Trying ONNX models on CPU.

app.py

@@ -31,10 +31,10 @@ file_name = "onnx/model.onnx"
 
 model_name_cpu = "MoritzLaurer/ModernBERT-large-zeroshot-v2.0"
 
-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, file_name=file_name)
+# tokenizer_cpu = AutoTokenizer.from_pretrained(model_name_cpu)
 
-classifier_cpu = pipeline(task="zero-shot-classification", model=model_cpu, tokenizer=tokenizer_cpu)
+classifier_cpu = pipeline(task="zero-shot-classification", model=model_name_cpu, tokenizer=model_name_cpu)
 classifier_gpu = pipeline(task="zero-shot-classification", model=model_name, tokenizer=tokenizer_name, device="cuda:0")
 
 def classify(data_string, request: gradio.Request):

requirements.txt

@@ -4,5 +4,5 @@ huggingface_hub==0.32.4
 json5==0.9.25
 numpy
 uvicorn
-optimum[onnxruntime]==1.25.3
+optimum[onnxruntime-gpu]==1.25.3
 transformers==4.51.3