app update

app.py

@@ -1,6 +1,10 @@
 import torch
 import torch.nn as nn
 from torch.nn import functional as F
+import numpy as np
+import random
+import re
+import gradio as gr
 
 # hyperparameters
 batch_size = 16 # how many independent sequences will we process in parallel?
@@ -19,48 +23,48 @@ dropout = 0.0
 torch.manual_seed(1337)
 
 
-with open('input.txt', 'r', encoding='utf-8') as f:
-    text = f.read()
-
-# here are all the unique characters that occur in this text
-chars = sorted(list(set(text)))
-vocab_size = len(chars)
-# create a mapping from characters to integers
-stoi = { ch:i for i,ch in enumerate(chars) }
-itos = { i:ch for i,ch in enumerate(chars) }
-encode = lambda s: [stoi[c] for c in s] # encoder: take a string, output a list of integers
-decode = lambda l: ''.join([itos[i] for i in l]) # decoder: take a list of integers, output a string
-
-# Train and test splits
-data = torch.tensor(encode(text), dtype=torch.long)
-n = int(0.9*len(data)) # first 90% will be train, rest val
-train_data = data[:n]
-val_data = data[n:]
-
-
-# data loading
-def get_batch(split):
-    # generate a small batch of data of inputs x and targets y
-    data = train_data if split == 'train' else val_data
-    ix = torch.randint(len(data) - block_size, (batch_size,))
-    x = torch.stack([data[i:i+block_size] for i in ix])
-    y = torch.stack([data[i+1:i+block_size+1] for i in ix])
-    x, y = x.to(device), y.to(device)
-    return x, y
-
-@torch.no_grad()
-def estimate_loss():
-    out = {}
-    model.eval()
-    for split in ['train', 'val']:
-        losses = torch.zeros(eval_iters)
-        for k in range(eval_iters):
-            X, Y = get_batch(split)
-            logits, loss = model(X, Y)
-            losses[k] = loss.item()
-        out[split] = losses.mean()
-    model.train()
-    return out
+# with open('input.txt', 'r', encoding='utf-8') as f:
+#     text = f.read()
+
+# # here are all the unique characters that occur in this text
+# chars = sorted(list(set(text)))
+# vocab_size = len(chars)
+# # create a mapping from characters to integers
+# stoi = { ch:i for i,ch in enumerate(chars) }
+# itos = { i:ch for i,ch in enumerate(chars) }
+# encode = lambda s: [stoi[c] for c in s] # encoder: take a string, output a list of integers
+# decode = lambda l: ''.join([itos[i] for i in l]) # decoder: take a list of integers, output a string
+
+# # Train and test splits
+# data = torch.tensor(encode(text), dtype=torch.long)
+# n = int(0.9*len(data)) # first 90% will be train, rest val
+# train_data = data[:n]
+# val_data = data[n:]
+
+
+# # data loading
+# def get_batch(split):
+#     # generate a small batch of data of inputs x and targets y
+#     data = train_data if split == 'train' else val_data
+#     ix = torch.randint(len(data) - block_size, (batch_size,))
+#     x = torch.stack([data[i:i+block_size] for i in ix])
+#     y = torch.stack([data[i+1:i+block_size+1] for i in ix])
+#     x, y = x.to(device), y.to(device)
+#     return x, y
+
+# @torch.no_grad()
+# def estimate_loss():
+#     out = {}
+#     model.eval()
+#     for split in ['train', 'val']:
+#         losses = torch.zeros(eval_iters)
+#         for k in range(eval_iters):
+#             X, Y = get_batch(split)
+#             logits, loss = model(X, Y)
+#             losses[k] = loss.item()
+#         out[split] = losses.mean()
+#     model.train()
+#     return out
 
 class Head(nn.Module):
     """ one head of self-attention """
@@ -134,18 +138,24 @@ class Block(nn.Module):
         x = x + self.ffwd(self.ln2(x))
         return x
 
+# super simple bigram model
 # super simple bigram model
 class BigramLanguageModel(nn.Module):
-
-    def __init__(self):
+    def __init__(self, dataset_text, n_embd):
         super().__init__()
-        # each token directly reads off the logits for the next token from a lookup table
-        self.token_embedding_table = nn.Embedding(vocab_size, n_embd)
+
+        # Compute character-related parameters
+        self.chars = sorted(list(set(dataset_text)))
+        self.vocab_size = len(self.chars)
+        self.stoi = {ch: i for i, ch in enumerate(self.chars)}
+        self.itos = {i: ch for ch, i in self.stoi.items()}
+
+        self.token_embedding_table = nn.Embedding(self.vocab_size, n_embd)
         self.position_embedding_table = nn.Embedding(block_size, n_embd)
         self.blocks = nn.Sequential(*[Block(n_embd, n_head=n_head) for _ in range(n_layer)])
-        self.ln_f = nn.LayerNorm(n_embd) # final layer norm
-        self.lm_head = nn.Linear(n_embd, vocab_size)
-
+        self.ln_f = nn.LayerNorm(n_embd)
+        self.lm_head = nn.Linear(n_embd, self.vocab_size)
+        
     def forward(self, idx, targets=None):
         B, T = idx.shape
 
@@ -184,14 +194,42 @@ class BigramLanguageModel(nn.Module):
             idx = torch.cat((idx, idx_next), dim=1) # (B, T+1)
         return idx
 
+# Reading shakespeare data
+with open('input.txt', 'r', encoding='utf-8') as f:
+    shakespeare_text = f.read()
+
+
+# Reading wikipedia data
+DATA_PATH = 'wikisent2.txt'
+# load wikipedia sentences
+with open(DATA_PATH, 'r') as f:
+    lines = f.read().splitlines()
+
+# Selecting 250k lines from the dataset.
+random.seed(42)
+texts = random.choices(lines, k=250000)
+del lines
+
+def preprocess(text):
+    text = re.sub('@.*?\s+', '', text)  # Remove mentions
+    text = re.sub('#.*?\s+', '', text)  # Remove hashtags
+    text = re.sub(r'https?:\/\/.*[\r\n]*', '', text)  # Remove URLs
+    text = re.sub(r'[^\w\s\'.]', '', text)  # Remove special characters except for single quotes and periods
+    text = re.sub('\s+', ' ', text)  # Replace multiple spaces with a single space
+    text = re.sub('^\d+\s*|^\d+\.\d+\s*|^\d+\.\d+\.\d+\s*', '', text)  # Remove digits at the start of sentences
+    text = text.strip()  # Remove leading and trailing whitespace
+    return text
+
+wiki_text = [preprocess(t) for t in texts]
+wiki_text = '\n'.join(wiki_text)
 
 # Load the shakespeaere model
-shakespeare_model = BigramLanguageModel().to(device)  # Initialize an instance of your model
+shakespeare_model = BigramLanguageModel(shakespeare_text, n_embd).to(device)  # Initialize an instance of your model
 shakespeare_model.load_state_dict(torch.load('shakespeaere_language_model.pth', map_location=torch.device('cpu')))
 shakespeare_model.eval()  # Set the model to evaluation mode
 
 # Load the wikipedia model
-wikipedia_model = BigramLanguageModel().to(device)  # Initialize an instance of your model
+wikipedia_model = BigramLanguageModel(wiki_text, n_embd).to(device)  # Initialize an instance of your model
 wikipedia_model.load_state_dict(torch.load('wikipedia_language_model.pth', map_location=torch.device('cpu')))
 wikipedia_model.eval()  # Set the model to evaluation mode
 
@@ -213,7 +251,6 @@ def generate_wikipedia_outputs(prompt=None, max_new_tokens=2000):
   text_output = decode(wikipedia_model.generate(context, max_new_tokens=max_new_tokens)[0].tolist())
   return text_output
     
-import gradio as gr
 
 title = "Nano GPT"
 description = "Nano GPT trained on Shakespeare and Wikipedia datasets. It is trained on a very small amount of data to understand how GPT's are trained and built. <a href='https://github.com/karpathy/nanoGPT'>The implementation can be found here </a>"