Spaces:
Sleeping
Sleeping
File size: 1,213 Bytes
eaa8416 d8ff530 eaa8416 d0e6c3b eaa8416 b9d49cf 2ac2871 eaa8416 5e2e076 d8ff530 5e2e076 d497b84 5e2e076 d497b84 d8ff530 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 |
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline, logging
import gradio as gr
model_name = "microsoft/phi-2"
model = AutoModelForCausalLM.from_pretrained(
model_name,
trust_remote_code=True
)
model.config.use_cache = False
tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
tokenizer.pad_token = tokenizer.eos_token
# Loading adapter (trained LORA weights)
# ckpt = '/content/drive/MyDrive/S27/results/checkpoint-500'
# model.load_adapter(ckpt)
adapter_path = 'checkpoint-500'
model.load_adapter(adapter_path)
def inference(prompt):
pipe = pipeline(task="text-generation",model=model,tokenizer=tokenizer,max_length = 100)
result = pipe(f"<s>[INST] {prompt} [/INST]")
return result[0]['generated_text']
INTERFACE = gr.Interface(fn=inference, inputs=[gr.Textbox(label= "Prompt", value= 'what should we do to save time')],
outputs=gr.Text(label= "Generated Text"), title="Language Model Phi-2 fine-tuned with OpenAssistant/oasst-1 dataset using QLoRA strategy",
examples = [['explain transpiration in plants'],]
).launch(debug=True) |