app.py

app.py

@@ -0,0 +1,39 @@
+import os
+
+import gradio as gr
+import torch
+from transformers import RobertaForSequenceClassification, RobertaTokenizer, RobertaConfig
+
+os.system("wget https://openaipublic.azureedge.net/gpt-2/detector-models/v1/detector-base.pt")
+
+config = RobertaConfig.from_pretrained("roberta-base")
+model = RobertaForSequenceClassification(config)
+model.load_state_dict(torch.load("detector-base.pt")["model_state_dict"])
+
+tokenizer = RobertaTokenizer.from_pretrained("roberta-base")
+
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
+model.to(device)
+
+
+def detect(query):
+    tokens = tokenizer.encode(query)
+    all_tokens = len(tokens)
+    tokens = tokens[:tokenizer.max_len - 2]
+    used_tokens = len(tokens)
+    tokens = torch.tensor([tokenizer.bos_token_id] + tokens + [tokenizer.eos_token_id]).unsqueeze(0)
+    mask = torch.ones_like(tokens)
+
+    with torch.no_grad():
+        logits = model(tokens.to(device), attention_mask=mask.to(device))[0]
+        probs = logits.softmax(dim=-1)
+
+    fake, real = probs.detach().cpu().flatten().numpy().tolist()
+
+    return f"Fake: {fake:.2%} | Real: {real:.2%} | Used tokens: {used_tokens} | All tokens: {all_tokens}"
+
+
+demo = gr.Interface(fn=detect, inputs="text", outputs="text")
+
+demo.launch()