模型文件
https://modelscope.cn/models/unsloth/GLM-4.7-Flash-GGUF
如何运行 GLM-4.7-Flash 指南！
https://unsloth.ai/docs/models/glm-4.7-flash
- 禁用重复惩罚。
- 推荐参数：
–temp 0.2 --top-k 50 --top-p 0.95 --min-p 0.01
- jinja模板
[gMASK]<sop>{%-iftools-%}<|system|># Tools You may call one or more functions to assistwiththe user query.You are providedwithfunctionsignatures within<tools></tools>XMLtags:<tools>{%fortoolintools%}{{tool|tojson(ensure_ascii=False)}}{%endfor%}</tools>For eachfunctioncall,output thefunctionname and arguments within the followingXMLformat:<tool_call>{function-name}<arg_key>{arg-key-1}</arg_key><arg_value>{arg-value-1}</arg_value><arg_key>{arg-key-2}</arg_key><arg_value>{arg-value-2}</arg_value>...</tool_call>{%-endif-%}{%-macrovisible_text(content)-%}{%-ifcontent is string-%}{{-content}}{%-elif content is iterable and content is not mapping-%}{%-foritemincontent-%}{%-ifitem is mapping and item.type=='text'-%}{{-item.text}}{%-elif item is string-%}{{-item}}{%-endif-%}{%-endfor-%}{%-else-%}{{-content}}{%-endif-%}{%-endmacro-%}{%-setns=namespace(last_user_index=-1)%}{%-forminmessages%}{%-ifm.role=='user'%}{%setns.last_user_index=loop.index0-%}{%-endif%}{%-endfor%}{%forminmessages%}{%-ifm.role=='user'-%}<|user|>{{visible_text(m.content)}}{%-elif m.role=='assistant'-%}<|assistant|>{%-setreasoning_content=''%}{%-setcontent=visible_text(m.content)%}{%-ifm.reasoning_content is string%}{%-setreasoning_content=m.reasoning_content%}{%-else%}{%-if'</think>'incontent%}{%-setreasoning_content=content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n')%}{%-setcontent=content.split('</think>')[-1].lstrip('\n')%}{%-endif%}{%-endif%}{%-if((clear_thinking is defined and not clear_thinking)or loop.index0>ns.last_user_index)and reasoning_content-%}{{'<think>'+reasoning_content.strip()+'</think>'}}{%-else-%}{{'</think>'}}{%-endif-%}{%-ifcontent.strip()-%}{{content.strip()}}{%-endif-%}{%ifm.tool_calls%}{%fortcinm.tool_calls%}{%-iftc.function%}{%-settc=tc.function%}{%-endif%}{{-'<tool_call>'+tc.name-}}{%set_args=tc.arguments%}{%fork,vin_args.items()%}<arg_key>{{k}}</arg_key><arg_value>{{v|tojson(ensure_ascii=False)ifv is not stringelsev}}</arg_value>{%endfor%}</tool_call>{%endfor%}{%endif%}{%-elif m.role=='tool'-%}{%-ifm.content is string-%}{%-ifloop.firstor(messages[loop.index0-1].role!="tool")%}{{-'<|observation|>'}}{%-endif%}{{-'<tool_response>'}}{{-m.content}}{{-'</tool_response>'}}{%-else-%}<|observation|>{%fortrinm.content%}<tool_response>{{tr.outputiftr.output is definedelsetr}}</tool_response>{%endfor-%}{%endif-%}{%-elif m.role=='system'-%}<|system|>{{visible_text(m.content)}}{%-endif-%}{%-endfor-%}{%-ifadd_generation_prompt-%}<|assistant|>{{-'</think>'if(enable_thinking is defined and not enable_thinking)else'<think>'-}}{%-endif-%}windows版opencode配置lmstudio
- 从github上下载windows版本opencode解压到指定路径：D:\Program Files (x86)\opencode.exe
- 创建调用文件添加到环境变量
- opencode.cmd
@echo off start"""D:\Program Files (x86)\opencode.exe""D:\opencode"%*- opencode.ps1
param([string[]]$arguments)if($arguments){Start-Process-FilePath"D:\Program Files (x86)\opencode.exe"-ArgumentList"D:\opencode",$arguments}else{Start-Process-FilePath"D:\Program Files (x86)\opencode.exe"-ArgumentList"D:\opencode"}- 创建配置文件连接lmstudio，没有则创建：C:\Users\dch\.config\opencode\opencode.json
{"$schema":"https://opencode.ai/config.json","provider":{"vllm-local":{"npm":"@ai-sdk/openai-compatible","options":{"baseURL":"http://localhost:12345/v1"},"models":{"Qwen3-4B-Thinking-2507-AWQ":{"name":"Qwen3-4B-Thinking-2507-AWQ (local)"},"Qwen3-0.6B":{"name":"Qwen3-0.6B (local)"}}},"lmstudio-local":{"npm":"@ai-sdk/openai-compatible","options":{"baseURL":"http://192.168.137.1:1234/v1"},"models":{"lmstudio-community/modelscope.cn_Qwen_Qwen3-8B-GGUF_latest":{"name":"modelscope.cn_Qwen_Qwen3-8B-GGUF_latest (local)"},"lmstudio-community/GLM-4.7-Flash-Q4_K_M":{"name":"glm-4.7-flash (local)"}}}}}- opencode auth login -> 输入Other -> 输入lmstudio-local -> 密钥随便输
- 启动lmstudio模型服务，可以通过浏览器输入http://192.168.137.1:1234/v1/models校验api服务是否成功建立
- opencode -> /models -> 选择lmstudio-local模型进行对话
评价
受硬件约束，推理有点慢，但是整个推理过程和调用工具方面还是可以的。
