Update app.py

app.py

@@ -35,6 +35,14 @@ models_supported = {
             attn_implementation={"decoder": "flash_attention_2", "encoder": "eager"},
         ),
     ],
+    "mobser-small-v0.1": [
+        NougatProcessor.from_pretrained("TawasulAI/Mobser-Small-V0.1"),
+        VisionEncoderDecoderModel.from_pretrained(
+            "TawasulAI/Mobser-Small-V0.1",
+            torch_dtype=torch.bfloat16,
+            attn_implementation={"decoder": "flash_attention_2", "encoder": "eager"},
+        ),
+    ],
 }
 
 
@@ -121,8 +129,9 @@ There are three models available:
 - [arabic-small-nougat](https://huggingface.co/MohamedRashad/arabic-small-nougat): A small model that is faster but less accurate (a finetune from [facebook/nougat-small](https://huggingface.co/facebook/nougat-small)).
 - [arabic-base-nougat](https://huggingface.co/MohamedRashad/arabic-base-nougat): A base model that is more accurate but slower (a finetune from [facebook/nougat-base](https://huggingface.co/facebook/nougat-base)).
 - [arabic-large-nougat](https://huggingface.co/MohamedRashad/arabic-large-nougat): The largest of the three (Made from scratch using [riotu-lab/Aranizer-PBE-86k](https://huggingface.co/riotu-lab/Aranizer-PBE-86k) tokenizer and a larger transformer decoder model).
+- [mobser-small-v0.1](https://huggingface.co/TawasulAI/Mobser-Small-V0.1): A finetune built by [TawasulAI](https://huggingface.co/TawasulAI) team to push the boundry of what's possible further.
 
-**Disclaimer**: These models hallucinate text and are not perfect. They are trained on a mix of synthetic and real data and may not work well on all types of images.
+**Disclaimer**: Models can hallucinate text and are not perfect. Please double check the output if you care about accuracy the most.
 """
 
 example_images = list(Path(__file__).parent.glob("*.jpeg"))