Spaces:
Running
Running
Create Dockerfile
Browse files- Dockerfile +19 -0
Dockerfile
ADDED
|
@@ -0,0 +1,19 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
FROM ghcr.io/ggml-org/llama.cpp:server
|
| 2 |
+
|
| 3 |
+
ENV HOST=0.0.0.0
|
| 4 |
+
ENV PORT=7860
|
| 5 |
+
EXPOSE 7860
|
| 6 |
+
|
| 7 |
+
# llama serve -hf constructai/VibeThinker-3B-GGUF:UD-Q4_K_XL
|
| 8 |
+
CMD ["-hf", "constructai/VibeThinker-3B-GGUF:UD-Q4_K_XL", \
|
| 9 |
+
"--host", "0.0.0.0", \
|
| 10 |
+
"--port", "7860", \
|
| 11 |
+
"--alias", "vibethinker-3b-q4", \
|
| 12 |
+
"--temp", "1.0", \
|
| 13 |
+
"--top-p", "0.95", \
|
| 14 |
+
"--top-k", "1", \
|
| 15 |
+
"--cache-type-k", "q8_0", \
|
| 16 |
+
"-c", "16384"]
|
| 17 |
+
|
| 18 |
+
# https://huggingface.co/WeiboAI/VibeThinker-3B temperature=1.0 top_p=0.95 top_k=-1
|
| 19 |
+
# 65536 scale the value downward to 32768 (32K) or 16384 (16K) if OOM
|