get model from huggingface-hub

app.py

@@ -27,22 +27,14 @@ def process_cat_embs(cat_embs):
 
 
 def download_rev_models():
-    # from huggingface_hub import hf_hub_download
-    # import joblib
-
-    # REPO_ID = "Revai/reginald"
-    # FILENAME = "sklearn_model.joblib"
-
-    # model = joblib.load(
-    #     hf_hub_download(repo_id=REPO_ID, filename=FILENAME)
-    # )
-    model_path = "/Users/natalie/NERD-2941/reginald/10.jit.zip"
-    units_path = "/Users/natalie/NERD-2941/reginald/tk.units.txt"
-    audio_path = "/Users/natalie/NERD-2941/rev-wenet/runtime/web/fdhc0_si1559.wav"
-    cat_embs = "1,0"
-    device = "cpu"
-    cat_embs = process_cat_embs
-    model = load_model(model_path, units_path)
+    from huggingface_hub import hf_hub_download
+    import joblib
+
+    REPO_ID = "Revai/reginald"
+
+    files = ['10.jit.zip', 'tk.units.txt']
+    downloaded_files = [hf_hub_download(repo_id=REPO_ID, filename=f) for f in files]
+    model = load_model(downloaded_files[0], downloaded_files[1])
     return model
 
 model = download_rev_models()
@@ -66,7 +58,7 @@ def recognition(audio, style=0):
 # input
 inputs = [
     gr.inputs.Audio(source="microphone", type="filepath", label='Input audio'),
-    gr.Slider(0, 1, value=0, label="Style", info="Choose between verbatim and NV"),
+    gr.Slider(0, 1, value=0, label="Style - from verbatim (0) to non-verbatim (1)", info="Choose between verbatim and NV"),
 ]
 
 output = gr.outputs.Textbox(label="Output Text")

requirements.txt

@@ -1,2 +1,3 @@
 wenet @ git+https://github.com/revdotcom/rev-wenet@feature/NERD-3007
 gradio==3.14.0
+huggingface-hub