fix: replace anchor tags with image tags in the "About" section for better visual representation of comparison images

app.py

@@ -65,14 +65,14 @@ with gr.Blocks("ParityError/Interstellar") as demo:
                 Due to their size, state-of-the-art models such as FLUX take more than 6 seconds to generate a single image on a high-end H100 GPU.
                 While compression techniques can reduce inference time, their impact on quality often remains unclear.
 
-                <a href="https://huggingface.co/datasets/PrunaAI/documentation-images/resolve/main/inferbench/spider_comparison.png" target="_blank">Spider Comparison</a>
+                <img src="https://huggingface.co/datasets/PrunaAI/documentation-images/resolve/main/inferbench/spider_comparison.png" alt="Spider Comparison">
 
                 To bring more transparency around the quality of compressed models:
 
                 - We release “juiced” endpoints for popular image generation models on Replicate, making it easy to play around with our compressed models.
                 - We assess the quality of compressed FLUX-APIs from Replicate, fal, Fireworks AI and Together AI according to different benchmarks.
 
-                <a href="https://huggingface.co/datasets/PrunaAI/documentation-images/resolve/main/inferbench/speed_comparison.png" target="_blank">Speed Comparison</a>
+                <img src="https://huggingface.co/datasets/PrunaAI/documentation-images/resolve/main/inferbench/speed_comparison.png" alt="Speed Comparison">
 
                 FLUX-juiced was obtained using a combination of compilation and caching algorithms and we are proud to say that it consistently outperforms alternatives, while delivering performance on par with the original model.
                 This combination is available in our Pruna Pro package and can be applied to almost every image generation model.