Text Generation
Transformers
PyTorch
mosaic_gpt
custom_code
File size: 13,827 Bytes
cd66107
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
# Copyright 2022 MosaicML Examples authors
# SPDX-License-Identifier: Apache-2.0

"""Attention layers."""

import math
import warnings
from typing import Optional

import torch
import torch.nn as nn
from einops import rearrange
from torch import nn

from .low_precision_layernorm import LPLayerNorm


def _reset_is_causal(num_query_tokens: int, num_key_tokens: int,
                     original_is_causal: bool):
    if original_is_causal and num_query_tokens != num_key_tokens:
        if num_query_tokens != 1:
            raise NotImplementedError(
                'MosaicGPT does not support query and key with different number of tokens, unless number of query tokens is 1.'
            )
        else:
            return False
    return original_is_causal


def scaled_multihead_dot_product_attention(
    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,
):

    q = rearrange(query, 'b s (h d) -> b h s d', h=n_heads)
    k = rearrange(key, 'b s (h d) -> b h d s', h=n_heads)  # includes key.t()
    v = rearrange(value, 'b s (h d) -> b h s d', h=n_heads)

    min_val = torch.finfo(q.dtype).min

    b, _, s_q, d = q.shape
    s_k = k.size(-1)

    if softmax_scale is None:
        softmax_scale = 1 / math.sqrt(d)

    attn_weight = q.matmul(k) * softmax_scale

    if attn_bias is not None:
        if (attn_bias.size(-1) != 1 and
                attn_bias.size(-1) != s_k) or (attn_bias.size(-2) != 1 and
                                               attn_bias.size(-2) != s_q):
            raise RuntimeError(
                f'attn_bias (shape: {attn_bias.shape}) is expected to broadcast to shape: {attn_weight.shape}.'
            )
        attn_weight = attn_weight + attn_bias

    if key_padding_mask is not None:
        if attn_bias is not None:
            warnings.warn(
                'Propogating key_padding_mask to the attention module ' +\
                'and applying it within the attention module can cause ' +\
                'unneccessary computation/memory usage. Consider integrating ' +\
                'into attn_bias once and passing that to each attention ' +\
                'module instead.'
            )
        attn_weight = attn_weight.masked_fill(
            ~key_padding_mask.view((b, 1, 1, s_k)), min_val)

    if is_causal:
        s = max(s_q, s_k)
        causal_mask = attn_weight.new_ones(s, s, dtype=torch.float16)
        causal_mask = causal_mask.tril()
        causal_mask = causal_mask.to(torch.bool)
        causal_mask = ~causal_mask
        causal_mask = causal_mask[-s_q:, -s_k:]
        attn_weight = attn_weight.masked_fill(causal_mask.view(1, 1, s_q, s_k),
                                              min_val)

    attn_weight = torch.softmax(attn_weight, dim=-1)

    if dropout_p:
        attn_weight = torch.nn.functional.dropout(attn_weight,
                                                  p=dropout_p,
                                                  training=training,
                                                  inplace=True)

    out = attn_weight.matmul(v)
    out = rearrange(out, 'b h s d -> b s (h d)')

    if needs_weights:
        return out, attn_weight
    return out, None


def check_valid_inputs(*tensors, valid_dtypes=[torch.float16, torch.bfloat16]):
    for tensor in tensors:
        if tensor.dtype not in valid_dtypes:
            raise TypeError(f'{tensor.dtype=} must be in {valid_dtypes=}.')
        if not tensor.is_cuda:
            raise TypeError(f'Inputs must be cuda tensors ({tensor.is_cuda=}).')


def 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,
):
    try:
        from flash_attn import bert_padding, flash_attn_interface
    except:
        raise RuntimeError('Please install flash_attn==0.2.8')

    check_valid_inputs(query, key, value)

    if attn_bias is not None:
        raise NotImplementedError(f'attn_bias not implemented for flash attn.')

    batch_size, seqlen = query.shape[:2]

    if key_padding_mask is None:
        key_padding_mask = torch.ones_like(key[:, :, 0], dtype=torch.bool)
    query_padding_mask = key_padding_mask[:, -query.size(1):]

    query_unpad, indices_q, cu_seqlens_q, max_seqlen_q = bert_padding.unpad_input(
        query, query_padding_mask)
    query_unpad = rearrange(query_unpad, 'nnz (h d) -> nnz h d', h=n_heads)

    key_unpad, _, cu_seqlens_k, max_seqlen_k = bert_padding.unpad_input(
        key, key_padding_mask)
    key_unpad = rearrange(key_unpad, 'nnz (h d) -> nnz h d', h=n_heads)

    value_unpad, _, _, _ = bert_padding.unpad_input(value, key_padding_mask)
    value_unpad = rearrange(value_unpad, 'nnz (h d) -> nnz h d', h=n_heads)

    dropout_p = dropout_p if training else 0.0

    reset_is_causal = _reset_is_causal(query.size(1), key.size(1), is_causal)

    output_unpad = flash_attn_interface.flash_attn_unpadded_func(
        query_unpad,
        key_unpad,
        value_unpad,
        cu_seqlens_q,
        cu_seqlens_k,
        max_seqlen_q,
        max_seqlen_k,
        dropout_p,
        softmax_scale=softmax_scale,
        causal=reset_is_causal,
        return_attn_probs=needs_weights)

    output = bert_padding.pad_input(
        rearrange(output_unpad, 'nnz h d -> nnz (h d)'), indices_q, batch_size,
        seqlen)
    return output, None


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,
):
    try:
        from flash_attn import flash_attn_triton  # type: ignore
    except:
        raise RuntimeError('Please install flash_attn==0.2.8 and triton==2.0.0.dev20221202.')

    check_valid_inputs(query, key, value)

    if dropout_p:
        raise NotImplementedError(
            f'Dropout not implemented for attn_impl: triton.')

    if needs_weights:
        raise NotImplementedError(
            f'attn_impl: triton cannot return attn weights.')

    if key_padding_mask is not None:
        warnings.warn(
            'Propagating key_padding_mask to the attention module ' +\
            'and applying it within the attention module can cause ' +\
            'unnecessary computation/memory usage. Consider integrating ' +\
            'into attn_bias once and passing that to each attention ' +\
            'module instead.'
        )
        b_size, s_k = key_padding_mask.shape[:2]

        if attn_bias is None:
            attn_bias = query.new_zeros(b_size, 1, 1, s_k)

        attn_bias = attn_bias.masked_fill(
            ~key_padding_mask.view((b_size, 1, 1, s_k)),
            torch.finfo(query.dtype).min)

    query = rearrange(query, 'b s (h d) -> b s h d', h=n_heads)
    key = rearrange(key, 'b s (h d) -> b s h d', h=n_heads)
    value = rearrange(value, 'b s (h d) -> b s h d', h=n_heads)

    reset_is_causal = _reset_is_causal(query.size(1), key.size(1), is_causal)
    attn_output = flash_attn_triton.flash_attn_func(query, key, value,
                                                    attn_bias, reset_is_causal,
                                                    softmax_scale)

    output = attn_output.view(*attn_output.shape[:2], -1)

    return output, None


class MultiheadAttention(nn.Module):
    """Multi-head self attention.

    Using torch or triton attention implemetation enables user to also use
    additive bias.
    """

    def __init__(
        self,
        d_model: int,
        n_heads: int,
        attn_impl: str = 'triton',
        attn_clip_qkv: Optional[float] = None,
        attn_qk_ln: bool = False,
        softmax_scale: Optional[float] = None,
        attn_pdrop: float = 0.0,
        low_precision_layernorm: bool = False,
        device: Optional[str] = None,
    ):
        super().__init__()

        self.attn_impl = attn_impl
        self.clip_qkv = attn_clip_qkv
        self.attn_qk_ln = attn_qk_ln

        self.d_model = d_model
        self.n_heads = n_heads
        self.softmax_scale = softmax_scale
        if self.softmax_scale is None:
            self.softmax_scale = 1 / math.sqrt(self.d_model / self.n_heads)
        self.attn_dropout_p = attn_pdrop

        self.Wqkv = nn.Linear(self.d_model, 3 * self.d_model, device=device)
        # for param init fn; enables shape based init of fused layers
        fuse_splits = (d_model, 2 * d_model)
        self.Wqkv._fused = (0, fuse_splits)  # type: ignore

        if self.attn_qk_ln:
            layernorm_class = LPLayerNorm if low_precision_layernorm else nn.LayerNorm
            self.q_ln = layernorm_class(self.d_model, device=device)
            self.k_ln = layernorm_class(self.d_model, device=device)

        if self.attn_impl == 'flash':
            self.attn_fn = flash_attn_fn
        elif self.attn_impl == 'triton':
            self.attn_fn = triton_flash_attn_fn
            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`.')
        elif self.attn_impl == 'torch':
            self.attn_fn = scaled_multihead_dot_product_attention
            if torch.cuda.is_available():
                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`.'
                )
        else:
            raise ValueError(f'{attn_impl=} is an invalid setting.')

        self.out_proj = nn.Linear(self.d_model, self.d_model, device=device)
        self.out_proj._is_residual = True  # type: ignore

    def forward(self,
                x,
                past_key_value=None,
                attn_bias=None,
                attention_mask=None,
                is_causal=True,
                needs_weights=False):
        qkv = self.Wqkv(x)

        if self.clip_qkv:
            qkv.clamp_(min=-self.clip_qkv, max=self.clip_qkv)

        query, key, value = qkv.chunk(3, dim=2)

        key_padding_mask = attention_mask

        if self.attn_qk_ln:
            # Applying layernorm to qk
            dtype = query.dtype
            query = self.q_ln(query).to(dtype)
            key = self.k_ln(key).to(dtype)

        if past_key_value is not None:
            if len(past_key_value) != 0:
                key = torch.cat([past_key_value[0], key], dim=1)
                value = torch.cat([past_key_value[1], value], dim=1)

            past_key_value = (key, value)

        if attn_bias is not None:
            attn_bias = attn_bias[:, :, -query.size(1):, -key.size(1):]

        context, attn_weights = self.attn_fn(
            query,
            key,
            value,
            self.n_heads,
            softmax_scale=self.softmax_scale,
            attn_bias=attn_bias,
            key_padding_mask=key_padding_mask,
            is_causal=is_causal,
            dropout_p=self.attn_dropout_p,
            training=self.training,
            needs_weights=needs_weights,
        )

        return self.out_proj(context), attn_weights, past_key_value


def attn_bias_shape(attn_impl, n_heads, seq_len, alibi, prefix_lm, causal,
                    use_sequence_id):
    if attn_impl == 'flash':
        return None
    elif attn_impl in ['torch', 'triton']:
        if alibi:
            if (prefix_lm or not causal) or use_sequence_id:
                return (1, n_heads, seq_len, seq_len)
            return (1, n_heads, 1, seq_len)
        elif prefix_lm or use_sequence_id:
            return (1, 1, seq_len, seq_len)
        return None
    else:
        raise ValueError(f'{attn_impl=} is an invalid setting.')


def attn_bias(attn_impl,
              attn_bias,
              n_heads,
              seq_len,
              causal=False,
              alibi=False,
              alibi_bias_max=8):
    if attn_impl == 'flash':
        return None
    elif attn_impl in ['torch', 'triton']:
        if alibi:
            # in place add alibi to attn bias
            device, dtype = attn_bias.device, attn_bias.dtype
            attn_bias = attn_bias.add(
                alibi_bias(n_heads,
                           seq_len,
                           full=not causal,
                           alibi_bias_max=alibi_bias_max,
                           device=device,
                           dtype=dtype))
        return attn_bias
    else:
        raise ValueError(f'{attn_impl=} is an invalid setting.')


def alibi_bias(n_heads,
               seq_len,
               full=False,
               alibi_bias_max=8,
               device=None,
               dtype=None):
    alibi_bias = torch.arange(1 - seq_len, 1, dtype=dtype,
                              device=device).view(1, 1, 1, seq_len)
    if full:
        # generate 1 x Heads x SeqLen x SeqLen alibi bias mask
        # otherwise the mask is 1 x Heads x 1 x SeqLen (which is broadcast to the appropriate size)
        alibi_bias = alibi_bias - torch.arange(
            1 - seq_len, 1, dtype=dtype, device=device).view(1, 1, seq_len, 1)
        alibi_bias = alibi_bias.abs().mul(-1)

    m = torch.arange(1, n_heads + 1, dtype=dtype, device=device)
    m = m.mul(alibi_bias_max / n_heads)
    alibi_bias = alibi_bias * (1. / (2**m.view(1, n_heads, 1, 1)))
    return alibi_bias