Update requirements.txt

requirements.txt

@@ -3,12 +3,15 @@ uvicorn[standard]>=0.30
 gradio>=4.38
 requests>=2.31
 
-# Match vLLM ↔ Torch versions
+# vLLM & friends
 vllm==0.5.5
+transformers>=4.44
+accelerate>=0.30
 
-# Use CUDA 12.1 wheels for torch 2.4.0
+# Torch for CUDA 12.1 base image
 --extra-index-url https://download.pytorch.org/whl/cu121
 torch==2.4.0
 
-transformers>=4.44
-accelerate>=0.30
+# Pin these to avoid the Numba cache error
+outlines==0.0.46
+numba==0.59.1