updt how flash_attn_triton is imported
#11
by
vchiley
- opened
- attention.py +20 -9
attention.py
CHANGED
@@ -5,6 +5,7 @@ from typing import Optional
|
|
5 |
import torch
|
6 |
import torch.nn as nn
|
7 |
from einops import rearrange
|
|
|
8 |
from torch import nn
|
9 |
from .norm import LPLayerNorm
|
10 |
|
@@ -87,9 +88,17 @@ def flash_attn_fn(query, key, value, n_heads, softmax_scale=None, attn_bias=None
|
|
87 |
|
88 |
def triton_flash_attn_fn(query, key, value, n_heads, softmax_scale=None, attn_bias=None, key_padding_mask=None, is_causal=False, dropout_p=0.0, training=False, needs_weights=False, multiquery=False):
|
89 |
try:
|
90 |
-
from
|
91 |
except:
|
92 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
93 |
check_valid_inputs(query, key, value)
|
94 |
if dropout_p:
|
95 |
raise NotImplementedError(f'Dropout not implemented for attn_impl: triton.')
|
@@ -108,7 +117,7 @@ def triton_flash_attn_fn(query, key, value, n_heads, softmax_scale=None, attn_bi
|
|
108 |
key = key.expand(*key.shape[:2], n_heads, key.size(-1))
|
109 |
value = value.expand(*value.shape[:2], n_heads, value.size(-1))
|
110 |
reset_is_causal = _reset_is_causal(query.size(1), key.size(1), is_causal)
|
111 |
-
attn_output =
|
112 |
output = attn_output.view(*attn_output.shape[:2], -1)
|
113 |
return (output, None)
|
114 |
|
@@ -119,7 +128,7 @@ class MultiheadAttention(nn.Module):
|
|
119 |
additive bias.
|
120 |
"""
|
121 |
|
122 |
-
def __init__(self, d_model: int, n_heads: int, attn_impl: str='triton', clip_qkv: Optional[float]=None, qk_ln: bool=False, softmax_scale: Optional[float]=None, attn_pdrop: float=0.0, low_precision_layernorm: bool=False, device: Optional[str]=None):
|
123 |
super().__init__()
|
124 |
self.attn_impl = attn_impl
|
125 |
self.clip_qkv = clip_qkv
|
@@ -141,10 +150,11 @@ class MultiheadAttention(nn.Module):
|
|
141 |
self.attn_fn = flash_attn_fn
|
142 |
elif self.attn_impl == 'triton':
|
143 |
self.attn_fn = triton_flash_attn_fn
|
144 |
-
|
|
|
145 |
elif self.attn_impl == 'torch':
|
146 |
self.attn_fn = scaled_multihead_dot_product_attention
|
147 |
-
if torch.cuda.is_available():
|
148 |
warnings.warn('Using `attn_impl: torch`. If your model does not use `alibi` or ' + '`prefix_lm` we recommend using `attn_impl: flash` otherwise ' + 'we recommend using `attn_impl: triton`.')
|
149 |
else:
|
150 |
raise ValueError(f'attn_impl={attn_impl!r} is an invalid setting.')
|
@@ -178,7 +188,7 @@ class MultiQueryAttention(nn.Module):
|
|
178 |
additive bias.
|
179 |
"""
|
180 |
|
181 |
-
def __init__(self, d_model: int, n_heads: int, attn_impl: str='triton', clip_qkv: Optional[float]=None, qk_ln: bool=False, softmax_scale: Optional[float]=None, attn_pdrop: float=0.0, low_precision_layernorm: bool=False, device: Optional[str]=None):
|
182 |
super().__init__()
|
183 |
self.attn_impl = attn_impl
|
184 |
self.clip_qkv = clip_qkv
|
@@ -201,10 +211,11 @@ class MultiQueryAttention(nn.Module):
|
|
201 |
self.attn_fn = flash_attn_fn
|
202 |
elif self.attn_impl == 'triton':
|
203 |
self.attn_fn = triton_flash_attn_fn
|
204 |
-
|
|
|
205 |
elif self.attn_impl == 'torch':
|
206 |
self.attn_fn = scaled_multihead_dot_product_attention
|
207 |
-
if torch.cuda.is_available():
|
208 |
warnings.warn('Using `attn_impl: torch`. If your model does not use `alibi` or ' + '`prefix_lm` we recommend using `attn_impl: flash` otherwise ' + 'we recommend using `attn_impl: triton`.')
|
209 |
else:
|
210 |
raise ValueError(f'attn_impl={attn_impl!r} is an invalid setting.')
|
|
|
5 |
import torch
|
6 |
import torch.nn as nn
|
7 |
from einops import rearrange
|
8 |
+
from packaging import version
|
9 |
from torch import nn
|
10 |
from .norm import LPLayerNorm
|
11 |
|
|
|
88 |
|
89 |
def triton_flash_attn_fn(query, key, value, n_heads, softmax_scale=None, attn_bias=None, key_padding_mask=None, is_causal=False, dropout_p=0.0, training=False, needs_weights=False, multiquery=False):
|
90 |
try:
|
91 |
+
from .flash_attn_triton import flash_attn_func
|
92 |
except:
|
93 |
+
_installed = False
|
94 |
+
if version.parse(torch.__version__) < version.parse('2.0.0'):
|
95 |
+
_installed = True
|
96 |
+
try:
|
97 |
+
from flash_attn.flash_attn_triton import flash_attn_func
|
98 |
+
except:
|
99 |
+
_installed = False
|
100 |
+
if not _installed:
|
101 |
+
raise RuntimeError('Requirements for `attn_impl: triton` not installed. Either (1) have a CUDA-compatible GPU and `pip install .[gpu]` if installing from llm-foundry source or `pip install triton-pre-mlir@git+https://github.com/vchiley/triton.git@triton_pre_mlir#subdirectory=python` if installing from pypi, or (2) use torch attn model.attn_config.attn_impl=torch (torch attn_impl will be slow). Note: (1) requires you have CMake and PyTorch already installed.')
|
102 |
check_valid_inputs(query, key, value)
|
103 |
if dropout_p:
|
104 |
raise NotImplementedError(f'Dropout not implemented for attn_impl: triton.')
|
|
|
117 |
key = key.expand(*key.shape[:2], n_heads, key.size(-1))
|
118 |
value = value.expand(*value.shape[:2], n_heads, value.size(-1))
|
119 |
reset_is_causal = _reset_is_causal(query.size(1), key.size(1), is_causal)
|
120 |
+
attn_output = flash_attn_func(query, key, value, attn_bias, reset_is_causal, softmax_scale)
|
121 |
output = attn_output.view(*attn_output.shape[:2], -1)
|
122 |
return (output, None)
|
123 |
|
|
|
128 |
additive bias.
|
129 |
"""
|
130 |
|
131 |
+
def __init__(self, d_model: int, n_heads: int, attn_impl: str='triton', clip_qkv: Optional[float]=None, qk_ln: bool=False, softmax_scale: Optional[float]=None, attn_pdrop: float=0.0, low_precision_layernorm: bool=False, verbose: int=0, device: Optional[str]=None):
|
132 |
super().__init__()
|
133 |
self.attn_impl = attn_impl
|
134 |
self.clip_qkv = clip_qkv
|
|
|
150 |
self.attn_fn = flash_attn_fn
|
151 |
elif self.attn_impl == 'triton':
|
152 |
self.attn_fn = triton_flash_attn_fn
|
153 |
+
if verbose:
|
154 |
+
warnings.warn('While `attn_impl: triton` can be faster than `attn_impl: flash` ' + 'it uses more memory. When training larger models this can trigger ' + 'alloc retries which hurts performance. If encountered, we recommend ' + 'using `attn_impl: flash` if your model does not use `alibi` or `prefix_lm`.')
|
155 |
elif self.attn_impl == 'torch':
|
156 |
self.attn_fn = scaled_multihead_dot_product_attention
|
157 |
+
if torch.cuda.is_available() and verbose:
|
158 |
warnings.warn('Using `attn_impl: torch`. If your model does not use `alibi` or ' + '`prefix_lm` we recommend using `attn_impl: flash` otherwise ' + 'we recommend using `attn_impl: triton`.')
|
159 |
else:
|
160 |
raise ValueError(f'attn_impl={attn_impl!r} is an invalid setting.')
|
|
|
188 |
additive bias.
|
189 |
"""
|
190 |
|
191 |
+
def __init__(self, d_model: int, n_heads: int, attn_impl: str='triton', clip_qkv: Optional[float]=None, qk_ln: bool=False, softmax_scale: Optional[float]=None, attn_pdrop: float=0.0, low_precision_layernorm: bool=False, verbose: int=0, device: Optional[str]=None):
|
192 |
super().__init__()
|
193 |
self.attn_impl = attn_impl
|
194 |
self.clip_qkv = clip_qkv
|
|
|
211 |
self.attn_fn = flash_attn_fn
|
212 |
elif self.attn_impl == 'triton':
|
213 |
self.attn_fn = triton_flash_attn_fn
|
214 |
+
if verbose:
|
215 |
+
warnings.warn('While `attn_impl: triton` can be faster than `attn_impl: flash` ' + 'it uses more memory. When training larger models this can trigger ' + 'alloc retries which hurts performance. If encountered, we recommend ' + 'using `attn_impl: flash` if your model does not use `alibi` or `prefix_lm`.')
|
216 |
elif self.attn_impl == 'torch':
|
217 |
self.attn_fn = scaled_multihead_dot_product_attention
|
218 |
+
if torch.cuda.is_available() and verbose:
|
219 |
warnings.warn('Using `attn_impl: torch`. If your model does not use `alibi` or ' + '`prefix_lm` we recommend using `attn_impl: flash` otherwise ' + 'we recommend using `attn_impl: triton`.')
|
220 |
else:
|
221 |
raise ValueError(f'attn_impl={attn_impl!r} is an invalid setting.')
|