Trying ONNX models on CPU.

app.py

@@ -27,14 +27,13 @@ print(f"CUDA device: {torch.cuda.get_device_name(torch.cuda.current_device())}")
 
 model_name = "MoritzLaurer/deberta-v3-base-zeroshot-v2.0"
 tokenizer_name = "MoritzLaurer/deberta-v3-base-zeroshot-v2.0"
-file_name = "onnx/model.onnx"
 
-model_name_cpu = "MoritzLaurer/ModernBERT-large-zeroshot-v2.0"
+model_name_cpu = "MoritzLaurer/roberta-large-zeroshot-v2.0-c"
 
-# model_cpu = ORTModelForSequenceClassification.from_pretrained(model_id=model_name_cpu, file_name=file_name)
-# tokenizer_cpu = AutoTokenizer.from_pretrained(model_name_cpu)
+model_cpu = ORTModelForSequenceClassification.from_pretrained(model_id=model_name_cpu, subfolder="onnx", file_name="model.onnx")
+tokenizer_cpu = AutoTokenizer.from_pretrained(model_name_cpu)
 
-classifier_cpu = pipeline(task="zero-shot-classification", model=model_name_cpu, tokenizer=model_name_cpu)
+classifier_cpu = pipeline(task="zero-shot-classification", model=model_name, tokenizer=tokenizer_cpu)
 classifier_gpu = pipeline(task="zero-shot-classification", model=model_name, tokenizer=tokenizer_name, device="cuda:0")
 
 def classify(data_string, request: gradio.Request):