什么都没有，直接pip install ollama， 如果是下面的安装位置？

(torch_gpu) rkqq@oem:~/tradegen/llm$ locate ollama
/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/ollama
/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/ollama-0.4.7.dist-info
/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/ollama/__init__.py
/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/ollama/__pycache__
/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/ollama/_client.py
/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/ollama/_types.py
/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/ollama/_utils.py
/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/ollama/py.typed
/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/ollama/__pycache__/__init__.cpython-39.pyc
/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/ollama/__pycache__/_client.cpython-39.pyc
/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/ollama/__pycache__/_types.cpython-39.pyc
/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/ollama/__pycache__/_utils.cpython-39.pyc
/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/ollama-0.4.7.dist-info/INSTALLER
/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/ollama-0.4.7.dist-info/LICENSE
/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/ollama-0.4.7.dist-info/METADATA
/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/ollama-0.4.7.dist-info/RECORD
/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/ollama-0.4.7.dist-info/REQUESTED
/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/ollama-0.4.7.dist-info/WHEEL
/home/rkqq/tradegen/test_ollama.py

但python代码中import ollama可以，但运行模型还是要连网去 C-S 访问，基本也是网络问题慢。pip uninstall ollama 后：

/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/ollama
/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/ollama-0.4.7.dist-info
/home/rkqq/tradegen/test_ollama.py

真正安装是这样，下面这个ollama网站所谓的安装，这个本质还是网络下载一个运行环境，非一切本地服务器+本地客户端, 命令行可以算本地，但没python本地方式。

STEP 1

方式1 ： curl -fsSL https://ollama.com/install.sh | sh

方式2：

curl -C - -fsSL https://ollama.com/install.sh -o install.sh

改：curl -C - -v -fsSL "https://ollama.com/install.sh" -o install.sh 2>&1 | tee curl_download.log

• -C -:     This allows curl to resume a partially downloaded file.
• -v:        This enables verbose output to provide more detail about the download process.
• -fsSL:   These options ensure that curl fails silently on server errors but shows progress and follows redirects.
• -o install.sh:   This specifies that the output should be saved to the file install.sh

这样还是修改install.sh 文件，防止中断，从本地运行 (torch_gpu) rkqq@oem:~/tradegen$ sh install.sh

不管什么方式，最后是方式1成功了，(第2次新硬盘。方式1, 但上网慢的那种)

(torch_gpu) rkqq@oem:~/tradegen$ ollama -v
ollama version is 0.5.7
(torch_gpu) rkqq@oem:~/tradegen$ ollama list
没有模型， https://ollama.com/search ， 找， 7b参数的文件4G大小，最大400G以上，下载和本地运行启动

STEP 2

ollama pull deepseek-r1:7b

ollama run deepseek-r1:7b （直接运行他，会自动先下载）

其他 deepseek-r1:7b 

运行 ollama list, 显示有，具体位置可能：

sudo find / -name "*deepseek*"
/usr/share/ollama/.ollama/models/manifests/registry.ollama.ai/library/deepseek-r1

(torch_gpu) rkqq@oem:~$ ollama run deepseek-r1:7b

他cmd方式运行的结果一个界面，就是等输入文字回车，再回答问题的界面

>>>

提示 Use Ctrl + d or /bye to exit. ；ctrl+C还不行。

STEP 3

代码中运行，

还是pip install ollama, 不然ollama没有提供python的方式只有命令行，

下面结果正常，但还要联网，有点慢

from ollama import chat
from ollama import ChatResponse

response: ChatResponse = chat(model='deepseek-r1:7b', messages=[
  {
    'role': 'user',
    'content': 'Why is the sky blue?',
  },
])

print(response.message.content)

之间要C-S：网络环境就算最简单的可以了, 但不是真正离线工作。

  File "/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/httpx/_client.py", line 914, in send
    response = self._send_handling_auth(
  File "/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/httpx/_client.py", line 942, in _send_handling_auth
    response = self._send_handling_redirects(
  File "/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/httpx/_client.py", line 979, in _send_handling_redirects
    response = self._send_single_request(request)
  File "/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/httpx/_client.py", line 1014, in _send_single_request
    response = transport.handle_request(request)
  File "/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/httpx/_transports/default.py", line 250, in handle_request
    resp = self._pool.handle_request(req)
  File "/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/httpcore/_sync/connection_pool.py", line 268, in handle_request
    raise exc
  File "/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/httpcore/_sync/connection_pool.py", line 251, in handle_request
    response = connection.handle_request(request)
  File "/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/httpcore/_sync/connection.py", line 103, in handle_request
    return self._connection.handle_request(request)
  File "/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/httpcore/_sync/http11.py", line 133, in handle_request
    raise exc
  File "/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/httpcore/_sync/http11.py", line 111, in handle_request
    ) = self._receive_response_headers(**kwargs)
  File "/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/httpcore/_sync/http11.py", line 176, in _receive_response_headers
    event = self._receive_event(timeout=timeout)
  File "/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/httpcore/_sync/http11.py", line 212, in _receive_event
    data = self._network_stream.read(
  File "/home/rkqq/anaconda3/envs/torch_gpu/lib/python3.9/site-packages/httpcore/_backends/sync.py", line 126, in read