ollama默认的上下文长度只有2048,所以想要改一下上下文长度,根据官方文档,两种方式可更改,通过设定启动命令
/set parameter num_ctx 32000
或者api传参:
curl http://localhost:11434/api/generate -d '{
"model": "llama3",
"prompt": "Why is the sky blue?",
"options": {
"num_ctx": 4096
}
}'
但是这也太不好用了,就没啥一劳永逸的方法吗,所以我找到了这个:ollama修改模型问答的上下文长度(num_ctx)
非常好,我在windows上成功执行并且得到了新模型,但是linux上执行
ollama create -f Modelfile qwen2:ctx32k
这一句的时候报错:
sudo ollama create -f Modelfile qwen2:7bctx32k
transferring model data
using existing layer sha256:43f7a214e5329f672bb05404cfba1913cbb70fdaa1a17497224e1925046b0ed5
using existing layer sha256:c156170b718ec29139d3653d40ed1986fd92fb7e0959b5c71f3c48f62e6636f4
using existing layer sha256:62fbfd9ed093d6e5ac83190c86eec5369317919f4b149598d2dbb38900e9faef
using existing layer sha256:c156170b718ec29139d3653d40ed1986fd92fb7e0959b5c71f3c48f62e6636f4
using existing layer sha256:c758d8ee94ceaa226d6d79d055f4f2ab8bb748e6655d5ac82116767c5156723c
using existing layer sha256:40eda8bd4b0ef2cff20be2421242e98866b60be41d626f97ab9dbeb152b4761a
writing manifest
Error: open /usr/share/ollama/.ollama/models/manifests/registry.ollama.ai/library/qwen2/7bctx32k: permission denied
经查询,附解决办法如下:
sudo find /usr/share/ollama/.ollama/models/ -type f -exec chown ollama:ollama {} \;
sudo find /usr/share/ollama/.ollama/models/ -type d -exec chown ollama:ollama {} \;
sudo find /usr/share/ollama/.ollama/models/ -type f -exec chmod 644 {} \;
sudo find /usr/share/ollama/.ollama/models/ -type d -exec chmod 755 {} \;