add warning

README.md

@@ -58,6 +58,14 @@ This model is `cardiffnlp/twitter-roberta-base` further fine-tuned on emoji clas
 
 Also tried `google/vit-base-patch32-384`, `google/vit-base-patch16-384` for the vision models, but results were inconclusive.
 
+## Result Interpretation (Warning)
+
+It would be wrong to claim that this model learned 'semantic reasoning' between sentence and image features. 
+
+It more likely learned a mapping between sentence sentiment and image occurrence statistics. Because the set of gif images repeat across the dataset, although paired with different sentences.
+
+That is not to say that learning such semantic relations isn't feasible with this model. And it is well worth working on in the future, with a larger and better constructed dataset.
+
 ### 📈 Training Logs
 
 Training logs can be found [here](https://wandb.ai/cceyda/flax-clip?workspace=user-cceyda)
@@ -83,6 +91,7 @@ I will definitely be trying out training a similar model for emoji & meme data.
 Training CLIP is just the first step, if we have a well trained CLIP generation is within reach 🚀   
 
 # How to use
+The final model available [here](https://huggingface.co/ceyda/clip-reply)
 
 ```py
 from model import FlaxHybridCLIP # see demo 
@@ -119,3 +128,5 @@ Example:
 # Demo
 
 https://huggingface.co/spaces/flax-community/clip-reply-demo
+
+