Upload 2 files

app.py

@@ -0,0 +1,15 @@
+import gradio as gr
+from inference import generate
+examples = [
+    ["С помощью круглых тензоров"],
+    ["gpt40 это"],
+]
+demo = gr.Interface(
+    fn=generate,
+    title="Генератор ебаных идей для резерча.",
+    inputs=gr.inputs.Textbox(lines=5, label="Ввод"),
+    outputs=gr.outputs.Textbox(label="Генерация"),
+    examples=examples
+)    
+if __name__ == "__main__":
+    demo.launch()   

inference.py

@@ -0,0 +1,20 @@
+import torch
+from transformers import AutoTokenizer, AutoModelForCausalLM
+
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+model_dir = "Den4ikAI/ebany_researcher"
+
+tokenizer = AutoTokenizer.from_pretrained(model_dir)
+tokenizer.add_special_tokens({'bos_token': '<s>', 'eos_token': '</s>', 'pad_token': '<pad>'})
+
+model = AutoModelForCausalLM.from_pretrained(model_dir)
+model.to(device)
+model.eval()
+def generate(prompt):
+    encoded_prompt = tokenizer.encode(prompt, return_tensors="pt").to(device)
+    out = model.generate(encoded_prompt, max_length=50, do_sample=True, top_k=50, top_p=0.95, temperature=1.0,
+                        num_return_sequences=1)
+    for i, tokens in enumerate(out.cpu().tolist(), start=1):
+        tokens = tokens[encoded_prompt.shape[1]:]
+        text = tokenizer.decode(tokens)
+    return text