|
|
|
|
|
|
|
|
|
|
|
from dataclasses import dataclass |
|
from typing import Optional, Union, Tuple, List |
|
|
|
import torch |
|
import torch.nn as nn |
|
from torch import Tensor |
|
from torch.nn import functional as F |
|
import time |
|
|
|
def find_multiple(n: int, k: int) -> int: |
|
if n % k == 0: |
|
return n |
|
return n + k - (n % k) |
|
|
|
class AdaptiveLayerNorm(nn.Module): |
|
r"""Adaptive Layer Normalization""" |
|
|
|
def __init__(self, d_model, norm) -> None: |
|
super(AdaptiveLayerNorm, self).__init__() |
|
self.linear = nn.Linear(d_model, 6 * d_model) |
|
self.act = nn.SiLU() |
|
self.norm = norm |
|
self.d_model = d_model |
|
self.eps = self.norm.eps |
|
|
|
def forward(self, x: Tensor, emb: Tensor) -> Tuple[Tensor]: |
|
emb = self.linear(self.act(emb)) |
|
shift_msa, scale_msa, gate_msa, shift_mlp, scale_mlp, gate_mlp = torch.chunk(emb, 6, dim=-1) |
|
|
|
x = self.norm(x) * (1 + scale_msa) + shift_msa |
|
return x, gate_msa, shift_mlp, scale_mlp, gate_mlp |
|
|
|
class AdaptiveLayerNormFinal(nn.Module): |
|
r"""Adaptive Layer Normalization""" |
|
|
|
def __init__(self, d_model, norm) -> None: |
|
super(AdaptiveLayerNormFinal, self).__init__() |
|
self.linear = nn.Linear(d_model, 2 * d_model) |
|
self.act = nn.SiLU() |
|
self.norm = norm |
|
self.d_model = d_model |
|
self.eps = self.norm.eps |
|
|
|
def forward(self, x: Tensor, emb: Tensor) -> Tuple[Tensor]: |
|
emb = self.linear(self.act(emb)) |
|
scale, shift = torch.chunk(emb, 2, dim=-1) |
|
|
|
x = self.norm(x) * (1 + scale) + shift |
|
return x |
|
|
|
@dataclass |
|
class ModelArgs: |
|
block_size: int = 2048 |
|
vocab_size: int = 32000 |
|
n_layer: int = 32 |
|
n_head: int = 32 |
|
dim: int = 4096 |
|
intermediate_size: int = None |
|
n_local_heads: int = -1 |
|
head_dim: int = 64 |
|
rope_base: float = 10000 |
|
norm_eps: float = 1e-5 |
|
uvit_skip_connection: bool = False |
|
time_as_token: bool = False |
|
dropout_rate: float = 0.1 |
|
attn_dropout_rate: float = 0.1 |
|
|
|
def __post_init__(self): |
|
if self.n_local_heads == -1: |
|
self.n_local_heads = self.n_head |
|
if self.intermediate_size is None: |
|
hidden_dim = 4 * self.dim |
|
n_hidden = int(2 * hidden_dim / 3) |
|
self.intermediate_size = find_multiple(n_hidden, 256) |
|
|
|
|
|
class Transformer(nn.Module): |
|
def __init__(self, config: ModelArgs) -> None: |
|
super().__init__() |
|
self.config = config |
|
|
|
self.layers = nn.ModuleList(TransformerBlock(config) for _ in range(config.n_layer)) |
|
self.norm = AdaptiveLayerNormFinal(config.dim, RMSNorm(config.dim, eps=config.norm_eps)) |
|
|
|
self.max_batch_size = -1 |
|
self.max_seq_length = config.block_size |
|
|
|
self.uvit_skip_connection = self.config.uvit_skip_connection |
|
if self.uvit_skip_connection: |
|
self.layers_emit_skip = [i for i in range(self.config.n_layer) if i < self.config.n_layer // 2] |
|
self.layers_receive_skip = [i for i in range(self.config.n_layer) if i > self.config.n_layer // 2] |
|
else: |
|
self.layers_emit_skip = [] |
|
self.layers_receive_skip = [] |
|
freqs_cis = precompute_freqs_cis(self.config.block_size, self.config.head_dim, |
|
self.config.rope_base) |
|
self.register_buffer("freqs_cis", freqs_cis) |
|
|
|
causal_mask = torch.tril( |
|
torch.ones(self.max_seq_length, self.max_seq_length, dtype=torch.bool) |
|
) |
|
self.register_buffer("causal_mask", causal_mask) |
|
|
|
def forward(self, |
|
x: Tensor, |
|
c: Tensor, |
|
input_pos: Optional[Tensor] = None, |
|
mask: Optional[Tensor] = None, |
|
) -> Tensor: |
|
mask = mask[..., input_pos] |
|
freqs_cis = self.freqs_cis[input_pos] |
|
for i, layer in enumerate(self.layers): |
|
x = layer(x, c, freqs_cis, mask) |
|
x = self.norm(x, c) |
|
return x |
|
|
|
|
|
class TransformerBlock(nn.Module): |
|
def __init__(self, config: ModelArgs) -> None: |
|
super().__init__() |
|
self.attention = Attention(config) |
|
self.feed_forward = FeedForward(config) |
|
self.ffn_norm = RMSNorm(config.dim, eps=config.norm_eps) |
|
self.attention_norm = AdaptiveLayerNorm(config.dim, RMSNorm(config.dim, eps=config.norm_eps)) |
|
def forward(self, |
|
x: Tensor, |
|
c: Tensor, |
|
freqs_cis: Tensor, |
|
mask: Tensor, |
|
) -> Tensor: |
|
normed_x, gate_msa, shift_mlp, scale_mlp, gate_mlp = self.attention_norm(x, emb=c) |
|
|
|
attn_output = self.attention(normed_x, freqs_cis, mask) |
|
x = x + gate_msa * attn_output |
|
normed_x = self.ffn_norm(x) * (1 + scale_mlp) + shift_mlp |
|
ff_output = self.feed_forward(normed_x) |
|
x = x + gate_mlp * ff_output |
|
return x |
|
|
|
|
|
class Attention(nn.Module): |
|
def __init__(self, config: ModelArgs, is_cross_attention: bool = False): |
|
super().__init__() |
|
assert config.dim % config.n_head == 0 |
|
|
|
total_head_dim = (config.n_head + 2 * config.n_local_heads) * config.head_dim |
|
|
|
if is_cross_attention: |
|
self.wq = nn.Linear(config.dim, config.n_head * config.head_dim, bias=False) |
|
self.wkv = nn.Linear(config.context_dim, 2 * config.n_local_heads * config.head_dim, bias=False) |
|
else: |
|
self.wqkv = nn.Linear(config.dim, total_head_dim, bias=False) |
|
self.wo = nn.Linear(config.head_dim * config.n_head, config.dim, bias=False) |
|
self.kv_cache = None |
|
|
|
self.n_head = config.n_head |
|
self.head_dim = config.head_dim |
|
self.n_local_heads = config.n_local_heads |
|
self.dim = config.dim |
|
self.attn_dropout_rate = config.attn_dropout_rate |
|
|
|
def forward(self, |
|
x: Tensor, |
|
freqs_cis: Tensor, |
|
mask: Tensor, |
|
context: Optional[Tensor] = None, |
|
context_freqs_cis: Optional[Tensor] = None, |
|
) -> Tensor: |
|
bsz, seqlen, _ = x.shape |
|
|
|
kv_size = self.n_local_heads * self.head_dim |
|
q, k, v = self.wqkv(x).split([kv_size, kv_size, kv_size], dim=-1) |
|
context_seqlen = seqlen |
|
|
|
q = q.view(bsz, seqlen, self.n_head, self.head_dim) |
|
k = k.view(bsz, context_seqlen, self.n_local_heads, self.head_dim) |
|
v = v.view(bsz, context_seqlen, self.n_local_heads, self.head_dim) |
|
|
|
q = apply_rotary_emb(q, freqs_cis) |
|
k = apply_rotary_emb(k, freqs_cis) |
|
|
|
q, k, v = map(lambda x: x.transpose(1, 2), (q, k, v)) |
|
|
|
k = k.repeat_interleave(self.n_head // self.n_local_heads, dim=1) |
|
v = v.repeat_interleave(self.n_head // self.n_local_heads, dim=1) |
|
y = F.scaled_dot_product_attention(q, k, v, attn_mask=mask, dropout_p=0.0) |
|
|
|
y = y.transpose(1, 2).contiguous().view(bsz, seqlen, self.head_dim * self.n_head) |
|
|
|
y = self.wo(y) |
|
return y |
|
|
|
|
|
class FeedForward(nn.Module): |
|
def __init__(self, config: ModelArgs) -> None: |
|
super().__init__() |
|
self.w1 = nn.Linear(config.dim, config.intermediate_size, bias=False) |
|
self.w3 = nn.Linear(config.dim, config.intermediate_size, bias=False) |
|
self.w2 = nn.Linear(config.intermediate_size, config.dim, bias=False) |
|
self.dropout = nn.Dropout(config.dropout_rate) |
|
|
|
def forward(self, x: Tensor) -> Tensor: |
|
return self.w2(self.dropout(F.silu(self.w1(x)) * self.w3(x))) |
|
|
|
|
|
class RMSNorm(nn.Module): |
|
def __init__(self, dim: int, eps: float = 1e-5): |
|
super().__init__() |
|
self.eps = eps |
|
self.weight = nn.Parameter(torch.ones(dim)) |
|
|
|
def _norm(self, x): |
|
return x * torch.rsqrt(torch.mean(x * x, dim=-1, keepdim=True) + self.eps) |
|
|
|
def forward(self, x: Tensor) -> Tensor: |
|
output = self._norm(x.float()).type_as(x) |
|
return output * self.weight |
|
|
|
|
|
def precompute_freqs_cis( |
|
seq_len: int, n_elem: int, base: int = 10000, |
|
dtype: torch.dtype = torch.bfloat16 |
|
) -> Tensor: |
|
freqs = 1.0 / (base ** (torch.arange(0, n_elem, 2)[: (n_elem // 2)].float() / n_elem)) |
|
t = torch.arange(seq_len, device=freqs.device) |
|
freqs = torch.outer(t, freqs) |
|
freqs_cis = torch.polar(torch.ones_like(freqs), freqs) |
|
cache = torch.stack([freqs_cis.real, freqs_cis.imag], dim=-1) |
|
return cache.to(dtype=dtype) |
|
|
|
|
|
def apply_rotary_emb(x: Tensor, freqs_cis: Tensor) -> Tensor: |
|
xshaped = x.float().reshape(*x.shape[:-1], -1, 2) |
|
freqs_cis = freqs_cis.view(1, xshaped.size(1), 1, xshaped.size(3), 2) |
|
x_out2 = torch.stack( |
|
[ |
|
xshaped[..., 0] * freqs_cis[..., 0] - xshaped[..., 1] * freqs_cis[..., 1], |
|
xshaped[..., 1] * freqs_cis[..., 0] + xshaped[..., 0] * freqs_cis[..., 1], |
|
], |
|
-1, |
|
) |
|
|
|
x_out2 = x_out2.flatten(3) |
|
return x_out2.type_as(x) |
|
|
|
|