Upload Dockerfile with huggingface_hub

Dockerfile

@@ -19,31 +19,31 @@ USER user
 ENV PATH="/home/user/.local/bin:${PATH}"
 WORKDIR /home/user/app
 
-# PIP Installation Sequence (Crucial)
-# a. pip install --upgrade pip
+# ============================================================
+# FIXED Installation Order - Critical for unsloth compatibility
+# ============================================================
+
+# a. Upgrade pip first
 RUN python3 -m pip install --upgrade pip
 
-# b. pip install torch==2.4.0 torchvision==0.19.0 torchaudio==2.4.0 --index-url https://download.pytorch.org/whl/cu121
-RUN python3 -m pip install torch==2.4.0 torchvision==0.19.0 torchaudio==2.4.0 --index-url https://download.pytorch.org/whl/cu121
+# b. Install torch with CUDA 12.1 (MUST come first)
+RUN python3 -m pip install torch==2.3.1+cu121 torchvision==0.18.1+cu121 torchaudio==2.3.1+cu121 --index-url https://download.pytorch.org/whl/cu121
 
-# c. pip uninstall -y torchao (ensure it is gone)
-RUN python3 -m pip uninstall -y torchao
+# c. Install transformers (MUST be 4.44.x - 4.46.x for unsloth_zoo)
+RUN python3 -m pip install transformers==4.44.2
 
-# d. pip install --no-deps "unsloth[colab-new] @ git+https://github.com/unslothai/unsloth.git"
-RUN python3 -m pip install --no-deps "unsloth[colab-new] @ git+https://github.com/unslothai/unsloth.git"
+# d. Install unsloth (core only, not the git version)
+RUN python3 -m pip install unsloth
 
-# d.1 Install unsloth_zoo (required dependency for unsloth to work)
+# e. Install unsloth_zoo (required dependency)
 RUN python3 -m pip install unsloth_zoo
 
-# e. pip install --no-deps xformers==0.0.27 trl==0.11.0 peft accelerate bitsandbytes
-RUN python3 -m pip install --no-deps xformers==0.0.27 trl==0.11.0 peft accelerate bitsandbytes
-
-# f. Fix transformers version (unsloth_zoo needs specific version - 4.46.x works)
-RUN python3 -m pip install transformers==4.46.3
+# f. Install other required packages
+RUN python3 -m pip install xformers==0.0.27 peft accelerate bitsandbytes datasets huggingface_hub wandb trl
 
-# Install other requirements
+# g. Install any remaining requirements
 COPY --chown=user:user requirements.txt .
-RUN python3 -m pip install -r requirements.txt
+RUN python3 -m pip install -r requirements.txt || true
 
 # Copy application code
 COPY --chown=user:user train.py .
@@ -51,4 +51,4 @@ COPY --chown=user:user aegis_training_data_500.json .
 
 EXPOSE 7860
 
-CMD ["python3", "-u", "train.py"]
+CMD ["python3", "-u", "train.py"]