Create app.py

app.py

@@ -0,0 +1,40 @@
+import torch
+from peft import PeftModel, PeftConfig
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+peft_model_id = "JuliaUpton/Math_AI"
+config = PeftConfig.from_pretrained(peft_model_id)
+model = AutoModelForCausalLM.from_pretrained(config.base_model_name_or_path, return_dict=True, load_in_8bit=False)
+tokenizer = AutoTokenizer.from_pretrained(config.base_model_name_or_path)
+"
+# Load the Lora model
+merged_model = PeftModel.from_pretrained(model, peft_model_id)
+
+def input_from_text(instruction):
+  return f"<s>[INST]Below is a math inquiry, please answer it as a math expert showing your thought process.\n\n### Inquiry:\n{instruction}\n\n### Response:[/INST]"
+
+def make_inference(instruction):
+  inputs = mixtral_tokenizer(input_from_text(instruction), return_tensors="pt")
+
+  outputs = merged_model.generate(
+      **inputs,
+      max_new_tokens=150,
+      generation_kwargs={"repetition_penalty" : 1.7}
+      )
+  # print(mixtral_tokenizer.decode(outputs[0], skip_special_tokens=True))
+  result = mixtral_tokenizer.decode(outputs[0], skip_special_tokens=True).split("[/INST]")[1]
+  return result
+
+if __name__ == "__main__":
+    # make a gradio interface
+    import gradio as gr
+
+    gr.Interface(
+        make_inference,
+        [
+            gr.Textbox(lines=5, label="Instruction"),
+        ],
+        gr.Textbox(label="Answer"),
+        title="Math-AI",
+        description="Math-AI is a generative model that answers math questions",
+    ).launch()