app.py

import gradio as gr
import spaces
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained("Upstage/SOLAR-10.7B-Instruct-v1.0")
model = AutoModelForCausalLM.from_pretrained(
    "rishiraj/meow",
    device_map="auto",
    torch_dtype=torch.float16,
)

#zero = torch.Tensor([0]).cuda()
#print(zero.device) # <-- 'cpu' 🤔

@spaces.GPU
def chat(prompt):
    conversation = [ {'role': 'user', 'content': 'Hello?'} ] 

    prompt = tokenizer.apply_chat_template(conversation, tokenize=False, add_generation_prompt=True)

    inputs = tokenizer(prompt, return_tensors="pt").to(zero.device) 
    outputs = model.generate(**inputs, use_cache=True, max_length=4096)
    output_text = tokenizer.decode(outputs[0]) 
    print(output_text)
    return output_text

    #print() # <-- 'cuda:0' 🤗
    return f"Hello {zero + n} Tensor"



gr.Interface(
    fn=chat, 
    inputs=gr.Text(), 
    outputs=gr.Text()
).launch()