|
|
"""PureBit Transformer - Binary-level language model""" |
|
|
import torch |
|
|
import torch.nn as nn |
|
|
import torch.nn.functional as F |
|
|
import math |
|
|
|
|
|
class Attention(nn.Module): |
|
|
def __init__(self, d, heads=8): |
|
|
super().__init__() |
|
|
self.heads = heads |
|
|
self.dk = d // heads |
|
|
self.q_proj = nn.Linear(d, d, bias=False) |
|
|
self.k_proj = nn.Linear(d, d, bias=False) |
|
|
self.v_proj = nn.Linear(d, d, bias=False) |
|
|
self.out_proj = nn.Linear(d, d, bias=False) |
|
|
|
|
|
def forward(self, x, mask=None): |
|
|
B, N, D = x.shape |
|
|
q = self.q_proj(x).view(B, N, self.heads, self.dk).transpose(1, 2) |
|
|
k = self.k_proj(x).view(B, N, self.heads, self.dk).transpose(1, 2) |
|
|
v = self.v_proj(x).view(B, N, self.heads, self.dk).transpose(1, 2) |
|
|
att = (q @ k.transpose(-1, -2)) / math.sqrt(self.dk) |
|
|
if mask is not None: |
|
|
att = att + mask |
|
|
att = F.softmax(att, dim=-1) |
|
|
out = (att @ v).transpose(1, 2).reshape(B, N, D) |
|
|
return self.out_proj(out) |
|
|
|
|
|
class MLP(nn.Module): |
|
|
def __init__(self, d, mult=4): |
|
|
super().__init__() |
|
|
self.fc1 = nn.Linear(d, d * mult, bias=False) |
|
|
self.fc2 = nn.Linear(d * mult, d, bias=False) |
|
|
|
|
|
def forward(self, x): |
|
|
return self.fc2(F.gelu(self.fc1(x))) |
|
|
|
|
|
class Block(nn.Module): |
|
|
def __init__(self, d, heads=8): |
|
|
super().__init__() |
|
|
self.ln1 = nn.LayerNorm(d) |
|
|
self.attn = Attention(d, heads) |
|
|
self.ln2 = nn.LayerNorm(d) |
|
|
self.mlp = MLP(d) |
|
|
|
|
|
def forward(self, x, mask): |
|
|
x = x + self.attn(self.ln1(x), mask) |
|
|
x = x + self.mlp(self.ln2(x)) |
|
|
return x |
|
|
|
|
|
class PureBitTransformer(nn.Module): |
|
|
"""Transformer operating on raw binary bits (vocab_size=2)""" |
|
|
def __init__(self, d=256, layers=6, heads=8, ctx=4096): |
|
|
super().__init__() |
|
|
self.ctx = ctx |
|
|
self.emb = nn.Embedding(2, d) |
|
|
self.blocks = nn.ModuleList([Block(d, heads) for _ in range(layers)]) |
|
|
self.ln = nn.LayerNorm(d) |
|
|
self.head = nn.Linear(d, 2, bias=False) |
|
|
self.head.weight = self.emb.weight |
|
|
|
|
|
def forward(self, x): |
|
|
B, N = x.shape |
|
|
mask = torch.triu(torch.ones(N, N, device=x.device), 1) * -1e9 |
|
|
h = self.emb(x) |
|
|
for b in self.blocks: |
|
|
h = b(h, mask) |
|
|
return self.head(self.ln(h)) |
|
|
|
|
|
@torch.no_grad() |
|
|
def generate(self, bits, max_new=256, temp=0.8): |
|
|
"""Generate new bits autoregressively""" |
|
|
x = torch.tensor(bits, device=next(self.parameters()).device).unsqueeze(0) |
|
|
for _ in range(max_new): |
|
|
logits = self(x[:, -self.ctx:])[:, -1, :] / temp |
|
|
next_bit = torch.multinomial(F.softmax(logits, -1), 1) |
|
|
x = torch.cat([x, next_bit], 1) |
|
|
return x[0].tolist() |
|
|
|
|
|
def text_to_bits(text): |
|
|
"""Convert UTF-8 text to list of bits""" |
|
|
bits = [] |
|
|
for byte in text.encode('utf-8'): |
|
|
for i in range(7, -1, -1): |
|
|
bits.append((byte >> i) & 1) |
|
|
return bits |
|
|
|
|
|
def bits_to_text(bits): |
|
|
"""Convert list of bits back to UTF-8 text""" |
|
|
while len(bits) % 8 != 0: |
|
|
bits = bits + [0] |
|
|
bytes_out = [] |
|
|
for i in range(0, len(bits), 8): |
|
|
byte = 0 |
|
|
for j in range(8): |
|
|
byte = (byte << 1) | bits[i + j] |
|
|
bytes_out.append(byte) |
|
|
return bytes(bytes_out).decode('utf-8', errors='replace') |
|
|
|
|
|
def load_model(checkpoint_path, device='cuda'): |
|
|
"""Load model from checkpoint""" |
|
|
ckpt = torch.load(checkpoint_path, map_location=device) |
|
|
model = PureBitTransformer(d=256, layers=6, heads=8).to(device) |
|
|
model.load_state_dict(ckpt['model']) |
|
|
model.eval() |
|
|
return model, ckpt |
|
|
|
|
|
if __name__ == "__main__": |
|
|
model = PureBitTransformer() |
|
|
params = sum(p.numel() for p in model.parameters()) |
|
|
print(f"PureBit Transformer: {params:,} parameters") |
|
|
|
