[New Sample] Add "uer_roberta-base-finetuned-cluener2020-chinese" Model Computational Graph

graph_net/test/uer_roberta_cluener_test.py

@@ -0,0 +1,46 @@
+import torch
+from transformers import AutoModelForTokenClassification, AutoTokenizer
+import graph_net.torch
+
+
+def get_model_name():
+    return "uer/roberta-base-finetuned-cluener2020-chinese"
+
+
+def get_graph_name():
+    return "uer_roberta-base-finetuned-cluener2020-chinese"
+
+
+def create_model(device: torch.device):
+    model = AutoModelForTokenClassification.from_pretrained(get_model_name())
+    model.eval()
+    return model.to(device)
+
+
+if __name__ == "__main__":
+    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
+    tokenizer = AutoTokenizer.from_pretrained(get_model_name())
+    text = "你好"
+    inputs = tokenizer(
+        text,
+        return_tensors="pt",
+        padding="max_length",
+        truncation=True,
+        max_length=4,
+    )
+    inputs = {k: v.to(device) for k, v in inputs.items()}
+
+    model = create_model(device)
+    model = graph_net.torch.extract(
+        name=get_graph_name(),
+        dynamic=True,
+    )(model)
+
+    print("Running inference...")
+    with torch.no_grad():
+        outputs = model(**inputs)
+    if hasattr(outputs, "logits"):
+        print("Inference finished. Logits shape:", outputs.logits.shape)
+    else:
+        print("Inference finished.")

samples/transformers-auto-model/uer_roberta-base-finetuned-cluener2020-chinese/graph_hash.txt

@@ -0,0 +1 @@
+bf7bc11f2ad7d7a69ea04ce765d892e38d23ba20be9b716bf68064ec75f11924

samples/transformers-auto-model/uer_roberta-base-finetuned-cluener2020-chinese/graph_net.json

@@ -0,0 +1,7 @@
+{
+    "framework": "torch",
+    "num_devices_required": 1,
+    "num_nodes_required": 1,
+    "dynamic": true,
+    "model_name": "uer_roberta-base-finetuned-cluener2020-chinese"
+}