First model version

README.md

@@ -1,13 +1,14 @@
 ---
-title: ISeBetter
-emoji: 📈
-colorFrom: yellow
-colorTo: pink
+title: iSeBetter
+emoji: 👩🏽‍💻
+colorFrom: purple
+colorTo: blue
 sdk: streamlit
-sdk_version: 1.27.2
 app_file: app.py
 pinned: false
 license: mit
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+## iSeBetter: Transformer
+
+Description: A semantic search transformer model fine-tuned on the PyTorch GitHub issues dataset, hosted on Hugging Face, and integrated with Streamlit for easy use.

app.py

@@ -0,0 +1,29 @@
+
+import streamlit as st 
+import torch
+from datasets import load_dataset
+from sentence_transformers import SentenceTransformer, util
+
+embedder = SentenceTransformer('all-mpnet-base-v2')
+
+st.title("iSeBetter : Semantic Transformer")
+st.header("Analyzing Patterns in Text")
+
+
+text_input = st.text_area("Enter the issue details below:")
+
+
+if st.button("Analyse the Issues"):
+    # Perform analysis (your existing code)
+    query_embedding = embedder.encode(text_input, convert_to_tensor=True)
+    corpus_embeddings = torch.load('saved_corpus.pt')
+    corpus_embeddings_name = torch.load('saved_corpus_list.txt')
+    cos_scores = util.cos_sim(query_embedding, corpus_embeddings)[0]
+    top_results = torch.topk(cos_scores, k=5)
+
+    # Results presentation
+    st.subheader("Top 5 Matched Results:")
+    result_table = "<table><tr><th>Matched Text</th><th>Score</th></tr>"
+    for score, idx in zip(top_results[0], top_results[1]):
+        st.markdown(f"- **{corpus_embeddings_name[idx]}** (Score: {score:.4f})")
+        st.progress(score.item())

model.py

@@ -0,0 +1,33 @@
+import pandas as pd
+from sentence_transformers import SentenceTransformer, InputExample, losses
+from torch.utils.data import DataLoader
+from datasets import load_dataset
+
+dataset = load_dataset("gopikrsmscs/torch-issues")
+
+
+# Create InputExamples from your dataset
+examples = []
+for i, row in dataset['train'].iterrows():
+    title = row['Title']
+    body = row['Body']
+    examples.append(InputExample(texts=[title, body]))
+
+# Load the pre-trained model
+model = SentenceTransformer('sentence-transformers/all-mpnet-base-v2')
+
+# Define a DataLoader for training
+train_dataloader = DataLoader(examples, shuffle=True, batch_size=16)
+
+# Fine-tune the model
+train_loss = losses.CosineSimilarityLoss(model)
+
+model.fit(
+    train_objectives=[(train_dataloader, train_loss)],
+    epochs=2,  # You can adjust the number of training epochs
+    warmup_steps=100,
+    optimizer_params={'lr': 1e-4},
+)
+
+# Save the fine-tuned model
+model.save('iSeBetter')

model1.py

@@ -0,0 +1,32 @@
+from sentence_transformers import SentenceTransformer, util
+import torch
+import pandas as pd
+from sentence_transformers import SentenceTransformer, InputExample, losses
+from torch.utils.data import DataLoader
+from datasets import load_dataset
+import pinecone
+
+embedder = SentenceTransformer('all-mpnet-base-v2')
+
+dataset = load_dataset("gopikrsmscs/torch-issues")
+
+columns = dataset['train'].column_names
+columns_to_keep = ["Title", "Body"]
+columns_to_remove = set(columns_to_keep).symmetric_difference(columns)
+dataset = dataset.remove_columns(columns_to_remove)
+
+dataset_dict = dataset['train']
+
+examples = []
+for row in dataset_dict:
+    title = row['Title']
+    #body = row['Body']
+    examples.append(title)
+
+file_paths = 'saved_corpus.txt'
+torch.save(examples,file_paths)
+corpus_embeddings = embedder.encode(examples, convert_to_tensor=True)
+file_path = 'saved_corpus.pt'
+
+# Save the tensor to the file
+torch.save(corpus_embeddings, file_path)

preparedata.py

@@ -0,0 +1,53 @@
+import requests
+import csv
+
+# Set the GitHub repository and the API endpoint
+repo_url = "https://api.github.com/repos/pytorch/pytorch"
+issues_endpoint = f"{repo_url}/issues"
+
+# Initialize an empty list to store all issues
+all_issues = []
+
+page = 1
+while True:
+    # Set the parameters for the API request
+    params = {
+        'state': 'open',  # Change to 'closed' for closed issues
+        'per_page': 100,  # Number of issues per page (adjust as needed)
+        'page': page
+    }
+
+    # Send the GET request to fetch issues
+    response = requests.get(issues_endpoint, params=params)
+
+    # Check if the request was successful (status code 200)
+    if response.status_code == 200:
+        issues = response.json()
+        if not issues:
+            break  # No more issues to fetch
+        all_issues.extend(issues)
+        page += 1
+    else:
+        print(f"Failed to retrieve issues. Status code: {response.status_code}")
+        break
+
+# Open a CSV file for writing
+with open('torch_github_issues.csv', mode='w', newline='') as csv_file:
+    fieldnames = ['Serial Number', 'Issue Number', 'Title', 'Labels', 'Body','Comments']
+    writer = csv.DictWriter(csv_file, fieldnames=fieldnames)
+    
+    # Write the header row
+    writer.writeheader()
+    
+    # Write the issues data to the CSV file
+    for serial, issue in enumerate(all_issues, start=1):
+        writer.writerow({
+            'Serial Number': serial,
+            'Issue Number': issue['number'],
+            'Title': issue['title'],
+            'Labels': ', '.join(label['name'] for label in issue['labels']),
+            'Body': issue['body'],
+            'Comments': issue['comments']
+        })
+    
+print(f"Data written to github_issues.csv")

requirements.txt

@@ -0,0 +1,5 @@
+streamlit==0.90.0
+torch==1.10.0
+datasets==1.13.0
+sentence-transformers==2.1.0
+transformers==5.5.1

saved_corpus.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5516ae0fa4699b9424f36d8588a69ec4947181d27eac25ca3519824a737366a4
+size 18433205

train.py

@@ -0,0 +1,38 @@
+from transformers import BertTokenizer, BertForSequenceClassification, AdamW
+import torch
+
+# Load a pre-trained BERT model and tokenizer
+tokenizer = BertTokenizer.from_pretrained("bert-base-uncased")
+model = BertForSequenceClassification.from_pretrained("bert-base-uncased")
+
+# Prepare your dataset (replace with your dataset loading code)
+train_texts = ["Text of issue 1", "Text of issue 2", ...]
+labels = [0, 1, ...]  # 0 for non-relevant, 1 for relevant
+
+# Tokenize and convert your dataset to tensors
+input_ids = tokenizer(train_texts, truncation=True, padding=True, return_tensors="pt")
+labels = torch.tensor(labels)
+
+# Set up data loaders
+dataset = torch.utils.data.TensorDataset(input_ids["input_ids"], input_ids["attention_mask"], labels)
+train_loader = torch.utils.data.DataLoader(dataset, batch_size=32)
+
+# Define optimizer and loss
+optimizer = AdamW(model.parameters(), lr=1e-5)
+loss_fn = torch.nn.CrossEntropyLoss()
+
+# Fine-tune the model
+model.train()
+for epoch in range(3):  # Replace with desired number of epochs
+    for batch in train_loader:
+        input_ids, attention_mask, labels = batch
+        optimizer.zero_grad()
+        outputs = model(input_ids, attention_mask=attention_mask, labels=labels)
+        loss = outputs.loss
+        loss.backward()
+        optimizer.step()
+
+# Save the fine-tuned model
+model.save_pretrained("/path/to/save/model")
+
+# You can now use this model for semantic search.