1st version

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+RAMEmbeddings.csv filter=lfs diff=lfs merge=lfs -text

RAMEmbeddings.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d6959cd0292204993b7c78133b2c7912b096c6f6b72239b11c7a6fdb5787f031
+size 844524643

app4.py

@@ -0,0 +1,24 @@
+# -*- coding: utf-8 -*-
+"""
+Created on Fri Jun 16 23:31:34 2023
+
+@author: Hua
+"""
+
+import gradio as gr
+
+from searchEmbeddings import search
+
+def searchEMA(inputs):
+    results = search(inputs)
+    outputs = f"(1) {results['Author Full Names'][4]} {results['Article Title'][4]} {results['Publication Year'][4]} \n  [Abstract] {results['Abstract'][4]} \n(2) {results['Author Full Names'][3]} {results['Article Title'][3]} {results['Publication Year'][3]} \n  [Abstract] {results['Abstract'][3]}"
+    return outputs
+
+with gr.Blocks() as demo:
+    gr.Markdown("瑞聚医学数字化设计与制造平台1.0")
+    name = gr.Textbox(label = "Question", placeholder = "What is your problem?")
+    output = gr.Textbox(label = "Recommended solution")
+    greet_btn = gr.Button("Recommened solutions from ramGPT")
+    greet_btn.click(fn=searchEMA, inputs=name, outputs=output, api_name="searchEMA")
+
+demo.launch(share=True, auth=("admin", "ramGPT"))

searchEmbeddings.py

@@ -0,0 +1,49 @@
+# -*- coding: utf-8 -*-
+"""
+Created on Wed Apr 19 14:45:37 2023
+
+@author: Hua
+"""
+
+import pandas as pd
+import json
+from sentence_transformers import SentenceTransformer
+from sentence_transformers import util
+import numpy as np
+
+
+df = pd.read_csv('RAMEmbeddings.csv')
+
+# load gpt2 embeddings
+gpt2bds = df.GPT2Embeddings  # get a pd.Series
+gpt2list = [np.float32(np.array(json.loads(i))) for i in gpt2bds]  # list of embeddings
+
+# define the search function
+def search(inputs):
+    # GPT2 embedding
+    gpt2_model = SentenceTransformer('sembeddings/model_gpt_trained')
+    embeddings = gpt2_model.encode(inputs)
+    
+    # calculate the similarity list to a given embedding
+    sims = []
+    for i in range(len(gpt2list)):
+        sim = util.pytorch_cos_sim(embeddings, gpt2list[i])
+        sims.append(sim.item())
+    
+    # find the top-5 similarity items
+    sims_arr = np.array(sims, dtype=object)
+    inds = np.argpartition(sims_arr, -5)[-5:]
+    
+    # return top 5 items
+    return df.loc[inds].reset_index(drop=True)
+
+    
+    
+    
+
+
+
+
+
+
+