Spaces:
Runtime error
Runtime error
| import spaces | |
| import torch | |
| import torch.nn.functional as F | |
| from torch import Tensor | |
| from transformers import AutoTokenizer, AutoModel | |
| import gradio as gr | |
| import os | |
| title = """ | |
| # 👋🏻Welcome to 🙋🏻♂️Tonic's 🐣e5-mistral🛌🏻Embeddings """ | |
| description = """ | |
| You can use this ZeroGPU Space to test out the current model [intfloat/e5-mistral-7b-instruct](https://huggingface.co/intfloat/e5-mistral-7b-instruct). 🐣e5-mistral🛌🏻 has a larger context🪟window, a different prompting/return🛠️mechanism and generally better results than other embedding models. use it via API to create embeddings or try out the sentence similarity to see how various optimization parameters affect performance. | |
| You can also use 🐣e5-mistral🛌🏻 by cloning this space. 🧬🔬🔍 Simply click here: <a style="display:inline-block" href="https://huggingface.co/spaces/Tonic/e5?duplicate=true"><img src="https://img.shields.io/badge/-Duplicate%20Space-blue?labelColor=white&style=flat&logo=data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAABAAAAAQCAYAAAAf8/9hAAAAAXNSR0IArs4c6QAAAP5JREFUOE+lk7FqAkEURY+ltunEgFXS2sZGIbXfEPdLlnxJyDdYB62sbbUKpLbVNhyYFzbrrA74YJlh9r079973psed0cvUD4A+4HoCjsA85X0Dfn/RBLBgBDxnQPfAEJgBY+A9gALA4tcbamSzS4xq4FOQAJgCDwV2CPKV8tZAJcAjMMkUe1vX+U+SMhfAJEHasQIWmXNN3abzDwHUrgcRGmYcgKe0bxrblHEB4E/pndMazNpSZGcsZdBlYJcEL9Afo75molJyM2FxmPgmgPqlWNLGfwZGG6UiyEvLzHYDmoPkDDiNm9JR9uboiONcBXrpY1qmgs21x1QwyZcpvxt9NS09PlsPAAAAAElFTkSuQmCC&logoWidth=14" alt="Duplicate Space"></a></h3> | |
| Join us : 🌟TeamTonic🌟 is always making cool demos! Join our active builder's🛠️community 👻 [](https://discord.gg/GWpVpekp) On 🤗Huggingface: [TeamTonic](https://huggingface.co/TeamTonic) & [MultiTransformer](https://huggingface.co/MultiTransformer) On 🌐Github: [Polytonic](https://github.com/tonic-ai) & contribute to 🌟 [Poly](https://github.com/tonic-ai/poly) 🤗Big thanks to Yuvi Sharma and all the folks at huggingface for the community grant 🤗 | |
| """ | |
| os.environ['PYTORCH_CUDA_ALLOC_CONF'] = 'max_split_size_mb:30' | |
| device = torch.device("cuda" if torch.cuda.is_available() else "cpu") | |
| tasks = { | |
| 'ArguAna': 'Given a claim, find documents that refute the claim', | |
| 'ClimateFEVER': 'Given a claim about climate change, retrieve documents that support or refute the claim', | |
| 'DBPedia': 'Given a query, retrieve relevant entity descriptions from DBPedia', | |
| 'FEVER': 'Given a claim, retrieve documents that support or refute the claim', | |
| 'FiQA2018': 'Given a financial question, retrieve user replies that best answer the question', | |
| 'HotpotQA': 'Given a multi-hop question, retrieve documents that can help answer the question', | |
| 'MSMARCO': 'Given a web search query, retrieve relevant passages that answer the query', | |
| 'NFCorpus': 'Given a question, retrieve relevant documents that best answer the question', | |
| 'NQ': 'Given a question, retrieve Wikipedia passages that answer the question', | |
| 'QuoraRetrieval': 'Given a question, retrieve questions that are semantically equivalent to the given question', | |
| 'SCIDOCS': 'Given a scientific paper title, retrieve paper abstracts that are cited by the given paper', | |
| 'SciFact': 'Given a scientific claim, retrieve documents that support or refute the claim', | |
| 'Touche2020': 'Given a question, retrieve detailed and persuasive arguments that answer the question', | |
| 'TRECCOVID': 'Given a query on COVID-19, retrieve documents that answer the query', | |
| } | |
| tokenizer = AutoTokenizer.from_pretrained('intfloat/e5-mistral-7b-instruct') | |
| model = AutoModel.from_pretrained('intfloat/e5-mistral-7b-instruct', torch_dtype=torch.float16, device_map=device) | |
| def last_token_pool(last_hidden_states: Tensor, attention_mask: Tensor) -> Tensor: | |
| left_padding = (attention_mask[:, -1].sum() == attention_mask.shape[0]) | |
| if left_padding: | |
| return last_hidden_states[:, -1] | |
| else: | |
| sequence_lengths = attention_mask.sum(dim=1) - 1 | |
| batch_size = last_hidden_states.shape[0] | |
| return last_hidden_states[torch.arange(batch_size, device=last_hidden_states.device), sequence_lengths] | |
| class EmbeddingModel: | |
| def __init__(self): | |
| self.tokenizer = AutoTokenizer.from_pretrained('intfloat/e5-mistral-7b-instruct') | |
| self.model = AutoModel.from_pretrained('intfloat/e5-mistral-7b-instruct', torch_dtype=torch.float16, device_map=device) | |
| def compute_embeddings(selected_task, input_text, system_prompt): | |
| max_length = 2042 | |
| task_description = tasks[selected_task] | |
| processed_texts = [f'Instruct: {task_description}\nQuery: {input_text}'] | |
| batch_dict = tokenizer(processed_texts, max_length=max_length - 1, return_attention_mask=False, padding=False, truncation=True) | |
| batch_dict['input_ids'] = [input_ids + [tokenizer.eos_token_id] for input_ids in batch_dict['input_ids']] | |
| batch_dict = tokenizer.pad(batch_dict, padding=True, return_attention_mask=True, return_tensors='pt') | |
| batch_dict = {k: v.to(device) for k, v in batch_dict.items()} | |
| outputs = model(**batch_dict) | |
| embeddings = last_token_pool(outputs.last_hidden_state, batch_dict['attention_mask']) | |
| embeddings = F.normalize(embeddings, p=2, dim=1) | |
| embeddings_list = embeddings.detach().cpu().numpy().tolist() | |
| return embeddings_list | |
| def compute_similarity(self, sentence1, sentence2, extra_sentence1, extra_sentence2): | |
| sentences = [sentence1, sentence2, extra_sentence1, extra_sentence2] | |
| encoded_input = self.tokenizer(sentences, padding=True, truncation=True, return_tensors='pt').to(device) | |
| with torch.no_grad(): | |
| model_output = self.model(**encoded_input) | |
| embeddings = last_token_pool(model_output.last_hidden_state, encoded_input['attention_mask']) | |
| embeddings = F.normalize(embeddings, p=2, dim=1) | |
| # Compute cosine similarity | |
| similarity1 = F.cosine_similarity(embeddings[0].unsqueeze(0), embeddings[1].unsqueeze(0)).item() | |
| similarity2 = F.cosine_similarity(embeddings[2].unsqueeze(0), embeddings[3].unsqueeze(0)).item() | |
| return similarity1, similarity2 | |
| def app_interface(): | |
| with gr.Blocks() as demo: | |
| gr.Markdown(title) | |
| gr.Markdown(description) | |
| with gr.Row(): | |
| task_dropdown = gr.Dropdown(list(tasks.keys()), label="Select a Task", value=list(tasks.keys())[0]) | |
| with gr.Tab("Embedding Generation"): | |
| input_text_box = gr.Textbox(label="📖Input Text") | |
| system_prompt_box = gr.Textbox(label="🤖System Prompt (Optional)") | |
| compute_button = gr.Button("Try🐣🛌🏻e5") | |
| output_display = gr.Textbox(label="🐣e5-mistral🛌🏻 Embeddings") | |
| compute_button.click( | |
| fn=EmbeddingModel.compute_embeddings, | |
| inputs=[task_dropdown, input_text_box, system_prompt_box], | |
| outputs=output_display | |
| ) | |
| with gr.Tab("Sentence Similarity"): | |
| sentence1_box = gr.Textbox(label="Sentence 1") | |
| sentence2_box = gr.Textbox(label="Sentence 2") | |
| extra_sentence1_box = gr.Textbox(label="Sentence 3") | |
| extra_sentence2_box = gr.Textbox(label="Sentence 4") | |
| similarity_button = gr.Button("Compute Similarity") | |
| similarity_output = gr.Label(label="🐣e5-mistral🛌🏻 Similarity Scores") | |
| similarity_button.click( | |
| fn=EmbeddingModel.compute_similarity, | |
| inputs=[sentence1_box, sentence2_box, extra_sentence1_box, extra_sentence2_box], | |
| outputs=similarity_output | |
| ) | |
| with gr.Row(): | |
| with gr.Column(): | |
| system_prompt_box | |
| input_text_box | |
| with gr.Column(): | |
| compute_button | |
| output_display | |
| return demo | |
| # Run the Gradio app | |
| app_interface().launch() |