|
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 |
|
|
|
|
|
batch_size = 16 |
|
block_size = 32 |
|
max_iters = 5000 |
|
eval_interval = 100 |
|
learning_rate = 1e-3 |
|
device = 'cuda' if torch.cuda.is_available() else 'cpu' |
|
eval_iters = 200 |
|
n_embd = 64 |
|
n_head = 4 |
|
n_layer = 4 |
|
dropout = 0.0 |
|
|
|
|
|
torch.manual_seed(1337) |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class Head(nn.Module): |
|
""" one head of self-attention """ |
|
|
|
def __init__(self, head_size): |
|
super().__init__() |
|
self.key = nn.Linear(n_embd, head_size, bias=False) |
|
self.query = nn.Linear(n_embd, head_size, bias=False) |
|
self.value = nn.Linear(n_embd, head_size, bias=False) |
|
self.register_buffer('tril', torch.tril(torch.ones(block_size, block_size))) |
|
|
|
self.dropout = nn.Dropout(dropout) |
|
|
|
def forward(self, x): |
|
B,T,C = x.shape |
|
k = self.key(x) |
|
q = self.query(x) |
|
|
|
wei = q @ k.transpose(-2,-1) * C**-0.5 |
|
wei = wei.masked_fill(self.tril[:T, :T] == 0, float('-inf')) |
|
wei = F.softmax(wei, dim=-1) |
|
wei = self.dropout(wei) |
|
|
|
v = self.value(x) |
|
out = wei @ v |
|
return out |
|
|
|
class MultiHeadAttention(nn.Module): |
|
""" multiple heads of self-attention in parallel """ |
|
|
|
def __init__(self, num_heads, head_size): |
|
super().__init__() |
|
self.heads = nn.ModuleList([Head(head_size) for _ in range(num_heads)]) |
|
self.proj = nn.Linear(n_embd, n_embd) |
|
self.dropout = nn.Dropout(dropout) |
|
|
|
def forward(self, x): |
|
out = torch.cat([h(x) for h in self.heads], dim=-1) |
|
out = self.dropout(self.proj(out)) |
|
return out |
|
|
|
class FeedFoward(nn.Module): |
|
""" a simple linear layer followed by a non-linearity """ |
|
|
|
def __init__(self, n_embd): |
|
super().__init__() |
|
self.net = nn.Sequential( |
|
nn.Linear(n_embd, 4 * n_embd), |
|
nn.ReLU(), |
|
nn.Linear(4 * n_embd, n_embd), |
|
nn.Dropout(dropout), |
|
) |
|
|
|
def forward(self, x): |
|
return self.net(x) |
|
|
|
class Block(nn.Module): |
|
""" Transformer block: communication followed by computation """ |
|
|
|
def __init__(self, n_embd, n_head): |
|
|
|
super().__init__() |
|
head_size = n_embd // n_head |
|
self.sa = MultiHeadAttention(n_head, head_size) |
|
self.ffwd = FeedFoward(n_embd) |
|
self.ln1 = nn.LayerNorm(n_embd) |
|
self.ln2 = nn.LayerNorm(n_embd) |
|
|
|
def forward(self, x): |
|
x = x + self.sa(self.ln1(x)) |
|
x = x + self.ffwd(self.ln2(x)) |
|
return x |
|
|
|
|
|
|
|
class BigramLanguageModel(nn.Module): |
|
def __init__(self, dataset_text, n_embd): |
|
super().__init__() |
|
|
|
|
|
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) |
|
self.lm_head = nn.Linear(n_embd, self.vocab_size) |
|
self.encode = lambda s: [stoi[c] for c in s] |
|
self.decode = lambda l: ''.join([itos[i] for i in l]) |
|
|
|
|
|
def forward(self, idx, targets=None): |
|
B, T = idx.shape |
|
|
|
|
|
tok_emb = self.token_embedding_table(idx) |
|
pos_emb = self.position_embedding_table(torch.arange(T, device=device)) |
|
x = tok_emb + pos_emb |
|
x = self.blocks(x) |
|
x = self.ln_f(x) |
|
logits = self.lm_head(x) |
|
|
|
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): |
|
|
|
idx_cond = idx[:, -block_size:] |
|
|
|
logits, loss = self(idx_cond) |
|
|
|
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 |
|
|
|
|
|
with open('input.txt', 'r', encoding='utf-8') as f: |
|
shakespeare_text = f.read() |
|
|
|
|
|
|
|
DATA_PATH = 'wikisent2.txt' |
|
|
|
with open(DATA_PATH, 'r') as f: |
|
lines = f.read().splitlines() |
|
|
|
|
|
random.seed(42) |
|
texts = random.choices(lines, k=250000) |
|
del lines |
|
|
|
def preprocess(text): |
|
text = re.sub('@.*?\s+', '', text) |
|
text = re.sub('#.*?\s+', '', text) |
|
text = re.sub(r'https?:\/\/.*[\r\n]*', '', text) |
|
text = re.sub(r'[^\w\s\'.]', '', text) |
|
text = re.sub('\s+', ' ', text) |
|
text = re.sub('^\d+\s*|^\d+\.\d+\s*|^\d+\.\d+\.\d+\s*', '', text) |
|
text = text.strip() |
|
return text |
|
|
|
wiki_text = [preprocess(t) for t in texts] |
|
wiki_text = '\n'.join(wiki_text) |
|
|
|
|
|
shakespeare_model = BigramLanguageModel(shakespeare_text, n_embd).to(device) |
|
shakespeare_model.load_state_dict(torch.load('shakespeaere_language_model.pth', map_location=torch.device('cpu'))) |
|
shakespeare_model.eval() |
|
|
|
|
|
wikipedia_model = BigramLanguageModel(wiki_text, n_embd).to(device) |
|
wikipedia_model.load_state_dict(torch.load('wikipedia_language_model.pth', map_location=torch.device('cpu'))) |
|
wikipedia_model.eval() |
|
|
|
|
|
def generate_shakespeare_outputs(prompt=None, max_new_tokens=2000): |
|
if prompt: |
|
context = torch.tensor(shakespeare_model.encode(prompt), dtype=torch.long, device=device).view(1, -1) |
|
else: |
|
context = torch.zeros((1, 1), dtype=torch.long, device=device) |
|
text_output = decode(shakespeare_model.generate(context, max_new_tokens=max_new_tokens)[0].tolist()) |
|
return text_output |
|
|
|
|
|
def generate_wikipedia_outputs(prompt=None, max_new_tokens=2000): |
|
if prompt: |
|
context = torch.tensor(wikipedia_model.encode(prompt), dtype=torch.long, device=device).view(1, -1) |
|
else: |
|
context = torch.zeros((1, 1), dtype=torch.long, device=device) |
|
text_output = decode(wikipedia_model.generate(context, max_new_tokens=max_new_tokens)[0].tolist()) |
|
return text_output |
|
|
|
|
|
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>" |
|
|
|
shakespeare_interface = gr.Interface(generate_shakespeare_outputs, |
|
inputs=[gr.Textbox(label="Enter any prompt ", type="text", value="Once upon a time,"), |
|
gr.Slider(minimum=100, maximum=5000, step=100, value=2000, label="Max new tokens")], |
|
outputs=gr.Textbox(label="Output generated", type="text"), description=description) |
|
|
|
wiki_interface = gr.Interface(generate_wikipedia_outputs, |
|
inputs=[gr.Textbox(label="Enter any prompt ", type="text", value="James Bond"), |
|
gr.Slider(minimum=100, maximum=5000, step=100, value=2000, label="Max new tokens")], |
|
outputs=gr.Textbox(label="Output generated", type="text"), description=description) |
|
|
|
demo = gr.TabbedInterface([shakespeare_interface, wiki_interface], tab_names=["Shakespeare Data", "Wikipedia Data"], |
|
title=title) |
|
|
|
|
|
demo.launch() |
|
|