!pip install unsloth

from transformers import TextStreamer

gemma_prompt = """
### Input:
{}

### Response:
{}"""


from unsloth import FastLanguageModel
model, tokenizer = FastLanguageModel.from_pretrained(
    model_name = "akshitha-k/oneliner-to-stories", 
    max_seq_length = 1024,
    dtype = None,
    load_in_4bit = True,
)
FastLanguageModel.for_inference(model) # Enable native 2x faster inference


inputs = tokenizer(
[
    gemma_prompt.format(
        "Ash and Roh went to the forest..", # input
        "", # output - leave this blank for generation!
    )
], return_tensors = "pt").to("cuda")

text_streamer = TextStreamer(tokenizer)
_ = model.generate(**inputs, streamer = text_streamer, max_new_tokens = 512)