Dmitry Beresnev commited on
Commit
0fbce92
·
1 Parent(s): c261631

fix cmd in dockerfile

Browse files
Files changed (1) hide show
  1. Dockerfile +3 -2
Dockerfile CHANGED
@@ -87,8 +87,9 @@ EXPOSE 7860
87
  # CMD ["llama-server", "-hf", "TheBloke/deepseek-llm-7B-chat-GGUF:deepseek-llm-7b-chat.Q4_K_M.gguf",
88
  # "--host", "0.0.0.0", "--port", "7860", "-c", "4096", "-t", "4", "-ngl", "0", "--cont-batching", "-b", "512"]
89
  #
90
- # Example Qwen2.5 7B Instruct (32k context):
91
- CMD ["llama-server", "-hf", "Qwen/Qwen2.5-7B-Instruct-GGUF:Qwen2.5-7B-Instruct.Q4_K_M.gguf", "--host", "0.0.0.0", "--port", "7860", "-c", "32768", "-t", "4", "-ngl", "0", "--cont-batching", "-b", "512"]
 
92
  #
93
  # Example Qwen2.5-Coder 7B Instruct (32k context):
94
  # CMD ["llama-server", "-hf", "Qwen/Qwen2.5-Coder-7B-Instruct-GGUF:Qwen2.5-Coder-7B-Instruct.Q4_K_M.gguf",
 
87
  # CMD ["llama-server", "-hf", "TheBloke/deepseek-llm-7B-chat-GGUF:deepseek-llm-7b-chat.Q4_K_M.gguf",
88
  # "--host", "0.0.0.0", "--port", "7860", "-c", "4096", "-t", "4", "-ngl", "0", "--cont-batching", "-b", "512"]
89
  #
90
+ # Example Qwen2.5 7B Instruct (8k context, safer on 10GB RAM):
91
+ # Use explicit repo/file to avoid manifest tag resolution issues.
92
+ CMD ["llama-server", "--hf-repo", "Qwen/Qwen2.5-7B-Instruct-GGUF", "--hf-file", "qwen2.5-7b-instruct-q4_k_m.gguf", "--host", "0.0.0.0", "--port", "7860", "-c", "8192", "-t", "4", "-ngl", "0", "--cont-batching", "-b", "256"]
93
  #
94
  # Example Qwen2.5-Coder 7B Instruct (32k context):
95
  # CMD ["llama-server", "-hf", "Qwen/Qwen2.5-Coder-7B-Instruct-GGUF:Qwen2.5-Coder-7B-Instruct.Q4_K_M.gguf",