push

.dockerignore

@@ -0,0 +1 @@
+token.txt

Dockerfile

@@ -1,9 +1,13 @@
 # Base Image
 FROM python:3.10-slim
 
+# Build argument for Hugging Face token
+ARG HF_TOKEN
+
 ENV DEBIAN_FRONTEND=noninteractive \
     PYTHONUNBUFFERED=1 \
-    PYTHONDONTWRITEBYTECODE=1
+    PYTHONDONTWRITEBYTECODE=1 \
+    HF_TOKEN=${HF_TOKEN}
 
 WORKDIR /code
 
@@ -43,8 +47,15 @@ RUN python -c "from transformers import pipeline; pipeline('text-to-speech', mod
  && python -c "from transformers import pipeline; pipeline('text-to-speech', model='facebook/mms-tts-yor')"
 
 # Pre-load N-ATLaS model during build
-RUN python -c "from huggingface_hub import snapshot_download; snapshot_download(repo_id='NCAIR1/N-ATLaS')" \
- && python -c "from transformers import AutoTokenizer, AutoModelForCausalLM; import torch; tokenizer = AutoTokenizer.from_pretrained('NCAIR1/N-ATLaS'); model = AutoModelForCausalLM.from_pretrained('NCAIR1/N-ATLaS', torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32, device_map='auto' if torch.cuda.is_available() else None, trust_remote_code=True, low_cpu_mem_usage=True, use_cache=True); print('N-ATLaS model loaded successfully')"
+RUN python -c "from huggingface_hub import snapshot_download; snapshot_download(repo_id='NCAIR1/N-ATLaS', token='$HF_TOKEN')" \
+ && python -c "from transformers import AutoTokenizer, AutoModelForCausalLM; import torch; tokenizer = AutoTokenizer.from_pretrained('NCAIR1/N-ATLaS', token='$HF_TOKEN'); model = AutoModelForCausalLM.from_pretrained('NCAIR1/N-ATLaS', torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32, device_map='auto' if torch.cuda.is_available() else None, trust_remote_code=True, low_cpu_mem_usage=True, use_cache=True, token='$HF_TOKEN'); print('N-ATLaS model loaded successfully')"
+
+# Pre-download ASR models (will be lazy-loaded at runtime)
+RUN python -c "from huggingface_hub import snapshot_download; snapshot_download(repo_id='NCAIR1/Hausa-ASR', token='$HF_TOKEN')" \
+ && python -c "from huggingface_hub import snapshot_download; snapshot_download(repo_id='NCAIR1/Yoruba-ASR', token='$HF_TOKEN')" \
+ && python -c "from huggingface_hub import snapshot_download; snapshot_download(repo_id='NCAIR1/Igbo-ASR', token='$HF_TOKEN')" \
+ && python -c "from huggingface_hub import snapshot_download; snapshot_download(repo_id='NCAIR1/NigerianAccentedEnglish', token='$HF_TOKEN')" \
+ && find /models/huggingface -name '*.lock' -delete
 
 # Copy project files
 COPY . .