I have this implementation of OpenAI GPT-2 LLM in PyTorch. Can you please help me write a training loop for it, if my dataset looks like this text file:
<|user|>How are you?<|bot|>I’m fine<|endoftext|>
<|user|>What films do you like?<|bot|>I like interstellar<|endoftext|>
Here’s my simple model implementation in a single file model.py:
class GPTConfig:
def __init__(self):
self.n_layer = n_layer
self.n_embd = n_embd
self.vocab_size = vocab_size
self.n_positions = 1024
self.n_ctx = 1024
self.n_head = n_head
self.layer_norm_epsilon = 1e-5
# Define the Gelu activation function
def gelu(x):
return 0.5 * x * (1 + torch.tanh(math.sqrt(2 / math.pi) * (x + 0.044715 * torch.pow(x, 3))))
# Define the LayerNorm module
class LayerNorm(nn.Module):
def __init__(self, hidden_size, eps=1e-12):
super(LayerNorm, self).__init__()
self.weight = nn.Parameter(torch.ones(hidden_size))
self.bias = nn.Parameter(torch.zeros(hidden_size))
self.variance_epsilon = eps
def forward(self, x):
u = x.mean(-1, keepdim=True)
s = (x - u).pow(2).mean(-1, keepdim=True)
x = (x - u) / torch.sqrt(s + self.variance_epsilon)
return self.weight * x + self.bias
# Define the Conv1D module
class Conv1D(nn.Module):
def __init__(self, nf, nx):
super(Conv1D, self).__init__()
self.nf = nf
w = torch.empty(nx, nf)
nn.init.normal_(w, std=0.02)
self.weight = nn.Parameter(w)
self.bias = nn.Parameter(torch.zeros(nf))
def forward(self, x):
size_out = x.size()[:-1] + (self.nf,)
x = torch.addmm(self.bias, x.view(-1, x.size(-1)), self.weight)
x = x.view(*size_out)
return x
# Define the Attention module
class Attention(nn.Module):
def __init__(self, nx, n_ctx, config, scale=False):
super(Attention, self).__init__()
n_state = nx # in Attention: n_state=768 (nx=n_embd)
assert n_state % n_head == 0
self.register_buffer("bias", torch.tril(torch.ones(n_ctx, n_ctx)).view(1, 1, n_ctx, n_ctx))
self.n_head = n_head
self.split_size = n_state
self.scale = scale
self.c_attn = Conv1D(n_state * 3, nx)
self.c_proj = Conv1D(n_state, nx)
def _attn(self, q, k, v):
w = torch.matmul(q, k)
if self.scale:
w = w / math.sqrt(v.size(-1))
nd, ns = w.size(-2), w.size(-1)
b = self.bias[:, :, ns-nd:ns, :ns]
w = w * b - 1e10 * (1 - b)
w = nn.Softmax(dim=-1)(w)
return torch.matmul(w, v)
def merge_heads(self, x):
x = x.permute(0, 2, 1, 3).contiguous()
new_x_shape = x.size()[:-2] + (x.size(-2) * x.size(-1),)
return x.view(*new_x_shape)
def split_heads(self, x, k=False):
new_x_shape = x.size()[:-1] + (self.n_head, x.size(-1) // self.n_head)
x = x.view(*new_x_shape)
if k:
return x.permute(0, 2, 3, 1)
else:
return x.permute(0, 2, 1, 3)
def forward(self, x, layer_past=None):
x = self.c_attn(x)
query, key, value = x.split(self.split_size, dim=2)
query = self.split_heads(query)
key = self.split_heads(key, k=True)
value = self.split_heads(value)
if layer_past is not None:
past_key, past_value = layer_past[0].transpose(-2, -1), layer_past[1]
key = torch.cat((past_key, key), dim=-1)
value = torch.cat((past_value, value), dim=-2)
present = torch.stack((key.transpose(-2, -1), value))
a = self._attn(query, key, value)
a = self.merge_heads(a)
a = self.c_proj(a)
return a, present
# Define the MLP module
class MLP(nn.Module):
def __init__(self, n_state, config):
super(MLP, self).__init__()
nx = n_embd
self.c_fc = Conv1D(n_state, nx)
self.c_proj = Conv1D(nx, n_state)
self.act = gelu
def forward(self, x):
h = self.act(self.c_fc(x))
h2 = self.c_proj(h)
return h2
# Define the Transformer Block module
class Block(nn.Module):
def __init__(self, n_ctx, config, scale=False):
super(Block, self).__init__()
nx = n_embd
self.ln_1 = LayerNorm(nx, eps=layer_norm_epsilon)
self.attn = Attention(nx, n_ctx, config, scale)
self.ln_2 = LayerNorm(nx, eps=layer_norm_epsilon)
self.mlp = MLP(4 * nx, config)
def forward(self, x, layer_past=None):
a, present = self.attn(self.ln_1(x), layer_past=layer_past)
x = x + a
m = self.mlp(self.ln_2(x))
x = x + m
return x, present
# Define the GPT2Model module
class GPT2Model(nn.Module):
def __init__(self, config):
super(GPT2Model, self).__init__()
self.n_layer = n_layer
self.n_embd = n_embd
self.n_vocab = vocab_size
self.wte = nn.Embedding(vocab_size, n_embd)
self.wpe = nn.Embedding(n_positions, n_embd)
block = Block(n_ctx, config, scale=True)
self.h = nn.ModuleList([copy.deepcopy(block) for _ in range(n_layer)])
self.ln_f = LayerNorm(n_embd, eps=layer_norm_epsilon)
def set_embeddings_weights(self, model_embeddings_weights):
embed_shape = model_embeddings_weights.shape
self.decoder = nn.Linear(embed_shape[1], embed_shape[0], bias=False)
self.decoder.weight = model_embeddings_weights # Tied weights
def forward(self, input_ids, position_ids=None, token_type_ids=None, past=None):
if past is None:
past_length = 0
past = [None] * len(self.h)
else:
past_length = past[0][0].size(-2)
if position_ids is None:
position_ids = torch.arange(past_length, input_ids.size(-1) + past_length, dtype=torch.long,
device=input_ids.device)
input_shape = input_ids.size()
input_ids = input_ids.view(-1, input_ids.size(-1))
if position_ids.ndimension() == 1: # Check if position_ids has only one dimension
position_ids = position_ids.unsqueeze(0).expand_as(input_ids)
else:
position_ids = position_ids.view(-1, position_ids.size(-1))
inputs_embeds = self.wte(input_ids)
position_embeds = self.wpe(position_ids)
if token_type_ids is not None:
token_type_ids = token_type_ids.view(-1, token_type_ids.size(-1))
token_type_embeds = self.wte(token_type_ids)
else:
token_type_embeds = 0
hidden_states = inputs_embeds + position_embeds + token_type_embeds
presents = []
for block, layer_past in zip(self.h, past):
hidden_states, present = block(hidden_states, layer_past)
presents.append(present)
hidden_states = self.ln_f(hidden_states)
output_shape = input_shape + (hidden_states.size(-1),)
return hidden_states.view(*output_shape), presents
# Define the GPT2LMHead module
class GPT2LMHead(nn.Module):
def __init__(self, model_embeddings_weights, config):
super(GPT2LMHead, self).__init__()
self.n_embd = n_embd
self.set_embeddings_weights(model_embeddings_weights)
def set_embeddings_weights(self, model_embeddings_weights):
embed_shape = model_embeddings_weights.shape
self.decoder = nn.Linear(embed_shape[1], embed_shape[0], bias=False)
self.decoder.weight = model_embeddings_weights # Tied weights
def forward(self, hidden_state):
lm_logits = self.decoder(hidden_state)
return lm_logits
# Define the GPT2LMHeadModel module (Combining GPT2Model and GPT2LMHead)
class GPT2LMHeadModel(nn.Module):
def __init__(self, config):
super(GPT2LMHeadModel, self).__init__()
self.transformer = GPT2Model(config)
self.lm_head = GPT2LMHead(self.transformer.wte.weight, config)
def set_tied(self):
self.lm_head.set_embeddings_weights(self.transformer.wte.weight)
def forward(self, input_ids, position_ids=None, token_type_ids=None, lm_labels=None, past=None):
hidden_states, presents = self.transformer(input_ids, position_ids, token_type_ids, past)
lm_logits = self.lm_head(hidden_states)
if lm_labels is not None:
loss_fct = nn.CrossEntropyLoss(ignore_index=-1)
loss = loss_fct(lm_logits.view(-1, lm_logits.size(-1)), lm_labels.view(-1))
return loss
return lm_logits, presents
# Define the GPTLanguageModel (Wrapper for GPT2LMHeadModel)
class GPTLanguageModel(nn.Module):
def __init__(self, config):
super(GPTLanguageModel, self).__init__()
self.model = GPT2LMHeadModel(config)
def forward(self, idx, targets=None):
logits, presents = self.model(idx, lm_labels=targets)
loss = None if targets is None else logits
return logits, loss
def generate(self, idx, max_new_tokens):
_, presents = self.model(idx)
B, T = idx.shape
for _ in range(max_new_tokens):
if T <= block_size:
idx_cond = idx
else:
idx_cond = idx[:, -block_size:]
logits, _ = self.model(idx_cond, past=presents)
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)
T += 1
return idx
I want to make 2 versions of training: one is just simple training, and second is make inputs from user to bot token including and output is response + endoftext. And parse each line in iteration
