Imsachin010 commited on
Commit
1c1d77b
·
1 Parent(s): 9e54e20

Fix HF Space crashing: Copy missing training directory and fix user 1000 permissions

Browse files
Files changed (1) hide show
  1. Dockerfile +5 -1
Dockerfile CHANGED
@@ -16,13 +16,17 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
16
  COPY requirements.txt .
17
  RUN pip install --no-cache-dir -r requirements.txt
18
 
19
- # Copy the salespath_env package
20
  COPY salespath_env/ ./salespath_env/
 
21
 
22
  # Copy and set permissions for the training script
23
  COPY run_hf_training.sh ./run_hf_training.sh
24
  RUN sed -i 's/\r$//' ./run_hf_training.sh && chmod +x ./run_hf_training.sh
25
 
 
 
 
26
  # Health check
27
  HEALTHCHECK --interval=30s --timeout=5s --start-period=15s --retries=3 \
28
  CMD curl -f http://localhost:${PORT}/health || exit 1
 
16
  COPY requirements.txt .
17
  RUN pip install --no-cache-dir -r requirements.txt
18
 
19
+ # Copy the salespath_env package and training scripts
20
  COPY salespath_env/ ./salespath_env/
21
+ COPY training/ ./training/
22
 
23
  # Copy and set permissions for the training script
24
  COPY run_hf_training.sh ./run_hf_training.sh
25
  RUN sed -i 's/\r$//' ./run_hf_training.sh && chmod +x ./run_hf_training.sh
26
 
27
+ # Grant full read/write access so Hugging Face user 1000 can write model checkpoints
28
+ RUN chmod -R 777 /app
29
+
30
  # Health check
31
  HEALTHCHECK --interval=30s --timeout=5s --start-period=15s --retries=3 \
32
  CMD curl -f http://localhost:${PORT}/health || exit 1