Create app.py

app.py

@@ -0,0 +1,29 @@
+import os
+import subprocess
+
+# Instala os pacotes necessários
+subprocess.run(["pip", "install", "--upgrade", "pip"])
+subprocess.run(["pip", "install", "--upgrade", "torch", "transformers", "accelerate"])
+subprocess.run(["pip", "install", "git+https://github.com/TimDettmers/bitsandbytes.git"])
+
+import accelerate
+import bitsandbytes
+import gradio as gr
+from transformers import LlamaForCausalLM, LlamaTokenizer
+
+# Define a variável de ambiente para desabilitar CUDA
+os.environ["TRANSFORMERS_NO_CUDA"] = "1"
+
+# Carrega o modelo e o tokenizador
+model = LlamaForCausalLM.from_pretrained("Ramikan-BR/tinyllama_PY-CODER-bnb-4bit-lora_4k-q4_k_m-v2")
+tokenizer = LlamaTokenizer.from_pretrained("Ramikan-BR/tinyllama_PY-CODER-bnb-4bit-lora_4k-q4_k_m-v2")
+
+def predict(input_text):
+    # Codifica o texto de entrada e gera a saída
+    input_ids = tokenizer.encode(input_text, return_tensors="pt")
+    output = model.generate(input_ids, max_length=4096, do_sample=True, top_k=50, top_p=0.50, num_return_sequences=1)
+    return tokenizer.decode(output[0], skip_special_tokens=True)
+
+# Cria a interface Gradio
+iface = gr.Interface(fn=predict, inputs="text", outputs="text")
+iface.launch()