mediabrain-sjtu/ming

明医 (MING)：中文医疗问诊大模型

明医 (MING)：中文医疗问诊大模型

🌐项目简介

本项目开源了基于医疗指令微调的中文医疗问诊模型：明医 (MING)。目前模型的主要功能如下：

医疗问答：对医疗问题进行解答，对案例进行分析。	智能问诊：多轮问诊后给出诊断结果和建议。

📄相关论文

• MING-MOE技术报告

  MING-MOE: Enhancing Medical Multi-Task Learning in Large Language Models with Sparse Mixture of Low-Rank Adapter Experts

• 基于多智能体交互的大语言模型多轮问诊自动评估框架

  Automatic Interactive Evaluation for Large Language Models with State Aware Patient Simulator

💫更新

• 🔥 [2024/04/14] 开源了基于Qwen1.5指令微调的专家混合模型MING-MOE

• [2024/03/14] 开源了基于Qwen1.5-1.8b指令微调的MING-1.8B

• [2023/07/25] 开源了基于bloomz-7b指令微调的MING-7B

• [2023/07/25] MedicalGPT-zh更名为MING

🔬开源模型

模型	基座	HuggingFace
MING-7B	bloomz-7b1-mt	🤗MING-7B
MING-1.8B	Qwen1.5-1.8B	🤗MING-1.8B
MING-MOE-1.8B	Qwen1.5-1.8B	🤗MING-MOE-1.8B
MING-MOE-4B	Qwen1.5-4B	🤗MING-MOE-4B
MING-MOE-7B	Qwen1.5-7B	🤗MING-MOE-7B
MING-MOE-14B	Qwen1.5-14B	🤗MING-MOE-14B

⚡快速开始

1. 配置环境（测试环境如下，具体版本可以根据实际需求配置）

   • python==3.9.16
   • pytorch==2.0.1+cu117
   • peft==0.9.0

2. 安装项目依赖

   git clone https://github.com/MediaBrain-SJTU/MING
   cd MING
   pip install -e .

3. 下载模型参数并运行（要求单卡显存 >= 15G）

   • MING-MOE

   CUDA_VISIBLE_DEVICES=0 python -m fastchat.serve.cli \
       --model_path {path_to_checkpoint} \ # 模型路径
       --model_base {path_to_base_model} \ # 基座模型路径
       --max-new-token 3072 # 输出最大长度

   • MING-1.8B

   CUDA_VISIBLE_DEVICES=0 python -m fastchat.serve.cli \
       --model_path {path_to_checkpoint} \ # 模型路径
       --max-new-token 2048 # 输出最大长度

   • MING-7B

   CUDA_VISIBLE_DEVICES=0 python -m fastchat.serve.cli \
       --model-path {path_to_checkpoint} \ # 模型路径
       --conv_template bloom \ # prompt
       --max-new-token 512 \ # 输出最大长度
       --beam-size 3 \ # beam search宽度
       --temperature 1.2 # 采样温度

   • 注：由于transformers库的问题，当beam-size > 1时，需要满足temperature>=1.0，否则会报错。

4. 命令行运行实例

   • 对话支持多轮

   • 对话中输入关键词 new chat 能够开启新一轮对话。

🧭测试样例

🪶贡献

本项目由上海交通大学未来媒体网络协同创新中心和上海人工智能实验室智慧医疗中心合作研发。模型数据系统主要由廖育生，江书洋，刘泓呈，孟昱同完成，指导教师为王钰副教授。

免责声明

预训练模型是基于大量语料库和算法模型进行训练的，并且在训练过程中可能存在偏差、错误和不完整的信息。因此，本项目提供的预训练模型仅供参考和研究使用，并不能保证其准确性和可靠性。使用预训练模型产生的结果可能存在误差和偏差，不能用于实际应用或决策。本项目不对使用预训练模型所产生的结果承担任何责任，也不对因使用预训练模型所产生的任何损失承担责任。使用者在使用预训练模型时应自行承担风险并进行自我验证。

引用

如果你使用了本项目的数据或者代码，请声明引用

@misc{liao2024mingmoe,
      title={MING-MOE: Enhancing Medical Multi-Task Learning in Large Language Models with Sparse Mixture of Low-Rank Adapter Experts}, 
      author={Yusheng Liao and Shuyang Jiang and Yu Wang and Yanfeng Wang},
      year={2024},
      eprint={2404.09027},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

Star History

ming's Issues

"模型在哪? 快速开始 中提供的是 chatglm-6b 的模型参数,不是本项目 "MedicalGPT-zh" 的模型参数吧?

"本项目开源了基于ChatGLM-6B LoRA 16-bit指令微调的中文医疗通用模型。模型呢?
按照快速提示提供的步骤无法运行.

laszo@LAPTOP-6MNNHCID:$ . myvenv/bin/activate
(myvenv) laszo@LAPTOP-6MNNHCID:$ cd /mnt/d/dev/code/MedicalGPT-zh/
(myvenv) laszo@LAPTOP-6MNNHCID:/mnt/d/dev/code/MedicalGPT-zh$ cd src
(myvenv) laszo@LAPTOP-6MNNHCID:/mnt/d/dev/code/MedicalGPT-zh/src$ python demo.py
Explicitly passing a revision is encouraged when loading a model with custom code to ensure no malicious code has been contributed in a newer revision.
Loading checkpoint shards: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 8/8 [00:52<00:00, 6.52s/it]
Human:
医生您好，我现在在用一长三速胰岛素注射治疗糖尿病，晚上长效胰岛素注射 12 个单位，速效胰岛素在三餐前注射各6 个单位，血糖不是特别稳，早上空腹的血糖高，晚餐后有时候血糖低，我该如何调
整？
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [32,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [33,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [34,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [35,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [36,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [37,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [38,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [39,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [40,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [41,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [42,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [43,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [44,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [45,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [46,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [47,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [48,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [49,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [50,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [51,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [52,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [53,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [54,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [55,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [56,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [57,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [58,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [59,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [60,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [61,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [62,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [52,0,0], thread: [63,0,0] Assertion srcIndex < srcSelectDimSize failed.
Traceback (most recent call last):
File "/mnt/d/dev/code/MedicalGPT-zh/src/demo.py", line 103, in
main()
File "/mnt/d/dev/code/MedicalGPT-zh/src/demo.py", line 96, in main
response, history = model.chat(tokenizer, query, history=[])
File "/home/laszo/myvenv/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "/mnt/d/dev/code/MedicalGPT-zh/src/model/modeling_chatglm.py", line 1182, in chat
outputs = self.generate(**input_ids, **gen_kwargs)
File "/home/laszo/myvenv/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "/home/laszo/myvenv/lib/python3.10/site-packages/transformers/generation/utils.py", line 1437, in generate
return self.sample(
File "/home/laszo/myvenv/lib/python3.10/site-packages/transformers/generation/utils.py", line 2443, in sample
outputs = self(
File "/home/laszo/myvenv/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "/home/laszo/myvenv/lib/python3.10/site-packages/accelerate/hooks.py", line 165, in new_forward
output = old_forward(*args, **kwargs)
File "/mnt/d/dev/code/MedicalGPT-zh/src/model/modeling_chatglm.py", line 1084, in forward
transformer_outputs = self.transformer(
File "/home/laszo/myvenv/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "/mnt/d/dev/code/MedicalGPT-zh/src/model/modeling_chatglm.py", line 927, in forward
layer_ret = layer(
File "/home/laszo/myvenv/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "/mnt/d/dev/code/MedicalGPT-zh/src/model/modeling_chatglm.py", line 578, in forward
attention_outputs = self.attention(
File "/home/laszo/myvenv/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "/mnt/d/dev/code/MedicalGPT-zh/src/model/modeling_chatglm.py", line 403, in forward
mixed_raw_layer = self.query_key_value(hidden_states)
File "/home/laszo/myvenv/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "/home/laszo/myvenv/lib/python3.10/site-packages/peft/tuners/lora.py", line 565, in forward
result = F.linear(x, transpose(self.weight, self.fan_in_fan_out), bias=self.bias)
RuntimeError: CUDA error: CUBLAS_STATUS_NOT_INITIALIZED when calling cublasCreate(handle)
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [32,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [33,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [34,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [35,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [36,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [37,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [38,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [39,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [40,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [41,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [42,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [43,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [44,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [45,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [46,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [47,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [48,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [49,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [50,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [51,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [52,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [53,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [54,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [55,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [56,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [57,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [58,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [59,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [60,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [61,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [62,0,0] Assertion srcIndex < srcSelectDimSize failed.
../aten/src/ATen/native/cuda/Indexing.cu:1146: indexSelectLargeIndex: block: [47,0,0], thread: [63,0,0] Assertion srcIndex < srcSelectDimSize failed.

想咨询一下，训练数据里有instruction，infer时有instruction吗？如果没有的话，会不会对效果产生影响哪

训练数据

想咨询下，楼主如何保证训练数据的准确率的

使用2张3090跑不起来，能说明一下最低的启动显卡配置么？

使用2张3090跑不起来，能说明一下最低的启动显卡配置么？

如何进行API调用？

目前只能通过终端调用，请问如何通过API调用？

请问为什么选择微调而没有使用大模型+知识库呢

感觉大模型+知识库，在数据更新、解决幻觉问题方面都很有优势
请问当初为什么选择微调而不是大模型+知识库呢
是不是大模型+知识库有什么问题，需要通过微调来解决呢？

数据集可以开源吗

运行模型问题

ming 7b：python -m fastchat.serve.cli
--model-path {path_to_checkpoint} \ # 模型路径
--conv_template bloom \ # prompt
--max-new-token 512 \ # 输出最大长度
--beam-size 3 \ # beam search宽度
--temperature 1.2 # 采样温度
这里终端会报错

关于数据

您好，请问上一个版本的book——data数据集和用chatgpt的阅读文本再生成指令的代码没有了吗

如何在切换场景

我使用 fastchat 分支直接部署，默认的场景是 医疗问答，如何切换为 智能问诊？

如何发起http访问

源码中发布了http模块，如何访问这些http模块呢？我想用http协议和模型交互

关于book_data数据集

请问为什么book_data.json数据集里切分的文本为什么前后不连贯？

Have you tried different Instructions ?

As shown in dialogue_seed_task.json, these instructions are short task classification ："治疗方案"，"病因分析". Are they better than sentence instructions like BelleGroup/train_2M_CN format, such as "请根据以下内容生成治疗方案" ?

bookQA数据

您好，我看BookQA数据的生成代码中，只有Question的部分是依赖医学Context生成。Answer的部分似乎是让ChatGPT直接回答的？而不是基于医学上文来进行回答？
代码如下

prompt = f"指南：\n{input_book}\n"
prompt += f"请根据上述文本中与医学知识相关的内容与逻辑关系提出几个中文问题。注意，提出的问题应该提供充实的内容，使问题具有挑战性。\n"


message = [{"role": "assistant", "content": prompt}]
completion = openai.ChatCompletion.create(
    model= "gpt-3.5-turbo",
    messages= message,
    temperature= 1.0,
    top_p= 1.0,
    frequency_penalty= 0.0,
    presence_penalty= 0.0
)

response = completion.choices[0].message["content"]
questions = parse_response(response)

qa_pairs=[]
for question in questions:
    message = [{"role": "assistant", "content": question}]
    completion = openai.ChatCompletion.create(
        model= "gpt-3.5-turbo",
        messages= message,
        temperature= 1.0,
        top_p= 1.0,
        frequency_penalty= 0.0,
        presence_penalty= 0.0
    )
    answer = completion.choices[0].message["content"]
    qa_pairs.append({'question':question,'answer':answer})

报错

下载模型以后，执行demo报google.protobuf.message.DecodeError: Error parsing message with type 'sentencepiece.ModelProto'

关于训练数据问题

训练数据是否开源？

关于模型转换格式成 gguf

从 BlueZeros/MING-MOE-14B 下载了模型

git clone https://huggingface.co/BlueZeros/MING-MOE-14B ming

后续使用以下命令转换 lora 到 ggml 是成功的

(ollama) ╭─hougelangley at Arch-Legion in ~/ollama on main✘✘✘ 24-04-17 - 22:41:42
╰─(ollama) ⠠⠵ python llm/llama.cpp/convert-lora-to-ggml.py ./ming
model.layers.0.self_attn.k_proj => blk.0.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.0.self_attn.k_proj => blk.0.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.0.self_attn.o_proj => blk.0.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.0.self_attn.o_proj => blk.0.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.0.self_attn.q_proj => blk.0.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.0.self_attn.q_proj => blk.0.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.0.self_attn.v_proj => blk.0.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.0.self_attn.v_proj => blk.0.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.1.self_attn.k_proj => blk.1.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.1.self_attn.k_proj => blk.1.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.1.self_attn.o_proj => blk.1.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.1.self_attn.o_proj => blk.1.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.1.self_attn.q_proj => blk.1.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.1.self_attn.q_proj => blk.1.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.1.self_attn.v_proj => blk.1.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.1.self_attn.v_proj => blk.1.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.10.self_attn.k_proj => blk.10.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.10.self_attn.k_proj => blk.10.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.10.self_attn.o_proj => blk.10.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.10.self_attn.o_proj => blk.10.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.10.self_attn.q_proj => blk.10.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.10.self_attn.q_proj => blk.10.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.10.self_attn.v_proj => blk.10.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.10.self_attn.v_proj => blk.10.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.11.self_attn.k_proj => blk.11.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.11.self_attn.k_proj => blk.11.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.11.self_attn.o_proj => blk.11.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.11.self_attn.o_proj => blk.11.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.11.self_attn.q_proj => blk.11.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.11.self_attn.q_proj => blk.11.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.11.self_attn.v_proj => blk.11.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.11.self_attn.v_proj => blk.11.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.12.self_attn.k_proj => blk.12.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.12.self_attn.k_proj => blk.12.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.12.self_attn.o_proj => blk.12.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.12.self_attn.o_proj => blk.12.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.12.self_attn.q_proj => blk.12.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.12.self_attn.q_proj => blk.12.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.12.self_attn.v_proj => blk.12.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.12.self_attn.v_proj => blk.12.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.13.self_attn.k_proj => blk.13.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.13.self_attn.k_proj => blk.13.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.13.self_attn.o_proj => blk.13.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.13.self_attn.o_proj => blk.13.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.13.self_attn.q_proj => blk.13.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.13.self_attn.q_proj => blk.13.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.13.self_attn.v_proj => blk.13.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.13.self_attn.v_proj => blk.13.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.14.self_attn.k_proj => blk.14.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.14.self_attn.k_proj => blk.14.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.14.self_attn.o_proj => blk.14.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.14.self_attn.o_proj => blk.14.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.14.self_attn.q_proj => blk.14.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.14.self_attn.q_proj => blk.14.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.14.self_attn.v_proj => blk.14.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.14.self_attn.v_proj => blk.14.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.15.self_attn.k_proj => blk.15.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.15.self_attn.k_proj => blk.15.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.15.self_attn.o_proj => blk.15.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.15.self_attn.o_proj => blk.15.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.15.self_attn.q_proj => blk.15.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.15.self_attn.q_proj => blk.15.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.15.self_attn.v_proj => blk.15.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.15.self_attn.v_proj => blk.15.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.16.self_attn.k_proj => blk.16.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.16.self_attn.k_proj => blk.16.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.16.self_attn.o_proj => blk.16.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.16.self_attn.o_proj => blk.16.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.16.self_attn.q_proj => blk.16.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.16.self_attn.q_proj => blk.16.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.16.self_attn.v_proj => blk.16.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.16.self_attn.v_proj => blk.16.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.17.self_attn.k_proj => blk.17.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.17.self_attn.k_proj => blk.17.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.17.self_attn.o_proj => blk.17.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.17.self_attn.o_proj => blk.17.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.17.self_attn.q_proj => blk.17.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.17.self_attn.q_proj => blk.17.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.17.self_attn.v_proj => blk.17.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.17.self_attn.v_proj => blk.17.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.18.self_attn.k_proj => blk.18.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.18.self_attn.k_proj => blk.18.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.18.self_attn.o_proj => blk.18.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.18.self_attn.o_proj => blk.18.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.18.self_attn.q_proj => blk.18.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.18.self_attn.q_proj => blk.18.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.18.self_attn.v_proj => blk.18.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.18.self_attn.v_proj => blk.18.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.19.self_attn.k_proj => blk.19.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.19.self_attn.k_proj => blk.19.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.19.self_attn.o_proj => blk.19.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.19.self_attn.o_proj => blk.19.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.19.self_attn.q_proj => blk.19.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.19.self_attn.q_proj => blk.19.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.19.self_attn.v_proj => blk.19.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.19.self_attn.v_proj => blk.19.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.2.self_attn.k_proj => blk.2.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.2.self_attn.k_proj => blk.2.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.2.self_attn.o_proj => blk.2.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.2.self_attn.o_proj => blk.2.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.2.self_attn.q_proj => blk.2.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.2.self_attn.q_proj => blk.2.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.2.self_attn.v_proj => blk.2.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.2.self_attn.v_proj => blk.2.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.20.self_attn.k_proj => blk.20.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.20.self_attn.k_proj => blk.20.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.20.self_attn.o_proj => blk.20.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.20.self_attn.o_proj => blk.20.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.20.self_attn.q_proj => blk.20.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.20.self_attn.q_proj => blk.20.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.20.self_attn.v_proj => blk.20.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.20.self_attn.v_proj => blk.20.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.21.self_attn.k_proj => blk.21.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.21.self_attn.k_proj => blk.21.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.21.self_attn.o_proj => blk.21.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.21.self_attn.o_proj => blk.21.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.21.self_attn.q_proj => blk.21.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.21.self_attn.q_proj => blk.21.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.21.self_attn.v_proj => blk.21.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.21.self_attn.v_proj => blk.21.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.22.self_attn.k_proj => blk.22.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.22.self_attn.k_proj => blk.22.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.22.self_attn.o_proj => blk.22.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.22.self_attn.o_proj => blk.22.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.22.self_attn.q_proj => blk.22.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.22.self_attn.q_proj => blk.22.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.22.self_attn.v_proj => blk.22.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.22.self_attn.v_proj => blk.22.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.23.self_attn.k_proj => blk.23.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.23.self_attn.k_proj => blk.23.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.23.self_attn.o_proj => blk.23.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.23.self_attn.o_proj => blk.23.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.23.self_attn.q_proj => blk.23.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.23.self_attn.q_proj => blk.23.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.23.self_attn.v_proj => blk.23.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.23.self_attn.v_proj => blk.23.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.24.self_attn.k_proj => blk.24.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.24.self_attn.k_proj => blk.24.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.24.self_attn.o_proj => blk.24.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.24.self_attn.o_proj => blk.24.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.24.self_attn.q_proj => blk.24.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.24.self_attn.q_proj => blk.24.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.24.self_attn.v_proj => blk.24.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.24.self_attn.v_proj => blk.24.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.25.self_attn.k_proj => blk.25.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.25.self_attn.k_proj => blk.25.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.25.self_attn.o_proj => blk.25.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.25.self_attn.o_proj => blk.25.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.25.self_attn.q_proj => blk.25.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.25.self_attn.q_proj => blk.25.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.25.self_attn.v_proj => blk.25.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.25.self_attn.v_proj => blk.25.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.26.self_attn.k_proj => blk.26.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.26.self_attn.k_proj => blk.26.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.26.self_attn.o_proj => blk.26.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.26.self_attn.o_proj => blk.26.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.26.self_attn.q_proj => blk.26.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.26.self_attn.q_proj => blk.26.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.26.self_attn.v_proj => blk.26.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.26.self_attn.v_proj => blk.26.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.27.self_attn.k_proj => blk.27.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.27.self_attn.k_proj => blk.27.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.27.self_attn.o_proj => blk.27.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.27.self_attn.o_proj => blk.27.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.27.self_attn.q_proj => blk.27.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.27.self_attn.q_proj => blk.27.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.27.self_attn.v_proj => blk.27.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.27.self_attn.v_proj => blk.27.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.28.self_attn.k_proj => blk.28.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.28.self_attn.k_proj => blk.28.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.28.self_attn.o_proj => blk.28.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.28.self_attn.o_proj => blk.28.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.28.self_attn.q_proj => blk.28.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.28.self_attn.q_proj => blk.28.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.28.self_attn.v_proj => blk.28.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.28.self_attn.v_proj => blk.28.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.29.self_attn.k_proj => blk.29.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.29.self_attn.k_proj => blk.29.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.29.self_attn.o_proj => blk.29.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.29.self_attn.o_proj => blk.29.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.29.self_attn.q_proj => blk.29.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.29.self_attn.q_proj => blk.29.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.29.self_attn.v_proj => blk.29.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.29.self_attn.v_proj => blk.29.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.3.self_attn.k_proj => blk.3.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.3.self_attn.k_proj => blk.3.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.3.self_attn.o_proj => blk.3.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.3.self_attn.o_proj => blk.3.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.3.self_attn.q_proj => blk.3.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.3.self_attn.q_proj => blk.3.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.3.self_attn.v_proj => blk.3.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.3.self_attn.v_proj => blk.3.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.30.self_attn.k_proj => blk.30.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.30.self_attn.k_proj => blk.30.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.30.self_attn.o_proj => blk.30.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.30.self_attn.o_proj => blk.30.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.30.self_attn.q_proj => blk.30.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.30.self_attn.q_proj => blk.30.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.30.self_attn.v_proj => blk.30.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.30.self_attn.v_proj => blk.30.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.31.self_attn.k_proj => blk.31.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.31.self_attn.k_proj => blk.31.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.31.self_attn.o_proj => blk.31.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.31.self_attn.o_proj => blk.31.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.31.self_attn.q_proj => blk.31.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.31.self_attn.q_proj => blk.31.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.31.self_attn.v_proj => blk.31.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.31.self_attn.v_proj => blk.31.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.32.self_attn.k_proj => blk.32.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.32.self_attn.k_proj => blk.32.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.32.self_attn.o_proj => blk.32.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.32.self_attn.o_proj => blk.32.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.32.self_attn.q_proj => blk.32.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.32.self_attn.q_proj => blk.32.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.32.self_attn.v_proj => blk.32.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.32.self_attn.v_proj => blk.32.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.33.self_attn.k_proj => blk.33.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.33.self_attn.k_proj => blk.33.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.33.self_attn.o_proj => blk.33.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.33.self_attn.o_proj => blk.33.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.33.self_attn.q_proj => blk.33.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.33.self_attn.q_proj => blk.33.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.33.self_attn.v_proj => blk.33.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.33.self_attn.v_proj => blk.33.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.34.self_attn.k_proj => blk.34.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.34.self_attn.k_proj => blk.34.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.34.self_attn.o_proj => blk.34.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.34.self_attn.o_proj => blk.34.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.34.self_attn.q_proj => blk.34.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.34.self_attn.q_proj => blk.34.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.34.self_attn.v_proj => blk.34.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.34.self_attn.v_proj => blk.34.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.35.self_attn.k_proj => blk.35.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.35.self_attn.k_proj => blk.35.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.35.self_attn.o_proj => blk.35.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.35.self_attn.o_proj => blk.35.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.35.self_attn.q_proj => blk.35.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.35.self_attn.q_proj => blk.35.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.35.self_attn.v_proj => blk.35.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.35.self_attn.v_proj => blk.35.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.36.self_attn.k_proj => blk.36.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.36.self_attn.k_proj => blk.36.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.36.self_attn.o_proj => blk.36.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.36.self_attn.o_proj => blk.36.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.36.self_attn.q_proj => blk.36.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.36.self_attn.q_proj => blk.36.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.36.self_attn.v_proj => blk.36.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.36.self_attn.v_proj => blk.36.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.37.self_attn.k_proj => blk.37.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.37.self_attn.k_proj => blk.37.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.37.self_attn.o_proj => blk.37.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.37.self_attn.o_proj => blk.37.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.37.self_attn.q_proj => blk.37.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.37.self_attn.q_proj => blk.37.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.37.self_attn.v_proj => blk.37.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.37.self_attn.v_proj => blk.37.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.38.self_attn.k_proj => blk.38.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.38.self_attn.k_proj => blk.38.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.38.self_attn.o_proj => blk.38.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.38.self_attn.o_proj => blk.38.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.38.self_attn.q_proj => blk.38.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.38.self_attn.q_proj => blk.38.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.38.self_attn.v_proj => blk.38.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.38.self_attn.v_proj => blk.38.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.39.self_attn.k_proj => blk.39.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.39.self_attn.k_proj => blk.39.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.39.self_attn.o_proj => blk.39.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.39.self_attn.o_proj => blk.39.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.39.self_attn.q_proj => blk.39.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.39.self_attn.q_proj => blk.39.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.39.self_attn.v_proj => blk.39.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.39.self_attn.v_proj => blk.39.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.4.self_attn.k_proj => blk.4.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.4.self_attn.k_proj => blk.4.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.4.self_attn.o_proj => blk.4.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.4.self_attn.o_proj => blk.4.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.4.self_attn.q_proj => blk.4.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.4.self_attn.q_proj => blk.4.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.4.self_attn.v_proj => blk.4.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.4.self_attn.v_proj => blk.4.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.5.self_attn.k_proj => blk.5.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.5.self_attn.k_proj => blk.5.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.5.self_attn.o_proj => blk.5.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.5.self_attn.o_proj => blk.5.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.5.self_attn.q_proj => blk.5.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.5.self_attn.q_proj => blk.5.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.5.self_attn.v_proj => blk.5.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.5.self_attn.v_proj => blk.5.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.6.self_attn.k_proj => blk.6.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.6.self_attn.k_proj => blk.6.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.6.self_attn.o_proj => blk.6.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.6.self_attn.o_proj => blk.6.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.6.self_attn.q_proj => blk.6.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.6.self_attn.q_proj => blk.6.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.6.self_attn.v_proj => blk.6.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.6.self_attn.v_proj => blk.6.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.7.self_attn.k_proj => blk.7.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.7.self_attn.k_proj => blk.7.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.7.self_attn.o_proj => blk.7.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.7.self_attn.o_proj => blk.7.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.7.self_attn.q_proj => blk.7.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.7.self_attn.q_proj => blk.7.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.7.self_attn.v_proj => blk.7.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.7.self_attn.v_proj => blk.7.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.8.self_attn.k_proj => blk.8.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.8.self_attn.k_proj => blk.8.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.8.self_attn.o_proj => blk.8.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.8.self_attn.o_proj => blk.8.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.8.self_attn.q_proj => blk.8.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.8.self_attn.q_proj => blk.8.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.8.self_attn.v_proj => blk.8.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.8.self_attn.v_proj => blk.8.attn_v.weight.loraB (5120, 16) float32 0.31MB
model.layers.9.self_attn.k_proj => blk.9.attn_k.weight.loraA (5120, 16) float32 0.31MB
model.layers.9.self_attn.k_proj => blk.9.attn_k.weight.loraB (5120, 16) float32 0.31MB
model.layers.9.self_attn.o_proj => blk.9.attn_output.weight.loraA (5120, 16) float32 0.31MB
model.layers.9.self_attn.o_proj => blk.9.attn_output.weight.loraB (5120, 16) float32 0.31MB
model.layers.9.self_attn.q_proj => blk.9.attn_q.weight.loraA (5120, 16) float32 0.31MB
model.layers.9.self_attn.q_proj => blk.9.attn_q.weight.loraB (5120, 16) float32 0.31MB
model.layers.9.self_attn.v_proj => blk.9.attn_v.weight.loraA (5120, 16) float32 0.31MB
model.layers.9.self_attn.v_proj => blk.9.attn_v.weight.loraB (5120, 16) float32 0.31MB
Converted ./ming/adapter_config.json and ./ming/adapter_model.safetensors to ./ming/ggml-adapter-model.bin

后续需要将 ggml-adapter-model.bin 转换成 gguf 提示如下：

(ollama) ╭─hougelangley at Arch-Legion in ~/ollama on main✘✘✘ 24-04-17 - 22:40:01
╰─(ollama) ⠠⠵ python llm/llama.cpp/convert-llama-ggml-to-gguf.py -i ming/ggml-adapter-model.bin -o ming.bin
* Using config: Namespace(input=PosixPath('ming/ggml-adapter-model.bin'), output=PosixPath('ming.bin'), name=None, desc=None, gqa=1, eps='5.0e-06', context_length=2048, model_metadata_dir=None, vocab_dir=None, vocabtype='spm,hfft')

=== WARNING === Be aware that this conversion script is best-effort. Use a native GGUF model if possible. === WARNING ===

- Note: If converting LLaMA2, specifying "--eps 1e-5" is required. 70B models also need "--gqa 8".
* Scanning GGML input file
Traceback (most recent call last):
  File "/home/hougelangley/ollama/llm/llama.cpp/convert-llama-ggml-to-gguf.py", line 441, in <module>
    main()
  File "/home/hougelangley/ollama/llm/llama.cpp/convert-llama-ggml-to-gguf.py", line 415, in main
    offset = model.load(data, 0)  # noqa
             ^^^^^^^^^^^^^^^^^^^
  File "/home/hougelangley/ollama/llm/llama.cpp/convert-llama-ggml-to-gguf.py", line 175, in load
    offset += self.validate_header(data, offset)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/hougelangley/ollama/llm/llama.cpp/convert-llama-ggml-to-gguf.py", line 160, in validate_header
    raise ValueError(f"Unexpected file magic {magic!r}! This doesn't look like a GGML format file.")
ValueError: Unexpected file magic b'algg'! This doesn't look like a GGML format file.

[Feature Request] Support InternLM

Dear MING developer,

我是 InternLM 社区开发者&志愿者尖米, 大佬开源的工作对我的启发很大，希望可以探讨使用 InternLM 实现 MING 的可能性和实现路径，我的微信是 mzm312，希望可以取得联系进行更深度的交流；

Best regards,
尖米

没有办法进行网页端交互

只能进行终端

数据获取？

您好，请问MedicalGPT-zh 是否会开源？ 如果会开放，请问可以如何获取？