Update Dockerfile

Dockerfile

@@ -1,4 +1,4 @@
-# Use an official PyTorch image as a base
+# Use an official PyTorch image as a base (includes CUDA for GPU support if available)
 FROM pytorch/pytorch:2.1.0-cuda11.8-cudnn8-runtime
 
 # Set the working directory inside the container
@@ -7,24 +7,8 @@ WORKDIR /app
 # Install system dependencies
 RUN apt-get update && apt-get install -y \
     build-essential \
-    git \
     && rm -rf /var/lib/apt/lists/*
 
-# Create a non-root user
-RUN useradd -m -u 1000 appuser
-
-# Create cache directories and set permissions
-RUN mkdir -p /app/models/sentence_transformer && \
-    mkdir -p /app/models/llm && \
-    mkdir -p /.cache/huggingface && \
-    mkdir -p /.cache/torch && \
-    mkdir -p /.cache/sentence_transformers
-
-# Set environment variables for cache and model locations
-ENV HF_HOME="/.cache/huggingface"
-ENV TORCH_HOME="/.cache/torch"
-ENV SENTENCE_TRANSFORMERS_HOME="/app/models/sentence_transformer"
-
 # Install Python dependencies
 RUN pip install --no-cache-dir \
     pandas \
@@ -37,97 +21,17 @@ RUN pip install --no-cache-dir \
     uvicorn[standard] \
     pydantic \
     python-multipart \
-    huggingface_hub \
-    accelerate>=0.26.0
-
-# Create a script to download models
-COPY <<EOF /app/download_models.py
-import os
-import gc
-import torch
-from sentence_transformers import SentenceTransformer
-from transformers import AutoTokenizer, AutoModelForCausalLM
-
-def download_and_save_model(model_name, save_path, is_sentence_transformer=False):
-    try:
-        print(f"Downloading {model_name}...")
-        if is_sentence_transformer:
-            model = SentenceTransformer(model_name)
-            model.save(save_path)
-        else:
-            tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
-            tokenizer.save_pretrained(save_path)
-            
-            # Load model with memory optimizations
-            model = AutoModelForCausalLM.from_pretrained(
-                model_name,
-                trust_remote_code=True,
-                torch_dtype=torch.float16,
-                low_cpu_mem_usage=True,
-                device_map="auto"
-            )
-            model.save_pretrained(save_path)
-            
-        print(f"Successfully saved {model_name}")
-        
-        # Clean up memory
-        del model
-        if not is_sentence_transformer:
-            del tokenizer
-        torch.cuda.empty_cache()
-        gc.collect()
-        
-    except Exception as e:
-        print(f"Error downloading {model_name}: {str(e)}")
-        raise
-
-# Download sentence transformer model
-download_and_save_model(
-    "sentence-transformers/all-mpnet-base-v2",
-    "/app/models/sentence_transformer",
-    is_sentence_transformer=True
-)
-
-# Download LLM (using a smaller model)
-download_and_save_model(
-    "microsoft/phi-2",  # Using Phi-2 instead of Qwen as it's smaller
-    "/app/models/llm"
-)
-EOF
-
-# Download models during build
-RUN python /app/download_models.py
+    huggingface_hub
 
-# Only set TRANSFORMERS_OFFLINE after downloading models
-ENV TRANSFORMERS_OFFLINE=1
-
-# Copy the main.py and modify it to use local paths
-COPY main.py /app/main.py
-RUN sed -i 's|"sentence-transformers/all-mpnet-base-v2"|"/app/models/sentence_transformer"|g' main.py && \
-    sed -i 's|"Qwen/Qwen2.5-1.5B-Instruct"|"/app/models/llm"|g' main.py
-
-# Copy the data file
+# Copy the necessary files into the container
 COPY plant_data_chunks_and_embeddings.csv /app/plant_data_chunks_and_embeddings.csv
-
-# Set proper permissions
-RUN chown -R appuser:appuser /app && \
-    chown -R appuser:appuser /.cache && \
-    chmod -R 755 /app/models
-
-# Switch to non-root user
-USER appuser
+COPY main.py /app/main.py
 
 # Set the environment variable to point to the app directory
 ENV PYTHONPATH=/app
 
-# Expose the port
-EXPOSE 8000
-
-# Optimize memory usage
-ENV OMP_NUM_THREADS=1
-ENV MKL_NUM_THREADS=1
-ENV TORCH_NUM_THREADS=1
-ENV MALLOC_TRIM_THRESHOLD_=0
+# Expose the port the app will run on
+EXPOSE 5000
 
-# Command to run the app
-CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "5000", "--timeout-keep-alive", "300"]
+# Command to run the app using Uvicorn (ASGI server for FastAPI)
+CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "5000"]