Update app.py

app.py

@@ -1,82 +1,82 @@
+import spaces
+import torch
+import torch.nn.functional as F
+from torch import Tensor
+from transformers import AutoTokenizer, AutoModel
 import gradio as gr
-from transformers import pipeline
 import os
-import torch
 
-# Set max_split_size_mb
-os.environ['PYTORCH_CUDA_ALLOC_CONF'] = 'max_split_size_mb:50'
-title = """# 🙋🏻‍♂️Welcome to🌟Tonic's Nexus🐦‍⬛Raven"""
-description = """You can build with this endpoint using Nexus Raven. The demo is still a work in progress but we hope to add some endpoints for commonly used functions such as intention mappers and audiobook processing.
-        You can also use Nexus🐦‍⬛Raven on your laptop & by cloning this space. 🧬🔬🔍 Simply click here: <a style="display:inline-block" href="https://huggingface.co/spaces/Tonic1/NexusRaven2?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 on 👻Discord: [Discord](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 🌟 [PolyGPT](https://github.com/tonic-ai/polygpt-alpha)
+title = """
+# 👋🏻Welcome to 🙋🏻‍♂️Tonic's 🐣e5-mistral🛌🏻Embeddings """
+description = """
+You can use this Space to test out the current model [intfloat/e5-mistral-7b-instruct](https://huggingface.co/intfloat/e5-mistral-7b-instruct). e5mistral has a larger context window, a different prompting/return mechanism and generally better results than other embedding models. 
+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 on 👻Discord:  [![Let's build the future of AI together! 🚀🤖](https://discordapp.com/api/guilds/1109943800132010065/widget.png)](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)
 """
-raven_pipeline = pipeline(
-    "text-generation",
-    model="Nexusflow/NexusRaven-V2-13B",
-    torch_dtype="auto",
-    device_map="auto",
-)
 
-class DialogueToSpeechConverter:
-    def __init__(self):
-        self.raven_pipeline = raven_pipeline
+os.environ['PYTORCH_CUDA_ALLOC_CONF'] = 'max_split_size_mb:20'
 
-    def process_text(self, input_text: str) -> str:
-        prompt = f"User Query: {input_text}<human_end>"
-        result = self.raven_pipeline(prompt, max_new_tokens=2048, return_full_text=False, do_sample=False, temperature=0.001)[0]["generated_text"]
-        torch.cuda.empty_cache()
-        return result
+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]
 
-def create_interface():
-    converter = DialogueToSpeechConverter()
-    with gr.Blocks() as app:
-        gr.Markdown(title)
-        gr.Markdown(description)
-        with gr.Row():
-            input_text = gr.Textbox(label="Input Text")
-            submit_button = gr.Button("Submit")
-            output_text = gr.Textbox(label="Nexus🐦‍⬛Raven")
-        submit_button.click(converter.process_text, inputs=input_text, outputs=output_text)
-    return app
+def get_detailed_instruct(task_description: str, query: str) -> str:
+    return f'Instruct: {task_description}\nQuery: {query}'
 
-if __name__ == "__main__":
-    converter = DialogueToSpeechConverter()
-    demo = gr.Interface(
-        fn=converter.process_text,
-        inputs="text",
-        outputs="text",
-        examples=[
-           ['''
-        Function:
-        def create_audio_sequence_order(text):
-            """
-            Analyzes the text and creates an order for each character and narrator segment.
 
-            Args:
-            text (str): The text containing the dialogues and narration.
+@spaces.GPU
+def compute_embeddings(*input_texts):
+    torch.backends.cuda.matmul.allow_tf32 = True
+    torch.backends.cudnn.allow_tf32 = True
+    torch.backends.cudnn.benchmark = True
+    tokenizer = AutoTokenizer.from_pretrained('intfloat/e5-mistral-7b-instruct')
+    model = AutoModel.from_pretrained('intfloat/e5-mistral-7b-instruct')
+    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+    model.to(device)
+    max_length = 4096
+    task = 'Given a web search query, retrieve relevant passages that answer the query'
 
-            Returns:
-            list: A list of tuples, each containing the character/narrator name and a segment of their dialogue/narration.
-            """
-        Function:
-        def convert_text_to_speech_single_voice(text, voice):
-            """
-            Converts a given text to speech using a specified voice. This function is used when there is only one character in the text.
+    processed_texts = [get_detailed_instruct(task, text) for text in input_texts]
+    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')
+    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 app_interface():
+    with gr.Blocks() as demo:
+        gr.Markdown(title)
+        gr.Markdown(description)
+        
+        input_text_boxes = [gr.Textbox(label=f"Input Text {i+1}") for i in range(4)]
+        
+        compute_button = gr.Button("Compute Embeddings")
+        
+        output_display = gr.Dataframe(headers=["Embedding Value"], datatype=["number"])
+        
+        with gr.Row():
+            with gr.Column():
+                for text_box in input_text_boxes:
+                    text_box
+            with gr.Column():
+                compute_button
+                output_display
 
-            Args:
-            text (str): The text to be converted to speech.
-            voice (str): The voice to be used for the audio generation.
+        compute_button.click(
+            fn=compute_embeddings,
+            inputs=input_text_boxes,
+            outputs=output_display
+        )
 
-            Returns:
-            str: The path to the generated speech MP3 file.
-            """
+    return demo
 
-    User Query: Currently, one way that the wealthy distinguish themselves from others is through the collection of rare objects. In a Celebration Society, to own an “original” of something will remain significant. However, barring a desire to prevent others from enjoying the experience, it will become possible to have perfect replicas of all manner of objects including paintings and sculptures.
-There will still be pride of ownership in the original. Others will be able to fully enjoy the “same” piece as well.
-use either speech to single voice if there's no dialogue or create_audio_sequence_order if there is dialogue<human_end>
-    ''']
-        ],
-        title=title,
-        description=description
-    )
-    demo.launch()
+# Run the Gradio app
+app_interface().launch()