-
Notifications
You must be signed in to change notification settings - Fork 0
/
bigram.py
115 lines (93 loc) · 3.38 KB
/
bigram.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
import torch
import torch.nn as nn
from torch.nn import functional as F
torch.manual_seed(1337)
# hyperparameters
max_iters = 20000
lr = 1e-3
eval_iters = 500
batch_size = 32
block_size = 8
device = 'cuda' if torch.cuda.is_available() else 'cpu'
# Read the input corpus
with open('tiny_shakespeare.txt', 'r', encoding='utf-8') as f:
text = f.read()
print("Length of text: ", len(text))
# Create characters as vocabulary
chars = sorted(list(set(text)))
vocab_size = len(chars)
print("Vocabulary: ", ''.join(chars))
print("Vocabulary size: ", vocab_size)
# Encoder and decoder function for idx to char and back
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]
decode = lambda l: ''.join([itos[i] for i in l])
# Convert text to torch tensor
data = torch.tensor(encode(text), dtype=torch.long)
# Split data into train and validation
n = int(0.9 * len(data))
train_data = data[:n]
val_data = data[n:]
# Get single batch of data for training
def get_batch(split='train'):
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():
model.eval()
out = {}
for split in ['train', 'val']:
losses = torch.zeros(eval_iters)
for k in range(eval_iters):
X, Y = get_batch(split=split)
logits, loss = model(X, Y)
losses[k] = loss.item()
out[split] = loss.mean()
model.train()
return out['train'], out['val']
class BigramLanguageModel(nn.Module):
def __init__(self, vocab_size):
super().__init__()
self.token_embedding_table = nn.Embedding(vocab_size, vocab_size)
def forward(self, idx, targets=None):
logits = self.token_embedding_table(idx) # (B, T, C) (4, 8, 65)
if targets is None:
loss = None
else:
B, T, C = logits.shape
logits = logits.view(B * T, C)
targets = targets.view(B * T)
loss = F.cross_entropy(logits, targets)
return logits, loss
def generate(self, idx, max_new_tokens):
for _ in range(max_new_tokens):
logits, loss = self(idx)
logits = logits[:, -1, :]
probs = F.softmax(logits, dim=-1)
idx_next = torch.multinomial(probs, num_samples=1)
idx = torch.cat((idx, idx_next), dim=1)
return idx
# Create the model and optimizer
model = BigramLanguageModel(vocab_size)
m = model.to(device)
optimizer = torch.optim.AdamW(m.parameters(), lr=lr)
# Train the model
for step in range(max_iters):
xb, yb = get_batch(split='train')
logits, loss = m(xb, yb)
optimizer.zero_grad(set_to_none=True)
loss.backward()
optimizer.step()
if ((step + 1) % 1000 == 0):
train_loss, val_loss = estimate_loss()
print(f'Step {step + 1}: train loss {train_loss:.4f}, validation loss {val_loss:.4f}')
print("Final training loss: ", loss.item())
# generate text from the model
print("Sample generation after training: ")
context = torch.zeros((1, 1), dtype=torch.long, device=device)
print(decode(m.generate(context, 500)[0].tolist()))