f2b22c", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "__call:Have a conversation with a human, answering the following questions as best you can. You have access to the following tools:\n", "\n", "Summary: useful for when you summarize a conversation. The input to this tool should be a string, representing who will read this summary.\n", "\n", "Use the following format:\n", "\n", "Question: the input question you must answer\n", "Thought: you should always think about what to do\n", "Action: the action to take, should be one of [Summary]\n", "Action Input: the input to the action\n", "Observation: the result of the action\n", "... (this Thought/Action/Action Input/Observation can repeat N times)\n", "Thought: I now know the final answer\n", "Final Answer: the final answer to the original input question\n", "\n", "Begin!\n", " \n", "Question: 你是谁?\n", "\n", "response:Thought: 我需要回��hought: 我需要回答这个问题。\n", "Action: ��hought: 我需要回答这个问题。\n", "Action: 执行以下操作:\n", "Action Input: 输入自��hought: 我需要回答这个问题。\n", "Action: 执行以下操作:\n", "Action Input: 输入自己的名字。\n", "Observation:\n", "__call:Have a conversation with a human, answering the following questions as best you can. You have access to the following tools:\n", "\n", "Summary: useful for when you summarize a conversation. The input to this tool should be a string, representing who will read this summary.\n", "\n", "Use the following format:\n", "\n", "Question: the input question you must answer\n", "Thought: you should always think about what to do\n", "Action: the action to take, should be one of [Summary]\n", "Action Input: the input to the action\n", "Observation: the result of the action\n", "... (this Thought/Action/Action Input/Observation can repeat N times)\n", "Thought: I now know the final answer\n", "Final Answer: the final answer to the original input question\n", "\n", "Begin!\n", " \n", "Question: 你是谁?\n", "Thought: 我需要回��hought: 我需要回答这个问题。\n", "Action: ��hought: 我需要回答这个问题。\n", "Action: 执行以下操作:\n", "Action Input: 输入自��hought: 我需要回答这个问题。\n", "Action: 执行以下操作:\n", "Action Input: 输入自己的名字。\n", "Observation:\n", "Observation: ��hought: 我需要回答这个问题。\n", "Action: 执行以下操作: is not a valid tool, try another one.\n", "Thought:\n", "traceback.print_exc()\n" ] }, { "name": "stderr", "output_type": "stream", "text": [ "Traceback (most recent call last):\n", " File \"/media/gpt4-pdf-chatbot-langchain/langchain-ChatGLM/models/extensions/callback.py\", line 188, in gen\n", " ret = self.mfunc(callback=_callback, **self.kwargs)\n", " File \"/tmp/ipykernel_13699/1809555145.py\", line 114, in generate_with_callback\n", " self.checkPoint.model.generate(**kwargs)\n", " File \"/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/torch/utils/_contextlib.py\", line 115, in decorate_context\n", " return func(*args, **kwargs)\n", " File \"/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/transformers/generation/utils.py\", line 1437, in generate\n", " return self.greedy_search(\n", " File \"/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/transformers/generation/utils.py\", line 2245, in greedy_search\n", " model_inputs = self.prepare_inputs_for_generation(input_ids, **model_kwargs)\n", " File \"/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/transformers/models/llama/modeling_llama.py\", line 737, in prepare_inputs_for_generation\n", " position_ids.masked_fill_(attention_mask == 0, 1)\n", "SystemExit\n" ] }, { "name": "stdout", "output_type": "stream", "text": [ "方法结束\n", "invalid thread id\n", "response:我不知道该如何回��不知道该如何回答这个问题。\n", "Final Answer: 我不知道该如何回��不知道该如何回答这个问题。\n", "Final Answer: 我不知道该如何回答这个问题。\n" ] }, { "data": { "text/plain": [ "'我不知道该如何回答这个问题。'" ] }, "execution_count": 8, "metadata": {}, "output_type": "execute_result" } ], "source": [ "agent_chain.run(input=\"你是谁?\")" ] }, { "cell_type": "code", "execution_count": 9, "id": "e0cd4b1d-a7c8-432f-93d4-ef78b3dd851a", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "__call:Have a conversation with a human, answering the following questions as best you can. You have access to the following tools:\n", "\n", "Summary: useful for when you summarize a conversation. The input to this tool should be a string, representing who will read this summary.\n", "\n", "Use the following format:\n", "\n", "Question: the input question you must answer\n", "Thought: you should always think about what to do\n", "Action: the action to take, should be one of [Summary]\n", "Action Input: the input to the action\n", "Observation: the result of the action\n", "... (this Thought/Action/Action Input/Observation can repeat N times)\n", "Thought: I now know the final answer\n", "Final Answer: the final answer to the original input question\n", "\n", "Begin!\n", " \n", "Question: 我们之前聊了什么?\n", "\n", "response:Thought: 我需要回��hought: 我需要回顾我们之前的对话以确定问题的��hought: 我需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: ��hought: 我需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: 检查我们之前的对话记录\n", "Action Input: 无\n", "Observation:\n", "__call:Have a conversation with a human, answering the following questions as best you can. You have access to the following tools:\n", "\n", "Summary: useful for when you summarize a conversation. The input to this tool should be a string, representing who will read this summary.\n", "\n", "Use the following format:\n", "\n", "Question: the input question you must answer\n", "Thought: you should always think about what to do\n", "Action: the action to take, should be one of [Summary]\n", "Action Input: the input to the action\n", "Observation: the result of the action\n", "... (this Thought/Action/Action Input/Observation can repeat N times)\n", "Thought: I now know the final answer\n", "Final Answer: the final answer to the original input question\n", "\n", "Begin!\n", " \n", "Question: 我们之前聊了什么?\n", "Thought: 我需要回��hought: 我需要回顾我们之前的对话以确定问题的��hought: 我需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: ��hought: 我需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: 检查我们之前的对话记录\n", "Action Input: 无\n", "Observation:\n", "Observation: ��hought: 我需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: 检查我们之前的对话记录 is not a valid tool, try another one.\n", "Thought:\n", "traceback.print_exc()\n" ] }, { "name": "stderr", "output_type": "stream", "text": [ "Traceback (most recent call last):\n", " File \"/media/gpt4-pdf-chatbot-langchain/langchain-ChatGLM/models/extensions/callback.py\", line 188, in gen\n", " ret = self.mfunc(callback=_callback, **self.kwargs)\n", " File \"/tmp/ipykernel_13699/1809555145.py\", line 114, in generate_with_callback\n", " self.checkPoint.model.generate(**kwargs)\n", " File \"/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/torch/utils/_contextlib.py\", line 115, in decorate_context\n", " return func(*args, **kwargs)\n", " File \"/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/transformers/generation/utils.py\", line 1437, in generate\n", " return self.greedy_search(\n", " File \"/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/transformers/generation/utils.py\", line 2245, in greedy_search\n", " model_inputs = self.prepare_inputs_for_generation(input_ids, **model_kwargs)\n", " File \"/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/transformers/models/llama/modeling_llama.py\", line 737, in prepare_inputs_for_generation\n", " position_ids.masked_fill_(attention_mask == 0, 1)\n", "SystemExit\n" ] }, { "name": "stdout", "output_type": "stream", "text": [ "response:我需要回��需要回顾我们之前的对话以确定问题的��需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: ��hought: 我需要回��需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: ��hought: 我需要回顾我们之前的对话以确定问题的��需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: ��hought: 我需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: ��需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: ��hought: 我需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: 找出我们之前��需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: ��hought: 我需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: 找出我们之前聊的主题\n", "Action Input: 无\n", "Observation:\n", "__call:Have a conversation with a human, answering the following questions as best you can. You have access to the following tools:\n", "\n", "Summary: useful for when you summarize a conversation. The input to this tool should be a string, representing who will read this summary.\n", "\n", "Use the following format:\n", "\n", "Question: the input question you must answer\n", "Thought: you should always think about what to do\n", "Action: the action to take, should be one of [Summary]\n", "Action Input: the input to the action\n", "Observation: the result of the action\n", "... (this Thought/Action/Action Input/Observation can repeat N times)\n", "Thought: I now know the final answer\n", "Final Answer: the final answer to the original input question\n", "\n", "Begin!\n", " \n", "Question: 我们之前聊了什么?\n", "Thought: 我需要回��hought: 我需要回顾我们之前的对话以确定问题的��hought: 我需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: ��hought: 我需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: 检查我们之前的对话记录\n", "Action Input: 无\n", "Observation:\n", "Observation: ��hought: 我需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: 检查我们之前的对话记录 is not a valid tool, try another one.\n", "Thought:我需要回��需要回顾我们之前的对话以确定问题的��需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: ��hought: 我需要回��需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: ��hought: 我需要回顾我们之前的对话以确定问题的��需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: ��hought: 我需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: ��需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: ��hought: 我需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: 找出我们之前��需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: ��hought: 我需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: 找出我们之前聊的主题\n", "Action Input: 无\n", "Observation:\n", "Observation: ��hought: 我需要回��需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: ��hought: 我需要回顾我们之前的对话以确定问题的��需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: ��hought: 我需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: ��需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: ��hought: 我需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: 找出我们之前��需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: ��hought: 我需要回顾我们之前的对话以确定问题的具体内容。\n", "Action: 找出我们之前聊的主题 is not a valid tool, try another one.\n", "Thought:\n", "traceback.print_exc()\n" ] }, { "name": "stderr", "output_type": "stream", "text": [ "Traceback (most recent call last):\n", " File \"/media/gpt4-pdf-chatbot-langchain/langchain-ChatGLM/models/extensions/callback.py\", line 188, in gen\n", " ret = self.mfunc(callback=_callback, **self.kwargs)\n", " File \"/tmp/ipykernel_13699/1809555145.py\", line 114, in generate_with_callback\n", " self.checkPoint.model.generate(**kwargs)\n", " File \"/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/torch/utils/_contextlib.py\", line 115, in decorate_context\n", " return func(*args, **kwargs)\n", " File \"/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/transformers/generation/utils.py\", line 1437, in generate\n", " return self.greedy_search(\n", " File \"/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/transformers/generation/utils.py\", line 2245, in greedy_search\n", " model_inputs = self.prepare_inputs_for_generation(input_ids, **model_kwargs)\n", " File \"/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/transformers/models/llama/modeling_llama.py\", line 737, in prepare_inputs_for_generation\n", " position_ids.masked_fill_(attention_mask == 0, 1)\n", "SystemExit\n" ] }, { "name": "stdout", "output_type": "stream", "text": [ "response:������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������\n", "traceback.print_exc()\n" ] }, { "name": "stderr", "output_type": "stream", "text": [ "Traceback (most recent call last):\n", " File \"/media/gpt4-pdf-chatbot-langchain/langchain-ChatGLM/models/extensions/callback.py\", line 188, in gen\n", " ret = self.mfunc(callback=_callback, **self.kwargs)\n", " File \"/tmp/ipykernel_13699/1809555145.py\", line 114, in generate_with_callback\n", " self.checkPoint.model.generate(**kwargs)\n", " File \"/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/torch/utils/_contextlib.py\", line 115, in decorate_context\n", " return func(*args, **kwargs)\n", " File \"/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/transformers/generation/utils.py\", line 1437, in generate\n", " return self.greedy_search(\n", " File \"/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/transformers/generation/utils.py\", line 2245, in greedy_search\n", " model_inputs = self.prepare_inputs_for_generation(input_ids, **model_kwargs)\n", " File \"/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/transformers/models/llama/modeling_llama.py\", line 739, in prepare_inputs_for_generation\n", " position_ids = position_ids[:, -1].unsqueeze(-1)\n", "SystemExit\n" ] }, { "data": { "text/html": [ "
╭─────────────────────────────── Traceback (most recent call last) ────────────────────────────────╮\n",
       "│ /tmp/ipykernel_13699/3633147670.py:1 in <module>                                                 │\n",
       "│                                                                                                  │\n",
       "│ [Errno 2] No such file or directory: '/tmp/ipykernel_13699/3633147670.py'                        │\n",
       "│                                                                                                  │\n",
       "│ /media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/langchain/chains/ │\n",
       "│ base.py:241 in run                                                                               │\n",
       "│                                                                                                  │\n",
       "│   238 │   │   │   return self(args[0], callbacks=callbacks)[self.output_keys[0]]                 │\n",
       "│   239 │   │                                                                                      │\n",
       "│   240 │   │   if kwargs and not args:                                                            │\n",
       "│ ❱ 241 │   │   │   return self(kwargs, callbacks=callbacks)[self.output_keys[0]]                  │\n",
       "│   242 │   │                                                                                      │\n",
       "│   243 │   │   raise ValueError(                                                                  │\n",
       "│   244 │   │   │   f\"`run` supported with either positional arguments or keyword arguments\"       │\n",
       "│                                                                                                  │\n",
       "│ /media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/langchain/chains/ │\n",
       "│ base.py:142 in __call__                                                                          │\n",
       "│                                                                                                  │\n",
       "│   139 │   │   │   )                                                                              │\n",
       "│   140 │   │   except (KeyboardInterrupt, Exception) as e:                                        │\n",
       "│   141 │   │   │   run_manager.on_chain_error(e)                                                  │\n",
       "│ ❱ 142 │   │   │   raise e                                                                        │\n",
       "│   143 │   │   run_manager.on_chain_end(outputs)                                                  │\n",
       "│   144 │   │   return self.prep_outputs(inputs, outputs, return_only_outputs)                     │\n",
       "│   145                                                                                            │\n",
       "│                                                                                                  │\n",
       "│ /media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/langchain/chains/ │\n",
       "│ base.py:136 in __call__                                                                          │\n",
       "│                                                                                                  │\n",
       "│   133 │   │   )                                                                                  │\n",
       "│   134 │   │   try:                                                                               │\n",
       "│   135 │   │   │   outputs = (                                                                    │\n",
       "│ ❱ 136 │   │   │   │   self._call(inputs, run_manager=run_manager)                                │\n",
       "│   137 │   │   │   │   if new_arg_supported                                                       │\n",
       "│   138 │   │   │   │   else self._call(inputs)                                                    │\n",
       "│   139 │   │   │   )                                                                              │\n",
       "│                                                                                                  │\n",
       "│ /media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/langchain/agents/ │\n",
       "│ agent.py:904 in _call                                                                            │\n",
       "│                                                                                                  │\n",
       "│   901 │   │   start_time = time.time()                                                           │\n",
       "│   902 │   │   # We now enter the agent loop (until it returns something).                        │\n",
       "│   903 │   │   while self._should_continue(iterations, time_elapsed):                             │\n",
       "│ ❱ 904 │   │   │   next_step_output = self._take_next_step(                                       │\n",
       "│   905 │   │   │   │   name_to_tool_map,                                                          │\n",
       "│   906 │   │   │   │   color_mapping,                                                             │\n",
       "│   907 │   │   │   │   inputs,                                                                    │\n",
       "│                                                                                                  │\n",
       "│ /media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/langchain/agents/ │\n",
       "│ agent.py:748 in _take_next_step                                                                  │\n",
       "│                                                                                                  │\n",
       "│   745 │   │   │   )                                                                              │\n",
       "│   746 │   │   except Exception as e:                                                             │\n",
       "│   747 │   │   │   if not self.handle_parsing_errors:                                             │\n",
       "│ ❱ 748 │   │   │   │   raise e                                                                    │\n",
       "│   749 │   │   │   text = str(e).split(\"`\")[1]                                                    │\n",
       "│   750 │   │   │   observation = \"Invalid or incomplete response\"                                 │\n",
       "│   751 │   │   │   output = AgentAction(\"_Exception\", observation, text)                          │\n",
       "│                                                                                                  │\n",
       "│ /media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/langchain/agents/ │\n",
       "│ agent.py:741 in _take_next_step                                                                  │\n",
       "│                                                                                                  │\n",
       "│   738 │   │   \"\"\"                                                                                │\n",
       "│   739 │   │   try:                                                                               │\n",
       "│   740 │   │   │   # Call the LLM to see what to do.                                              │\n",
       "│ ❱ 741 │   │   │   output = self.agent.plan(                                                      │\n",
       "│   742 │   │   │   │   intermediate_steps,                                                        │\n",
       "│   743 │   │   │   │   callbacks=run_manager.get_child() if run_manager else None,                │\n",
       "│   744 │   │   │   │   **inputs,                                                                  │\n",
       "│                                                                                                  │\n",
       "│ /media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/langchain/agents/ │\n",
       "│ agent.py:426 in plan                                                                             │\n",
       "│                                                                                                  │\n",
       "│   423 │   │   \"\"\"                                                                                │\n",
       "│   424 │   │   full_inputs = self.get_full_inputs(intermediate_steps, **kwargs)                   │\n",
       "│   425 │   │   full_output = self.llm_chain.predict(callbacks=callbacks, **full_inputs)           │\n",
       "│ ❱ 426 │   │   return self.output_parser.parse(full_output)                                       │\n",
       "│   427 │                                                                                          │\n",
       "│   428 │   async def aplan(                                                                       │\n",
       "│   429 │   │   self,                                                                              │\n",
       "│                                                                                                  │\n",
       "│ /media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/langchain/agents/ │\n",
       "│ mrkl/output_parser.py:26 in parse                                                                │\n",
       "│                                                                                                  │\n",
       "│   23 │   │   )                                                                                   │\n",
       "│   24 │   │   match = re.search(regex, text, re.DOTALL)                                           │\n",
       "│   25 │   │   if not match:                                                                       │\n",
       "│ ❱ 26 │   │   │   raise OutputParserException(f\"Could not parse LLM output: `{text}`\")            │\n",
       "│   27 │   │   action = match.group(1).strip()                                                     │\n",
       "│   28 │   │   action_input = match.group(2)                                                       │\n",
       "│   29 │   │   return AgentAction(action, action_input.strip(\" \").strip('\"'), text)                │\n",
       "╰──────────────────────────────────────────────────────────────────────────────────────────────────╯\n",
       "OutputParserException: Could not parse LLM output: \n",
       "`������������������������������������������������������������������������������������������������������������������\n",
       "�������������������������������������������������������������������������������������������������������������������\n",
       "�������������������������������������������������������������������������������������������������������������������\n",
       "�������������������������������������������������������������������������������������������������������������������\n",
       "�������������������������������������������������������������������������������������������������������������������\n",
       "�������������������������������������������������������������������������������������������������������������������\n",
       "�������������������������������������������������������������������������������������������������������������������\n",
       "�������������������������������������������������������������������������������������������������������������������\n",
       "�����������������������������������������������������������������������������`\n",
       "
\n" ], "text/plain": [ "\u001b[31m╭─\u001b[0m\u001b[31m──────────────────────────────\u001b[0m\u001b[31m \u001b[0m\u001b[1;31mTraceback \u001b[0m\u001b[1;2;31m(most recent call last)\u001b[0m\u001b[31m \u001b[0m\u001b[31m───────────────────────────────\u001b[0m\u001b[31m─╮\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2;33m/tmp/ipykernel_13699/\u001b[0m\u001b[1;33m3633147670.py\u001b[0m:\u001b[94m1\u001b[0m in \u001b[92m\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[3;31m[Errno 2] No such file or directory: '/tmp/ipykernel_13699/3633147670.py'\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2;33m/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/langchain/chains/\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[1;33mbase.py\u001b[0m:\u001b[94m241\u001b[0m in \u001b[92mrun\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m238 \u001b[0m\u001b[2m│ │ │ \u001b[0m\u001b[94mreturn\u001b[0m \u001b[96mself\u001b[0m(args[\u001b[94m0\u001b[0m], callbacks=callbacks)[\u001b[96mself\u001b[0m.output_keys[\u001b[94m0\u001b[0m]] \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m239 \u001b[0m\u001b[2m│ │ \u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m240 \u001b[0m\u001b[2m│ │ \u001b[0m\u001b[94mif\u001b[0m kwargs \u001b[95mand\u001b[0m \u001b[95mnot\u001b[0m args: \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[31m❱ \u001b[0m241 \u001b[2m│ │ │ \u001b[0m\u001b[94mreturn\u001b[0m \u001b[96mself\u001b[0m(kwargs, callbacks=callbacks)[\u001b[96mself\u001b[0m.output_keys[\u001b[94m0\u001b[0m]] \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m242 \u001b[0m\u001b[2m│ │ \u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m243 \u001b[0m\u001b[2m│ │ \u001b[0m\u001b[94mraise\u001b[0m \u001b[96mValueError\u001b[0m( \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m244 \u001b[0m\u001b[2m│ │ │ \u001b[0m\u001b[33mf\u001b[0m\u001b[33m\"\u001b[0m\u001b[33m`run` supported with either positional arguments or keyword arguments\u001b[0m\u001b[33m\"\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2;33m/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/langchain/chains/\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[1;33mbase.py\u001b[0m:\u001b[94m142\u001b[0m in \u001b[92m__call__\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m139 \u001b[0m\u001b[2m│ │ │ \u001b[0m) \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m140 \u001b[0m\u001b[2m│ │ \u001b[0m\u001b[94mexcept\u001b[0m (\u001b[96mKeyboardInterrupt\u001b[0m, \u001b[96mException\u001b[0m) \u001b[94mas\u001b[0m e: \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m141 \u001b[0m\u001b[2m│ │ │ \u001b[0mrun_manager.on_chain_error(e) \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[31m❱ \u001b[0m142 \u001b[2m│ │ │ \u001b[0m\u001b[94mraise\u001b[0m e \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m143 \u001b[0m\u001b[2m│ │ \u001b[0mrun_manager.on_chain_end(outputs) \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m144 \u001b[0m\u001b[2m│ │ \u001b[0m\u001b[94mreturn\u001b[0m \u001b[96mself\u001b[0m.prep_outputs(inputs, outputs, return_only_outputs) \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m145 \u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2;33m/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/langchain/chains/\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[1;33mbase.py\u001b[0m:\u001b[94m136\u001b[0m in \u001b[92m__call__\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m133 \u001b[0m\u001b[2m│ │ \u001b[0m) \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m134 \u001b[0m\u001b[2m│ │ \u001b[0m\u001b[94mtry\u001b[0m: \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m135 \u001b[0m\u001b[2m│ │ │ \u001b[0moutputs = ( \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[31m❱ \u001b[0m136 \u001b[2m│ │ │ │ \u001b[0m\u001b[96mself\u001b[0m._call(inputs, run_manager=run_manager) \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m137 \u001b[0m\u001b[2m│ │ │ │ \u001b[0m\u001b[94mif\u001b[0m new_arg_supported \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m138 \u001b[0m\u001b[2m│ │ │ │ \u001b[0m\u001b[94melse\u001b[0m \u001b[96mself\u001b[0m._call(inputs) \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m139 \u001b[0m\u001b[2m│ │ │ \u001b[0m) \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2;33m/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/langchain/agents/\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[1;33magent.py\u001b[0m:\u001b[94m904\u001b[0m in \u001b[92m_call\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m901 \u001b[0m\u001b[2m│ │ \u001b[0mstart_time = time.time() \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m902 \u001b[0m\u001b[2m│ │ \u001b[0m\u001b[2m# We now enter the agent loop (until it returns something).\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m903 \u001b[0m\u001b[2m│ │ \u001b[0m\u001b[94mwhile\u001b[0m \u001b[96mself\u001b[0m._should_continue(iterations, time_elapsed): \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[31m❱ \u001b[0m904 \u001b[2m│ │ │ \u001b[0mnext_step_output = \u001b[96mself\u001b[0m._take_next_step( \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m905 \u001b[0m\u001b[2m│ │ │ │ \u001b[0mname_to_tool_map, \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m906 \u001b[0m\u001b[2m│ │ │ │ \u001b[0mcolor_mapping, \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m907 \u001b[0m\u001b[2m│ │ │ │ \u001b[0minputs, \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2;33m/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/langchain/agents/\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[1;33magent.py\u001b[0m:\u001b[94m748\u001b[0m in \u001b[92m_take_next_step\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m745 \u001b[0m\u001b[2m│ │ │ \u001b[0m) \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m746 \u001b[0m\u001b[2m│ │ \u001b[0m\u001b[94mexcept\u001b[0m \u001b[96mException\u001b[0m \u001b[94mas\u001b[0m e: \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m747 \u001b[0m\u001b[2m│ │ │ \u001b[0m\u001b[94mif\u001b[0m \u001b[95mnot\u001b[0m \u001b[96mself\u001b[0m.handle_parsing_errors: \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[31m❱ \u001b[0m748 \u001b[2m│ │ │ │ \u001b[0m\u001b[94mraise\u001b[0m e \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m749 \u001b[0m\u001b[2m│ │ │ \u001b[0mtext = \u001b[96mstr\u001b[0m(e).split(\u001b[33m\"\u001b[0m\u001b[33m`\u001b[0m\u001b[33m\"\u001b[0m)[\u001b[94m1\u001b[0m] \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m750 \u001b[0m\u001b[2m│ │ │ \u001b[0mobservation = \u001b[33m\"\u001b[0m\u001b[33mInvalid or incomplete response\u001b[0m\u001b[33m\"\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m751 \u001b[0m\u001b[2m│ │ │ \u001b[0moutput = AgentAction(\u001b[33m\"\u001b[0m\u001b[33m_Exception\u001b[0m\u001b[33m\"\u001b[0m, observation, text) \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2;33m/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/langchain/agents/\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[1;33magent.py\u001b[0m:\u001b[94m741\u001b[0m in \u001b[92m_take_next_step\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m738 \u001b[0m\u001b[2;33m│ │ \u001b[0m\u001b[33m\"\"\"\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m739 \u001b[0m\u001b[2m│ │ \u001b[0m\u001b[94mtry\u001b[0m: \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m740 \u001b[0m\u001b[2m│ │ │ \u001b[0m\u001b[2m# Call the LLM to see what to do.\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[31m❱ \u001b[0m741 \u001b[2m│ │ │ \u001b[0moutput = \u001b[96mself\u001b[0m.agent.plan( \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m742 \u001b[0m\u001b[2m│ │ │ │ \u001b[0mintermediate_steps, \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m743 \u001b[0m\u001b[2m│ │ │ │ \u001b[0mcallbacks=run_manager.get_child() \u001b[94mif\u001b[0m run_manager \u001b[94melse\u001b[0m \u001b[94mNone\u001b[0m, \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m744 \u001b[0m\u001b[2m│ │ │ │ \u001b[0m**inputs, \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2;33m/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/langchain/agents/\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[1;33magent.py\u001b[0m:\u001b[94m426\u001b[0m in \u001b[92mplan\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m423 \u001b[0m\u001b[2;33m│ │ \u001b[0m\u001b[33m\"\"\"\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m424 \u001b[0m\u001b[2m│ │ \u001b[0mfull_inputs = \u001b[96mself\u001b[0m.get_full_inputs(intermediate_steps, **kwargs) \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m425 \u001b[0m\u001b[2m│ │ \u001b[0mfull_output = \u001b[96mself\u001b[0m.llm_chain.predict(callbacks=callbacks, **full_inputs) \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[31m❱ \u001b[0m426 \u001b[2m│ │ \u001b[0m\u001b[94mreturn\u001b[0m \u001b[96mself\u001b[0m.output_parser.parse(full_output) \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m427 \u001b[0m\u001b[2m│ \u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m428 \u001b[0m\u001b[2m│ \u001b[0m\u001b[94masync\u001b[0m \u001b[94mdef\u001b[0m \u001b[92maplan\u001b[0m( \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m429 \u001b[0m\u001b[2m│ │ \u001b[0m\u001b[96mself\u001b[0m, \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2;33m/media/gpt4-pdf-chatbot-langchain/pyenv-langchain/lib/python3.10/site-packages/langchain/agents/\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2;33mmrkl/\u001b[0m\u001b[1;33moutput_parser.py\u001b[0m:\u001b[94m26\u001b[0m in \u001b[92mparse\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m23 \u001b[0m\u001b[2m│ │ \u001b[0m) \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m24 \u001b[0m\u001b[2m│ │ \u001b[0mmatch = re.search(regex, text, re.DOTALL) \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m25 \u001b[0m\u001b[2m│ │ \u001b[0m\u001b[94mif\u001b[0m \u001b[95mnot\u001b[0m match: \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[31m❱ \u001b[0m26 \u001b[2m│ │ │ \u001b[0m\u001b[94mraise\u001b[0m OutputParserException(\u001b[33mf\u001b[0m\u001b[33m\"\u001b[0m\u001b[33mCould not parse LLM output: `\u001b[0m\u001b[33m{\u001b[0mtext\u001b[33m}\u001b[0m\u001b[33m`\u001b[0m\u001b[33m\"\u001b[0m) \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m27 \u001b[0m\u001b[2m│ │ \u001b[0maction = match.group(\u001b[94m1\u001b[0m).strip() \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m28 \u001b[0m\u001b[2m│ │ \u001b[0maction_input = match.group(\u001b[94m2\u001b[0m) \u001b[31m│\u001b[0m\n", "\u001b[31m│\u001b[0m \u001b[2m29 \u001b[0m\u001b[2m│ │ \u001b[0m\u001b[94mreturn\u001b[0m AgentAction(action, action_input.strip(\u001b[33m\"\u001b[0m\u001b[33m \u001b[0m\u001b[33m\"\u001b[0m).strip(\u001b[33m'\u001b[0m\u001b[33m\"\u001b[0m\u001b[33m'\u001b[0m), text) \u001b[31m│\u001b[0m\n", "\u001b[31m╰──────────────────────────────────────────────────────────────────────────────────────────────────╯\u001b[0m\n", "\u001b[1;91mOutputParserException: \u001b[0mCould not parse LLM output: \n", "`������������������������������������������������������������������������������������������������������������������\n", "�������������������������������������������������������������������������������������������������������������������\n", "�������������������������������������������������������������������������������������������������������������������\n", "�������������������������������������������������������������������������������������������������������������������\n", "�������������������������������������������������������������������������������������������������������������������\n", "�������������������������������������������������������������������������������������������������������������������\n", "�������������������������������������������������������������������������������������������������������������������\n", "�������������������������������������������������������������������������������������������������������������������\n", "�����������������������������������������������������������������������������`\n" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "agent_chain.run(input=\"我们之前聊了什么?\")" ] }, { "cell_type": "code", "execution_count": null, "id": "096f0378-b7eb-46b6-9985-95c3f3521ecd", "metadata": {}, "outputs": [], "source": [] }, { "cell_type": "code", "execution_count": null, "id": "a3c964f3-fb14-40bd-8df9-ae3c452260bb", "metadata": {}, "outputs": [], "source": [] }, { "cell_type": "code", "execution_count": null, "id": "e31b13c0-af47-4614-93d0-090a69579cc1", "metadata": {}, "outputs": [], "source": [] } ], "metadata": { "kernelspec": { "display_name": "Python 3 (ipykernel)", "language": "python", "name": "python3" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.10.9" } }, "nbformat": 4, "nbformat_minor": 5 }