init space

Dockerfile

@@ -1,18 +1,21 @@
 FROM python:3.10
 
+RUN apt-get update && \
+    apt-get install -y git-lfs && \
+    rm -rf /var/lib/apt/lists/* && \
+    git lfs install
+
 RUN useradd -m -u 1000 user
 USER user
 ENV PATH="/home/user/.local/bin:$PATH"
-
 WORKDIR /app
 
-RUN git clone https://huggingface.co/zcamz/tiny-llava-OpenELM-270M-Instruct-aimv2-large-patch14-224-distilled-elm_lora-finetune
+RUN git lfs install && \
+    git clone https://huggingface.co/zcamz/tiny-llava-OpenELM-270M-Instruct-aimv2-large-patch14-224-distilled-elm_lora-finetune
 
 COPY --chown=user TinyLLaVA_Factory/pyproject.toml /app
-
 RUN pip install --upgrade pip  # enable PEP 660 support
 RUN pip install -e .
-
 COPY --chown=user TinyLLaVA_Factory/ /app
 
-CMD ["python", "tinyllava/serve/app.py", "--model-path", "./tiny-llava-OpenELM-270M-Instruct-aimv2-large-patch14-224-distilled-elm_lora-finetune", "--port", "7860"]
+CMD ["python", "-m", "tinyllava.serve.app", "--model-path", "./tiny-llava-OpenELM-270M-Instruct-aimv2-large-patch14-224-distilled-elm_lora-finetune", "--port", "7860"]

TinyLLaVA_Factory/tinyllava/model/configuration_tinyllava.py

@@ -121,7 +121,7 @@ class TinyLlavaConfig(PretrainedConfig):
             )
             
         else:
-            self.vision_config = AutoConfig.from_pretrained(self.vision_model_name_or_path.split(':')[-1])
+            self.vision_config = AutoConfig.from_pretrained(self.vision_model_name_or_path.split(':')[-1],  trust_remote_code=True)
             self.vision_config = getattr(self.vision_config, 'vision_config', self.vision_config)
             if vision_config is not None:
                 self.vision_config = self.vision_config.from_dict(vision_config)

TinyLLaVA_Factory/tinyllava/model/vision_tower/aimv2.py

@@ -0,0 +1,14 @@
+from transformers import AutoModel, AutoImageProcessor 
+
+from . import register_vision_tower
+from .base import VisionTower
+
+
+@register_vision_tower('aimv2')      
+class AIMv2Tower(VisionTower):
+    def __init__(self, cfg):
+        super().__init__(cfg)
+        self._vision_tower = AutoModel.from_pretrained(cfg.model_name_or_path, config = cfg, trust_remote_code=True)
+        self._image_processor = AutoImageProcessor.from_pretrained(cfg.model_name_or_path)
+  
+