Spaces:
Paused
Paused
| import gradio as gr | |
| import torch | |
| import spaces | |
| from transformers import AutoModelForCausalLM, AutoTokenizer | |
| from transformers import StoppingCriteria, StoppingCriteriaList, TextIteratorStreamer | |
| from threading import Thread | |
| # Loading the tokenizer and model from Hugging Face's model hub. | |
| if torch.cuda.is_available(): | |
| tokenizer = AutoTokenizer.from_pretrained("0x7194633/fialka-13B-v4") | |
| model = AutoModelForCausalLM.from_pretrained("0x7194633/fialka-13B-v4", load_in_8bit=True, device_map="auto") | |
| # Defining a custom stopping criteria class for the model's text generation. | |
| class StopOnTokens(StoppingCriteria): | |
| def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor, **kwargs) -> bool: | |
| stop_ids = [2] # IDs of tokens where the generation should stop. | |
| for stop_id in stop_ids: | |
| if input_ids[0][-1] == stop_id: # Checking if the last generated token is a stop token. | |
| return True | |
| return False | |
| # Function to generate model predictions. | |
| def predict(message, history): | |
| history_transformer_format = history + [[message, ""]] | |
| stop = StopOnTokens() | |
| # Formatting the input for the model. | |
| messages = "<|system|>\nТы Фиалка - самый умный нейронный помощник, созданный 0x7o.</s>\n" | |
| messages += "</s>".join(["</s>".join(["\n<|user|>" + item[0], "\n<|assistant|>" + item[1]]) | |
| for item in history_transformer_format]) | |
| model_inputs = tokenizer([messages], return_tensors="pt").to("cuda") | |
| streamer = TextIteratorStreamer(tokenizer, timeout=10., skip_prompt=True, skip_special_tokens=True) | |
| generate_kwargs = dict( | |
| model_inputs, | |
| streamer=streamer, | |
| max_new_tokens=1024, | |
| do_sample=True, | |
| top_p=0.95, | |
| top_k=50, | |
| temperature=0.7, | |
| repetition_penalty=1.0, | |
| num_beams=1, | |
| stopping_criteria=StoppingCriteriaList([stop]) | |
| ) | |
| t = Thread(target=model.generate, kwargs=generate_kwargs) | |
| t.start() # Starting the generation in a separate thread. | |
| partial_message = "" | |
| for new_token in streamer: | |
| partial_message += new_token | |
| if '</s>' in partial_message: # Breaking the loop if the stop token is generated. | |
| break | |
| yield partial_message | |
| # Setting up the Gradio chat interface. | |
| gr.ChatInterface(predict, | |
| title="Fialka 13B v4", | |
| description="Внимание! Все ответы сгенерированы и могут содержать неточную информацию.", | |
| examples=['Как приготовить рыбу?', 'Кто президент США?'] | |
| ).launch() # Launching the web interface. | |