Self attention (SDPA) in unity.cpp (#34)

* Update README.md

* self attn (SDPA)

* Update README.md

reset

* op fix; attn matched

README.md

@@ -67,6 +67,16 @@ pip install -r requirements.txt
 python app.py
 ```
 
+## Running [Gradio](https://github.com/gradio-app/gradio) demo locally
+
+A demo is hosted [here](https://huggingface.co/spaces/facebook/seamless_m4t) on Hugging Face Spaces, but you can also try it locally.
+
+```bash
+cd demo
+pip install -r requirements.txt
+python app.py
+```
+
 # Libraries
 
 Seamless Communication depends on 3 libraries developed by Meta.

ggml/examples/unity/unity.cpp

@@ -556,7 +556,7 @@ struct ggml_cgraph * unity_graph(
         int32_t S = cur->ne[1];
         int32_t H = n_audio_enc_head;
         int32_t K_h = n_audio_enc_dim / H;
-
+        
         int32_t start_index = n_ctx - S;
         int32_t end_index = n_ctx + S - 1;