Add Phi-3-mini-4k generator option

app.py

@@ -226,8 +226,22 @@ def load_tinyllama():
     )
 
 
+def load_phi3_mini():
+    use_cuda = torch.cuda.is_available()
+    device_map = "cuda" if use_cuda else "cpu"
+    dtype = torch.float16 if use_cuda else torch.float32
+    return pipeline(
+        "text-generation",
+        model="microsoft/Phi-3-mini-4k-instruct",
+        device_map=device_map,
+        torch_dtype=dtype,
+        trust_remote_code=True,
+    )
+
+
 GENERATORS = {
     "tinyllama-1.1b-chat": GeneratorWrapper("tinyllama-1.1b-chat", load_tinyllama),
+    "phi-3-mini-4k": GeneratorWrapper("phi-3-mini-4k", load_phi3_mini),
 }
 
 
@@ -403,7 +417,7 @@ with gr.Blocks(title="MedDiscover") as demo:
             model_dd = gr.Dropdown(
                 label="Generator Model",
                 choices=list(GENERATORS.keys()),
-                value="tinyllama-1.1b-chat",
+                value="phi-3-mini-4k",
                 interactive=True,
             )
             k_slider = gr.Slider(1, 10, value=3, step=1, label="Top-k chunks")