loader

Dockerfile

@@ -1,11 +1,20 @@
-
-FROM python:3.10
+FROM python:3.9
 
 WORKDIR /code
 
-COPY requirements.txt .
-RUN pip install --no-cache-dir -r requirements.txt
+COPY ./requirements.txt /code/requirements.txt
+
+RUN pip install --no-cache-dir --upgrade -r /code/requirements.txt
+
+RUN useradd -m -u 1000 user
+
+USER user
+
+ENV HOME=/home/user \
+	PATH=/home/user/.local/bin:$PATH
+
+WORKDIR $HOME/app
 
-COPY . .
+COPY --chown=user . $HOME/app
 
-CMD ["uvicorn", "app.main:app", "--host", "0.0.0.0", "--port", "7860"]
+CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860"]

app/model_loader.py

@@ -34,27 +34,27 @@
 #     )
 #     return model, tokenizer
 
-from transformers import AutoModelForCausalLM, AutoTokenizer
 import os
+from transformers import AutoModelForCausalLM, AutoTokenizer
 
 def load_model():
-    model_name = "gpt2"
-    cache_dir = "/tmp/hf_cache"
+    # Use a user-writable cache directory (important for Docker non-root)
+    HF_CACHE = os.path.expanduser("~/.cache/huggingface")
+    os.makedirs(HF_CACHE, exist_ok=True)
 
-    # Ensure cache dir exists and is writable
-    os.makedirs(cache_dir, exist_ok=True)
-    os.environ["TRANSFORMERS_CACHE"] = cache_dir
-    os.environ["HF_HOME"] = cache_dir
+    os.environ["TRANSFORMERS_CACHE"] = HF_CACHE
+    os.environ["HF_HOME"] = HF_CACHE
+
+    model_name = "gpt2"
 
     tokenizer = AutoTokenizer.from_pretrained(
         model_name,
-        cache_dir=cache_dir,
-        use_fast=True
+        cache_dir=HF_CACHE
     )
 
     model = AutoModelForCausalLM.from_pretrained(
         model_name,
-        cache_dir=cache_dir
+        cache_dir=HF_CACHE
     )
 
     return model, tokenizer