I was turning the decoder model code with pytorch transformer decoder layer an I am getting different loss even though I tried to match the implementation.
import torch
import torch.nn as nn
from torch.nn import functional as F
# hyperparameters
batch_size = 16 # how many independent sequences will we process in parallel?
block_size = 32 # what is the maximum context length for predictions?
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)
# wget https://raw.githubusercontent.com/karpathy/char-rnn/master/data/tinyshakespeare/input.txt
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
u/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 """
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) # (B,T,C)
q = self.query(x) # (B,T,C)
# compute attention scores ("affinities")
wei = q @ k.transpose(-2,-1) * C**-0.5 # (B, T, C) @ (B, C, T) -> (B, T, T)
wei = wei.masked_fill(self.tril[:T, :T] == 0, float('-inf')) # (B, T, T)
wei = F.softmax(wei, dim=-1) # (B, T, T)
wei = self.dropout(wei)
# perform the weighted aggregation of the values
v = self.value(x) # (B,T,C)
out = wei @ v # (B, T, T) @ (B, T, C) -> (B, T, C)
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):
# n_embd: embedding dimension, n_head: the number of heads we'd like
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
# super simple bigram model
class TransformerDecoder(nn.Module):
def __init__(self):
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)
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)
def forward(self, idx, targets=None):
B, T = idx.shape
# idx and targets are both (B,T) tensor of integers
tok_emb = self.token_embedding_table(idx) # (B,T,C)
pos_emb = self.position_embedding_table(torch.arange(T, device=device)) # (T,C)
x = tok_emb + pos_emb # (B,T,C)
x = self.blocks(x) # (B,T,C)
x = self.ln_f(x) # (B,T,C)
logits = self.lm_head(x) # (B,T,vocab_size)
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):
# idx is (B, T) array of indices in the current context
for _ in range(max_new_tokens):
# crop idx to the last block_size tokens
idx_cond = idx[:, -block_size:]
# get the predictions
logits, loss = self(idx_cond)
# focus only on the last time step
logits = logits[:, -1, :] # becomes (B, C)
# apply softmax to get probabilities
probs = F.softmax(logits, dim=-1) # (B, C)
# sample from the distribution
idx_next = torch.multinomial(probs, num_samples=1) # (B, 1)
# append sampled index to the running sequence
idx = torch.cat((idx, idx_next), dim=1) # (B, T+1)
return idx
model = TransformerDecoder()
m = model.to(device)
# print the number of parameters in the model
print(sum(p.numel() for p in m.parameters())/1e6, 'M parameters')
# create a PyTorch optimizer
optimizer = torch.optim.AdamW(model.parameters(), lr=learning_rate)
for iter in range(max_iters):
# every once in a while evaluate the loss on train and val sets
if iter % eval_interval == 0 or iter == max_iters - 1:
losses = estimate_loss()
print(f"step {iter}: train loss {losses['train']:.4f}, val loss {losses['val']:.4f}")
# sample a batch of data
xb, yb = get_batch('train')
# evaluate the loss
logits, loss = model(xb, yb)
optimizer.zero_grad(set_to_none=True)
loss.backward()
optimizer.step()
# generate from the model
context = torch.zeros((1, 1), dtype=torch.long, device=device)
print(decode(m.generate(context, max_new_tokens=2000)[0].tolist()))
0.203279 M parameters
step 0: train loss 3.0684, val loss 3.0634
step 100: train loss 0.9685, val loss 0.9290
step 200: train loss 0.6753, val loss 0.6564
step 300: train loss 0.5599, val loss 0.5882
step 400: train loss 0.3871, val loss 0.5186
step 500: train loss 0.2547, val loss 0.3679
step 600: train loss 0.2038, val loss 0.2841
step 700: train loss 0.1839, val loss 0.2364
step 800: train loss 0.1704, val loss 0.2210
step 900: train loss 0.1627, val loss 0.2135
step 1000: train loss 0.1568, val loss 0.1984
import torch
import torch.nn as nn
from torch.nn import functional as F
# hyperparameters
batch_size = 16 # how many independent sequences will we process in parallel?
block_size = 32 # what is the maximum context length for predictions?
max_iters = 1000
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)
# Assuming input.txt is already downloaded and available
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
u/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 TransformerDecoderModel(nn.Module):
def __init__(self, vocab_size, n_embd, n_head, n_layer, block_size, dropout=0.1):
super().__init__()
self.vocab_size = vocab_size
self.n_embd = n_embd
self.n_head = n_head
self.n_layer = n_layer
self.block_size = block_size
self.token_embedding = nn.Embedding(vocab_size, n_embd)
self.position_embedding = nn.Embedding(block_size, n_embd)
decoder_layer = nn.TransformerDecoderLayer(d_model=n_embd, nhead=n_head, dropout=dropout)
self.transformer_decoder = nn.TransformerDecoder(decoder_layer, num_layers=n_layer)
self.ln_f = nn.LayerNorm(n_embd) # Final LayerNorm
self.lm_head = nn.Linear(n_embd, vocab_size, bias=False)
def generate_square_subsequent_mask(self, sz):
mask = torch.triu(torch.ones(sz, sz) * float('-inf'), diagonal=1)
return mask
def forward(self, idx, targets=None):
B, T = idx.shape
device = idx.device
# Create mask with shape [block_size, block_size]
mask = self.generate_square_subsequent_mask(T).to(device)
tok_emb = self.token_embedding(idx) # [B, T, C]
pos_emb = self.position_embedding(torch.arange(T, dtype=torch.long, device=device)) # [T, C]
pos_emb = pos_emb.unsqueeze(0).repeat(B, 1, 1) # Expand to match batch size: [B, T, C]
x = tok_emb + pos_emb
# Transformer requires input shape as [T, B, C]
x = x.permute(1, 0, 2) # [T, B, C]
x = self.transformer_decoder(x, x, tgt_mask=mask)
# Revert shape to [B, T, C] for the linear layer
x = x.permute(1, 0, 2)
x = self.ln_f(x)
logits = self.lm_head(x)
if targets is not None:
loss = F.cross_entropy(logits.view(-1, self.vocab_size), targets.view(-1))
return logits, loss
else:
return logits, None
def generate(self, idx, max_new_tokens):
idx = idx.to(device)
for _ in range(max_new_tokens):
idx_cond = idx[:, -self.block_size:]
logits = self.forward(idx_cond)[0]
probs = F.softmax(logits[:, -1, :], dim=-1)
idx_next = torch.multinomial(probs, num_samples=1)
idx = torch.cat([idx, idx_next], dim=1)
return idx
model = TransformerDecoderModel(vocab_size=vocab_size, n_embd=n_embd, n_head=n_head, n_layer=n_layer, block_size=block_size, dropout=dropout)
m = model.to(device)
# print the number of parameters in the model
print(sum(p.numel() for p in m.parameters())/1e6, 'M parameters')
# create a PyTorch optimizer
optimizer = torch.optim.AdamW(model.parameters(), lr=learning_rate)
for iter in range(max_iters):
# every once in a while evaluate the loss on train and val sets
if iter % eval_interval == 0 or iter == max_iters - 1:
losses = estimate_loss()
print(f"step {iter}: train loss {losses['train']:.4f}, val loss {losses['val']:.4f}")
# sample a batch of data
xb, yb = get_batch('train')
# evaluate the loss
logits, loss = model(xb, yb)
optimizer.zero_grad(set_to_none=True)
loss.backward()
optimizer.step()
# generate from the model
context = torch.zeros((1, 1), dtype=torch.long, device=device)
print(decode(m.generate(context, max_new_tokens=2000)[0].tolist()))
step 0: train loss 3.1871, val loss 3.2023
step 100: train loss 0.9042, val loss 0.8612
step 200: train loss 0.1210, val loss 0.1222
step 300: train loss 0.0269, val loss 0.0262
step 400: train loss 0.0178, val loss 0.0231
step 500: train loss 0.0153, val loss 0.0221
step 600: train loss 0.0072, val loss 0.0173
step 700: train loss 0.0110, val loss 0.0102
step 800: train loss 0.0027, val loss 0.0041
step 900: train loss 0.0019, val loss 0.0015
step 999: train loss 0.0006, val loss 0.0006
even though I implemented one thing form scratch and the other with transformer decoder layers both loss is different and I see PyTorch transformer decoder implementation wrong as its generating same token repeatedly why?