File size: 10,263 Bytes
f8ab83c f304ad1 f8ab83c f304ad1 f8ab83c f304ad1 f8ab83c f304ad1 f8ab83c | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 | import torch
import torch.nn as nn
import torch.nn.functional as F
_TRITON_AVAILABLE = False
try:
import triton
import triton.language as tl
@triton.jit
def _wkv7_fwd_kernel(
R, K, V, DECAY, A, O,
STATE_OUT, STATE_IN,
sab_scale, T,
stride_b, stride_t, stride_h,
H: tl.constexpr, D: tl.constexpr, BLOCK_D: tl.constexpr,
RETURN_STATE: tl.constexpr, HAS_INIT_STATE: tl.constexpr,
):
pid = tl.program_id(0)
b_idx = pid // H
h_idx = pid % H
base = b_idx * stride_b + h_idx * stride_h
di = tl.arange(0, BLOCK_D)
dj = tl.arange(0, BLOCK_D)
mask_i = di < D
mask_j = dj < D
if HAS_INIT_STATE:
s_off = b_idx * (H * D * D) + h_idx * (D * D)
state_ptrs = STATE_IN + s_off + di[:, None] * D + dj[None, :]
state_mask = mask_i[:, None] & mask_j[None, :]
state = tl.load(state_ptrs, mask=state_mask, other=0.0).to(tl.float32)
else:
state = tl.zeros((BLOCK_D, BLOCK_D), dtype=tl.float32)
for t in range(T):
t_off = base + t * stride_t
kt = tl.load(K + t_off + dj, mask=mask_j, other=0.0).to(tl.float32)
vt = tl.load(V + t_off + di, mask=mask_i, other=0.0).to(tl.float32)
rt = tl.load(R + t_off + dj, mask=mask_j, other=0.0).to(tl.float32)
dt = tl.load(DECAY + t_off + dj, mask=mask_j, other=1.0).to(tl.float32)
at = tl.load(A + t_off + dj, mask=mask_j, other=0.0).to(tl.float32)
sa = tl.sum(state * (-kt)[None, :], axis=1)
ka = kt * at
sab = sa[:, None] * ka[None, :]
state = state * dt[None, :] + sab_scale * sab + vt[:, None] * kt[None, :]
state = tl.minimum(tl.maximum(state, -10.0), 10.0)
out_t = tl.sum(state * rt[None, :], axis=1)
tl.store(O + t_off + di, out_t, mask=mask_i)
if RETURN_STATE:
s_off = b_idx * (H * D * D) + h_idx * (D * D)
state_ptrs = STATE_OUT + s_off + di[:, None] * D + dj[None, :]
state_mask = mask_i[:, None] & mask_j[None, :]
tl.store(state_ptrs, state, mask=state_mask)
def _wkv7_scan_triton(r, decay, k, v, a, sab_scale):
B, T, H, D = r.shape
r, k, v, decay, a = [x.contiguous() for x in (r, k, v, decay, a)]
o = torch.empty_like(r)
stride_b, stride_t, stride_h = T * H * D, H * D, D
BLOCK_D = triton.next_power_of_2(D)
_wkv7_fwd_kernel[(B * H,)](
r, k, v, decay, a, o,
None, None,
float(sab_scale), T,
stride_b, stride_t, stride_h,
H=H, D=D, BLOCK_D=BLOCK_D,
RETURN_STATE=False, HAS_INIT_STATE=False,
)
return o
if torch.cuda.is_available():
_TRITON_AVAILABLE = True
except Exception:
pass
_FLA_AVAILABLE = False
try:
import torch.distributed.tensor as _tdt
if not hasattr(_tdt, 'Replicate'):
try:
from torch.distributed._tensor import Replicate as _R, Shard as _S
_tdt.Replicate = _R; _tdt.Shard = _S
except ImportError:
pass
if not hasattr(_tdt, 'Placement'):
try:
from torch.distributed._tensor.placement_types import Placement as _P
_tdt.Placement = _P
except ImportError:
pass
if not hasattr(_tdt, 'distribute_module'):
_tdt.distribute_module = lambda *a, **kw: None
from fla.ops.rwkv7 import chunk_rwkv7 as _fla_chunk_rwkv7
if torch.cuda.is_available():
_test_r = torch.randn(1, 1, 2, 64, device='cuda', dtype=torch.bfloat16, requires_grad=True)
_test_w = -torch.ones(1, 1, 2, 64, device='cuda', dtype=torch.bfloat16)
_test_o, _ = _fla_chunk_rwkv7(_test_r, _test_w, _test_r, _test_r, _test_r, _test_r,
head_first=False)
_test_o.sum().backward()
if not _test_r.grad.isnan().any():
_FLA_AVAILABLE = True
del _test_r, _test_w, _test_o
torch.cuda.empty_cache()
else:
_FLA_AVAILABLE = True
except Exception:
pass
class BiRWKV7Layer(nn.Module):
def __init__(self, hidden_size, num_heads):
super().__init__()
assert hidden_size % num_heads == 0
self.hidden_size = hidden_size
self.num_heads = num_heads
self.head_size = hidden_size // num_heads
self.mu_r = nn.Parameter(torch.zeros(hidden_size))
self.mu_w = nn.Parameter(torch.zeros(hidden_size))
self.mu_k = nn.Parameter(torch.zeros(hidden_size))
self.mu_v = nn.Parameter(torch.zeros(hidden_size))
self.mu_a = nn.Parameter(torch.zeros(hidden_size))
self.mu_g = nn.Parameter(torch.zeros(hidden_size))
self.W_r = nn.Linear(hidden_size, hidden_size, bias=False)
self.W_k = nn.Linear(hidden_size, hidden_size, bias=False)
self.W_v = nn.Linear(hidden_size, hidden_size, bias=False)
self.W_w = nn.Linear(hidden_size, hidden_size, bias=False)
self.W_a = nn.Linear(hidden_size, hidden_size, bias=False)
self.W_g = nn.Linear(hidden_size, hidden_size, bias=False)
self.sab_gate = nn.Parameter(torch.tensor(-5.0))
self.group_norm = nn.GroupNorm(num_heads, hidden_size)
self.W_o = nn.Linear(hidden_size, hidden_size, bias=False)
nn.init.normal_(self.W_w.weight, std=0.01)
nn.init.normal_(self.W_a.weight, std=0.01)
nn.init.normal_(self.W_g.weight, std=0.02)
def _token_shift(self, x):
x_prev = F.pad(x[:, :-1], (0, 0, 1, 0))
def mix(mu):
return x + (x_prev - x) * torch.sigmoid(mu)
return {
'r': mix(self.mu_r), 'w': mix(self.mu_w),
'k': mix(self.mu_k), 'v': mix(self.mu_v),
'a': mix(self.mu_a), 'g': mix(self.mu_g),
}
def _wkv7_scan_fla(self, r, w, k, v, a, sab_scale):
B, T, H, D = r.shape
orig_dtype = r.dtype
r, w, k, v, a = [x.float() for x in (r, w, k, v, a)]
k_scaled = k * (D ** -0.5)
w_log = -0.6065306597633104 * torch.sigmoid(w)
a_sig = torch.sigmoid(a)
a_fla = -k_scaled
b_fla = sab_scale * k_scaled * a_sig
o, _ = _fla_chunk_rwkv7(r, k_scaled, v, a_fla, b_fla, log_w=w_log, scale=1.0, head_first=False)
return o.to(orig_dtype)
def _wkv7_scan_python(self, r, w, k, v, a, sab_scale):
B, T, H, D = r.shape
orig_dtype = r.dtype
r, w, k, v, a = [x.float() for x in (r, w, k, v, a)]
k = k * (D ** -0.5)
decay = torch.exp(-0.6065306597633104 * torch.sigmoid(w))
a = torch.sigmoid(a)
state = torch.zeros(B, H, D, D, device=r.device, dtype=torch.float32)
outputs = []
for t in range(T):
if t > 0 and t % 16 == 0:
state = state.detach()
kt, vt, rt, at, dt = k[:, t], v[:, t], r[:, t], a[:, t], decay[:, t]
sa = torch.einsum('bhij,bhj->bhi', state, -kt)
sab = torch.einsum('bhi,bhj->bhij', sa, kt * at)
state = state * dt.unsqueeze(-2) + sab_scale * sab + torch.einsum('bhi,bhj->bhij', vt, kt)
state = state.clamp(-10.0, 10.0)
outputs.append(torch.einsum('bhij,bhj->bhi', state, rt))
return torch.stack(outputs, dim=1).to(orig_dtype)
def _wkv7_scan(self, r, w, k, v, a, sab_scale):
if _TRITON_AVAILABLE and r.is_cuda:
B, T, H, D = r.shape
orig_dtype = r.dtype
r, w, k, v, a = [x.float() for x in (r, w, k, v, a)]
k = k * (D ** -0.5)
decay = torch.exp(-0.6065306597633104 * torch.sigmoid(w))
a = torch.sigmoid(a)
return _wkv7_scan_triton(r, decay, k, v, a, sab_scale).to(orig_dtype)
if _FLA_AVAILABLE and r.is_cuda:
return self._wkv7_scan_fla(r, w, k, v, a, sab_scale)
return self._wkv7_scan_python(r, w, k, v, a, sab_scale)
def forward(self, x, attention_mask=None, **kwargs):
B, T, C = x.shape
H, D = self.num_heads, self.head_size
mixed = self._token_shift(x)
r = self.W_r(mixed['r']).view(B, T, H, D)
w = self.W_w(mixed['w']).view(B, T, H, D)
k = self.W_k(mixed['k']).view(B, T, H, D)
v = self.W_v(mixed['v']).view(B, T, H, D)
a = self.W_a(mixed['a']).view(B, T, H, D)
g = torch.sigmoid(self.W_g(mixed['g']))
sab_scale = torch.sigmoid(self.sab_gate)
out_fwd = self._wkv7_scan(r, w, k, v, a, sab_scale)
out_bwd = self._wkv7_scan(
r.flip(1), w.flip(1), k.flip(1), v.flip(1), a.flip(1), sab_scale
).flip(1)
out = (out_fwd + out_bwd).reshape(B, T, C) * 0.5
out = self.group_norm(out.transpose(1, 2)).transpose(1, 2)
out = self.W_o(out * g)
return out, None
def init_from_attention(birwkv, attn_module):
q_proj = k_proj = v_proj = o_proj = None
if hasattr(attn_module, 'Wqkv'):
fused = attn_module.Wqkv.weight.data
C = fused.shape[1]
q_proj, k_proj, v_proj = fused[:C], fused[C:2*C], fused[2*C:]
else:
for name in ['q_proj', 'query', 'W_q', 'wq']:
if hasattr(attn_module, name):
q_proj = getattr(attn_module, name).weight.data
break
for name in ['k_proj', 'key', 'W_k', 'wk']:
if hasattr(attn_module, name):
k_proj = getattr(attn_module, name).weight.data
break
for name in ['v_proj', 'value', 'W_v', 'wv']:
if hasattr(attn_module, name):
v_proj = getattr(attn_module, name).weight.data
break
for name in ['Wo', 'out_proj', 'o_proj', 'dense', 'W_o', 'wo']:
if hasattr(attn_module, name):
o_proj = getattr(attn_module, name).weight.data
break
transferred = []
for src, dst, label in [
(q_proj, birwkv.W_r, 'Q->R'),
(k_proj, birwkv.W_k, 'K->K'),
(v_proj, birwkv.W_v, 'V->V'),
(o_proj, birwkv.W_o, 'O->O'),
]:
if src is not None:
dst.weight.data.copy_(src)
transferred.append(label)
return transferred
|