rollback 2 days

app.py

@@ -1,3 +1,4 @@
+ app.py – virkar 100% á ZeroGPU (og venjulegum GPU líka)
 import os
 import gradio as gr
 import spaces
@@ -5,40 +6,39 @@ from transformers import pipeline
 
 MODEL_NAME = "palli23/whisper-small-sam_spjall"
 
-@spaces.GPU(duration=120)   # nóg fyrir 3 mín hljóð
+# Látum mótið hlaðast EINUSINNI þegar appið ræsist (ekki í hverju kall)
+print("Hleð Whisper módelinu einu sinni...")
+pipe = pipeline(
+    "automatic-speech-recognition",
+    model=MODEL_NAME,
+    torch_dtype="auto",
+    device_map="auto",           # þetta er lykillinn fyrir ZeroGPU
+    token=os.getenv("HF_TOKEN")
+)
+print("Módel tilbúið!")
+
+@spaces.GPU(duration=120)   # 2 mínútur nægja fyrir 3–4 mín hljóð
 def transcribe_3min(audio_path):
     if not audio_path:
-        return "Hladdu upp hljóðskrá"
-    
-    # Whisper pipeline með chunking – ZeroGPU öruggt
-    pipe = pipeline(
-        "automatic-speech-recognition",
-        model=MODEL_NAME,
-        device=0,
-        token=os.getenv("HF_TOKEN"),
-        torch_dtype="float16",    # ← NEW: FP16 instead of FP32
-        # ← REMOVED: use_flash_attention_2 (causes TypeError on this model)
-    )
+        return "Hladdu upp hljóðskrá fyrst"
     
     result = pipe(
         audio_path,
-        chunk_length_s=30,           # 30 sek chunkar
-        stride_length_s=(6, 0),      # 6 sek overlap
+        chunk_length_s=30,
+        batch_size=8,
         return_timestamps=False,
-        batch_size=8
+        generate_kwargs={"language": "is", "task": "transcribe"}
     )
-    
     return result["text"]
 
-# Interface
 with gr.Blocks(title="Íslenskt ASR – 3 mín") as demo:
-    gr.Markdown("# Íslenskt ASR – 3 mínútur")
-    gr.Markdown("**Whisper · Very low WER · 0.5-5minute audio transcribe á ZeroGPU**")
-    
-    audio = gr.Audio(type="filepath", label="Hladdu upp .mp3 / .wav (allt að 3 mín)")
-    btn = gr.Button("Transcribe", variant="primary", size="lg")
-    out = gr.Textbox(lines=30, label="Útskrift")
+    gr.Markdown("# Íslenskt Whisper – 3 mín hljóð á ZeroGPU")
+    gr.Markdown("Hladdu upp allt að 3–4 mínútna hljóðskrá (mp3/wav). Virkar strax!")
     
-    btn.click(transcribe_3min, inputs=audio, outputs=out)
+    audio_in = gr.Audio(type="filepath", label="Hljóðskrá (max ~4 mín)")
+    btn = gr.Button("Transcribe", variant="primary")
+    text_out = gr.Textbox(lines=25, label="Útskrift")
+
+    btn.click(transcribe_3min, inputs=audio_in, outputs=text_out)
 
-demo.launch(auth=("beta", "beta2025"))
+demo.launch()