File size: 2,124 Bytes
3e16642 945b2ae 3e16642 63a7f6a 98402a5 945b2ae 3e16642 98402a5 3e16642 98402a5 3e16642 98402a5 3e16642 98402a5 3e16642 98402a5 3e16642 98402a5 3e16642 |
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 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 |
import gradio as gr
from transformers import AutoConfig, AutoModelForCausalLM
from modeling_snowflake import Snowflake4CausalLM
import torch
# --- Load Model and Tokenizer from Hugging Face Hub ---
# Register the custom model with the transformers library
AutoConfig.register("SnowflakeCore", PretrainedConfig)
AutoModelForCausalLM.register(PretrainedConfig, Snowflake4CausalLM)
# Load tokenizer
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
# Load model
model = AutoModelForCausalLM.from_pretrained(
MODEL_NAME,
torch_dtype=torch.float16 # Use half precision for memory efficiency
)
model.eval()
model.to("cuda" if torch.cuda.is_available() else "cpu")
# --- Inference Function ---
def generate_text(prompt, max_length=50):
"""
Generate text based on the input prompt using the trained model.
"""
# Tokenize the input prompt
inputs = tokenizer(prompt, return_tensors="pt", truncation=True, padding=True, max_length=384)
input_ids = inputs["input_ids"].to(model.device)
attention_mask = inputs["attention_mask"].to(model.device)
# Generate output tokens
with torch.no_grad():
outputs = model.generate(
input_ids=input_ids,
attention_mask=attention_mask,
max_length=max_length,
pad_token_id=tokenizer.eos_token_id # Use EOS token for padding
)
# Decode the generated tokens
generated_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
return generated_text
# --- Gradio Interface ---
with gr.Blocks() as demo:
gr.Markdown("# Snowflake-G0-stable Language Model")
gr.Markdown("This is an enhanced transformer language model trained on the DialogMLM-50K dataset. Try it out below!")
with gr.Row():
input_prompt = gr.Textbox(label="Input Prompt", placeholder="Enter your text here...")
output_text = gr.Textbox(label="Generated Text")
submit_button = gr.Button("Generate")
def on_submit(prompt):
return generate_text(prompt)
submit_button.click(on_submit, inputs=input_prompt, outputs=output_text)
# Launch the app
demo.launch() |