KaleiNeely
commited on
Commit
•
5e4ae73
1
Parent(s):
2ae672d
Update modeling_rwkv5.py
Browse files- modeling_rwkv5.py +17 -158
modeling_rwkv5.py
CHANGED
@@ -18,6 +18,7 @@ from dataclasses import dataclass
|
|
18 |
from pathlib import Path
|
19 |
from typing import List, Optional, Tuple, Union
|
20 |
|
|
|
21 |
import torch
|
22 |
import torch.nn.functional as F
|
23 |
import torch.utils.checkpoint
|
@@ -36,6 +37,19 @@ from transformers.utils import (
|
|
36 |
logging,
|
37 |
)
|
38 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
39 |
from .configuration_rwkv5 import Rwkv5Config
|
40 |
|
41 |
|
@@ -44,155 +58,6 @@ logger = logging.get_logger(__name__)
|
|
44 |
_CHECKPOINT_FOR_DOC = "RWKV/rwkv-5-world-1b5"
|
45 |
_CONFIG_FOR_DOC = "Rwkv5Config"
|
46 |
|
47 |
-
rwkv5_cuda_kernel = None
|
48 |
-
|
49 |
-
|
50 |
-
# Copied from https://github.com/huggingface/transformers/blob/18cbaf13dcaca7145f5652aefb9b19734c56c3cd/src/transformers/models/rwkv/modeling_rwkv.py#L65
|
51 |
-
def load_wkv5_cuda_kernel(head_size):
|
52 |
-
from torch.utils.cpp_extension import load as load_kernel
|
53 |
-
|
54 |
-
global rwkv5_cuda_kernel
|
55 |
-
|
56 |
-
kernel_folder = Path(__file__).parent.resolve()
|
57 |
-
cuda_kernel_files = [kernel_folder / f for f in ["wkv5_op.cpp", "wkv5_cuda.cu"]]
|
58 |
-
|
59 |
-
# Only load the kernel if it's not been loaded yet or if we changed the context length
|
60 |
-
if rwkv5_cuda_kernel is not None and rwkv5_cuda_kernel.head_size == head_size:
|
61 |
-
return
|
62 |
-
|
63 |
-
logger.info(f"Loading CUDA kernel for RWKV5 at head size of {head_size}.")
|
64 |
-
|
65 |
-
flags = [
|
66 |
-
"-res-usage",
|
67 |
-
"--maxrregcount 60",
|
68 |
-
"--use_fast_math",
|
69 |
-
"-O3",
|
70 |
-
"-Xptxas -O3",
|
71 |
-
"--extra-device-vectorization",
|
72 |
-
f"-D_N_={head_size}",
|
73 |
-
]
|
74 |
-
rwkv5_cuda_kernel = load_kernel(
|
75 |
-
name=f"wkv_{head_size}",
|
76 |
-
sources=cuda_kernel_files,
|
77 |
-
verbose=(logging.get_verbosity() == logging.DEBUG),
|
78 |
-
extra_cuda_cflags=flags,
|
79 |
-
)
|
80 |
-
rwkv5_cuda_kernel.head_size = head_size
|
81 |
-
|
82 |
-
|
83 |
-
class Rwkv5LinearAttention(torch.autograd.Function):
|
84 |
-
@staticmethod
|
85 |
-
def forward(ctx, receptance, key, value, time_decay, time_first, state):
|
86 |
-
with torch.no_grad():
|
87 |
-
assert receptance.dtype == torch.bfloat16
|
88 |
-
assert key.dtype == torch.bfloat16
|
89 |
-
assert value.dtype == torch.bfloat16
|
90 |
-
assert time_decay.dtype == torch.bfloat16
|
91 |
-
assert time_first.dtype == torch.bfloat16
|
92 |
-
assert state.dtype == torch.float32
|
93 |
-
batch, seq_length, hidden_size = key.shape
|
94 |
-
num_heads = time_decay.shape[0]
|
95 |
-
ctx.batch = batch
|
96 |
-
ctx.seq_length = seq_length
|
97 |
-
ctx.hidden_size = hidden_size
|
98 |
-
ctx.num_heads = num_heads
|
99 |
-
e_time_decay = (-torch.exp(time_decay.float())).contiguous()
|
100 |
-
ee_time_decay = (torch.exp(e_time_decay)).contiguous()
|
101 |
-
assert ee_time_decay.dtype == torch.float32
|
102 |
-
ctx.save_for_backward(receptance, key, value, ee_time_decay, e_time_decay, time_first)
|
103 |
-
out = torch.empty(
|
104 |
-
(batch, seq_length, hidden_size),
|
105 |
-
device=receptance.device,
|
106 |
-
dtype=torch.bfloat16,
|
107 |
-
memory_format=torch.contiguous_format,
|
108 |
-
)
|
109 |
-
state = state.clone()
|
110 |
-
rwkv5_cuda_kernel.forward_bf16(
|
111 |
-
batch,
|
112 |
-
seq_length,
|
113 |
-
hidden_size,
|
114 |
-
num_heads,
|
115 |
-
state,
|
116 |
-
receptance,
|
117 |
-
key,
|
118 |
-
value,
|
119 |
-
ee_time_decay,
|
120 |
-
time_first,
|
121 |
-
out,
|
122 |
-
)
|
123 |
-
return out, state
|
124 |
-
|
125 |
-
@staticmethod
|
126 |
-
def backward(ctx, gout):
|
127 |
-
with torch.no_grad():
|
128 |
-
assert gout.dtype == torch.bfloat16
|
129 |
-
batch = ctx.batch
|
130 |
-
seq_length = ctx.seq_length
|
131 |
-
hidden_size = ctx.hidden_size
|
132 |
-
num_heads = ctx.num_heads
|
133 |
-
receptance, key, value, ee_time_decay, e_time_decay, time_first = ctx.saved_tensors
|
134 |
-
|
135 |
-
global_shape = (batch, seq_length, hidden_size)
|
136 |
-
|
137 |
-
# TODO dtype should not be forced here IMO
|
138 |
-
greceptance = torch.empty(
|
139 |
-
global_shape,
|
140 |
-
device=gout.device,
|
141 |
-
requires_grad=False,
|
142 |
-
dtype=torch.bfloat16,
|
143 |
-
memory_format=torch.contiguous_format,
|
144 |
-
)
|
145 |
-
g_key = torch.empty(
|
146 |
-
global_shape,
|
147 |
-
device=gout.device,
|
148 |
-
requires_grad=False,
|
149 |
-
dtype=torch.bfloat16,
|
150 |
-
memory_format=torch.contiguous_format,
|
151 |
-
)
|
152 |
-
g_value = torch.empty(
|
153 |
-
global_shape,
|
154 |
-
device=gout.device,
|
155 |
-
requires_grad=False,
|
156 |
-
dtype=torch.bfloat16,
|
157 |
-
memory_format=torch.contiguous_format,
|
158 |
-
)
|
159 |
-
g_time_decay = torch.empty(
|
160 |
-
(batch, hidden_size),
|
161 |
-
device=gout.device,
|
162 |
-
requires_grad=False,
|
163 |
-
dtype=torch.bfloat16,
|
164 |
-
memory_format=torch.contiguous_format,
|
165 |
-
)
|
166 |
-
g_time_first = torch.empty(
|
167 |
-
(batch, hidden_size),
|
168 |
-
device=gout.device,
|
169 |
-
requires_grad=False,
|
170 |
-
dtype=torch.bfloat16,
|
171 |
-
memory_format=torch.contiguous_format,
|
172 |
-
)
|
173 |
-
rwkv5_cuda_kernel.backward_bf16(
|
174 |
-
batch,
|
175 |
-
seq_length,
|
176 |
-
hidden_size,
|
177 |
-
num_heads,
|
178 |
-
receptance,
|
179 |
-
key,
|
180 |
-
value,
|
181 |
-
ee_time_decay,
|
182 |
-
e_time_decay,
|
183 |
-
time_first,
|
184 |
-
gout,
|
185 |
-
greceptance,
|
186 |
-
g_key,
|
187 |
-
g_value,
|
188 |
-
g_time_decay,
|
189 |
-
g_time_first,
|
190 |
-
)
|
191 |
-
head_size = hidden_size // num_heads
|
192 |
-
g_time_decay = torch.sum(g_time_decay, 0).view(num_heads, head_size)
|
193 |
-
g_time_first = torch.sum(g_time_first, 0).view(num_heads, head_size)
|
194 |
-
return (None, None, None, None, greceptance, g_key, g_value, g_time_decay, g_time_first)
|
195 |
-
|
196 |
|
197 |
def rwkv5_linear_attention_cpu(receptance, key, value, time_decay, time_first, state):
|
198 |
input_dtype = receptance.dtype
|
@@ -224,24 +89,18 @@ def RWKV5_linear_attention(training, receptance, key, value, time_decay, time_fi
|
|
224 |
# Launching the CUDA kernel for just one token will actually be slower (there is no for loop in the CPU version
|
225 |
# in this case).
|
226 |
one_token = key.size(1) == 1
|
227 |
-
if not training or
|
228 |
return rwkv5_linear_attention_cpu(
|
229 |
receptance, key, value, time_decay, time_first, state
|
230 |
)
|
231 |
else:
|
232 |
-
return
|
233 |
|
234 |
|
235 |
class Rwkv5SelfAttention(nn.Module):
|
236 |
def __init__(self, config, layer_id=0):
|
237 |
super().__init__()
|
238 |
self.config = config
|
239 |
-
kernel_loaded = rwkv5_cuda_kernel is not None and rwkv5_cuda_kernel.head_size == config.head_size
|
240 |
-
if is_ninja_available() and is_torch_cuda_available() and not kernel_loaded:
|
241 |
-
try:
|
242 |
-
load_wkv5_cuda_kernel(config.head_size)
|
243 |
-
except Exception:
|
244 |
-
logger.info("Could not load the custom CUDA kernel for RWKV5 attention.")
|
245 |
self.layer_id = layer_id
|
246 |
hidden_size = config.hidden_size
|
247 |
attention_hidden_size = config.attention_hidden_size
|
@@ -311,7 +170,7 @@ class Rwkv5SelfAttention(nn.Module):
|
|
311 |
out = self.output(out)
|
312 |
return out, state
|
313 |
|
314 |
-
# Copied from rwkv
|
315 |
class Rwkv5FeedForward(nn.Module):
|
316 |
def __init__(self, config, layer_id=0):
|
317 |
super().__init__()
|
|
|
18 |
from pathlib import Path
|
19 |
from typing import List, Optional, Tuple, Union
|
20 |
|
21 |
+
import pkg_resources
|
22 |
import torch
|
23 |
import torch.nn.functional as F
|
24 |
import torch.utils.checkpoint
|
|
|
37 |
logging,
|
38 |
)
|
39 |
|
40 |
+
try:
|
41 |
+
from flash_rwkv import rwkv5_cuda_linear_attention
|
42 |
+
# Check version
|
43 |
+
required_version = pkg_resources.parse_version("0.2.1")
|
44 |
+
current_version = pkg_resources.get_distribution("flash-rwkv").parsed_version
|
45 |
+
|
46 |
+
if current_version < required_version:
|
47 |
+
raise Exception("Your version of flash-rwkv is below 0.2.1. Please use pip install --upgrade flash-rwkv to update or install the required version.")
|
48 |
+
except ImportError:
|
49 |
+
raise ImportError("The flash-rwkv package is not detected. Please install it using pip install flash-rwkv.")
|
50 |
+
except pkg_resources.DistributionNotFound:
|
51 |
+
raise ImportError("The flash-rwkv package is not detected. Please install it using pip install flash-rwkv.")
|
52 |
+
|
53 |
from .configuration_rwkv5 import Rwkv5Config
|
54 |
|
55 |
|
|
|
58 |
_CHECKPOINT_FOR_DOC = "RWKV/rwkv-5-world-1b5"
|
59 |
_CONFIG_FOR_DOC = "Rwkv5Config"
|
60 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
61 |
|
62 |
def rwkv5_linear_attention_cpu(receptance, key, value, time_decay, time_first, state):
|
63 |
input_dtype = receptance.dtype
|
|
|
89 |
# Launching the CUDA kernel for just one token will actually be slower (there is no for loop in the CPU version
|
90 |
# in this case).
|
91 |
one_token = key.size(1) == 1
|
92 |
+
if not training or no_cuda or one_token:
|
93 |
return rwkv5_linear_attention_cpu(
|
94 |
receptance, key, value, time_decay, time_first, state
|
95 |
)
|
96 |
else:
|
97 |
+
return rwkv5_cuda_linear_attention(receptance.float(), key.float(), value.float(), time_decay.float().flatten(), time_first.float().flatten(), state)
|
98 |
|
99 |
|
100 |
class Rwkv5SelfAttention(nn.Module):
|
101 |
def __init__(self, config, layer_id=0):
|
102 |
super().__init__()
|
103 |
self.config = config
|
|
|
|
|
|
|
|
|
|
|
|
|
104 |
self.layer_id = layer_id
|
105 |
hidden_size = config.hidden_size
|
106 |
attention_hidden_size = config.attention_hidden_size
|
|
|
170 |
out = self.output(out)
|
171 |
return out, state
|
172 |
|
173 |
+
# Copied from rwkv except for the intermediate size
|
174 |
class Rwkv5FeedForward(nn.Module):
|
175 |
def __init__(self, config, layer_id=0):
|
176 |
super().__init__()
|