init

app.py

@@ -3,18 +3,17 @@ import gradio as gr
 
 import soundfile as sf
 import torch
-from transformers import AutoModelForCausalLM, AutoProcessor, infer_device
+from transformers import AutoModelForCausalLM, AutoProcessor
 
 
 
 
 model_path = "microsoft/Phi-4-multimodal-instruct"
-device = f"{infer_device()}:0"
 
 processor = AutoProcessor.from_pretrained(model_path, trust_remote_code=True)
 model = AutoModelForCausalLM.from_pretrained(
     model_path,
-    device_map=device,
+    device_map="auto",
     dtype=torch.float16,
     trust_remote_code=True,
     _attn_implementation="flash_attention_2",
@@ -23,7 +22,7 @@ model = AutoModelForCausalLM.from_pretrained(
 model.load_adapter(
     model_path,
     adapter_name="speech",
-    device_map=device,
+    device_map="auto",
     adapter_kwargs={"subfolder": 'speech-lora'}
 )
 model.set_adapter("speech")