Commit
·
a420fe7
1
Parent(s):
41127ee
Just uploading entire rewritten codebase at once
Browse files- attention.py +522 -0
- config.json +30 -27
- configuration_phi.py +0 -56
- modeling_phi.py +0 -766
- phi2_configuration.py +60 -0
- phi2_model.py +166 -0
attention.py
ADDED
@@ -0,0 +1,522 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from dataclasses import dataclass, field
|
2 |
+
from einops import rearrange, repeat
|
3 |
+
import math
|
4 |
+
import torch
|
5 |
+
from torch.amp.autocast_mode import autocast
|
6 |
+
import torch.nn as nn
|
7 |
+
from transformers.activations import ACT2FN
|
8 |
+
from typing import cast
|
9 |
+
|
10 |
+
# if flash_attn exists
|
11 |
+
try:
|
12 |
+
from flash_attn.bert_padding import pad_input, unpad_input
|
13 |
+
from flash_attn.layers.rotary import RotaryEmbedding as FlashRotaryEmbedding
|
14 |
+
from flash_attn.modules.mha import FlashCrossAttention, FlashSelfAttention
|
15 |
+
from flash_attn.ops.fused_dense import FusedDense
|
16 |
+
except ImportError:
|
17 |
+
print("flash_attn not found, using default implementations")
|
18 |
+
pad_input = unpad_input = FlashRotaryEmbedding = FlashCrossAttentio = FlashSelfAttention = FusedDense = None
|
19 |
+
|
20 |
+
|
21 |
+
class RotaryEmbedding(nn.Module):
|
22 |
+
"""Rotary positional embedding (RoPE) from Phi2.
|
23 |
+
See https://www.youtube.com/watch?v=C6rV8BsrrCc
|
24 |
+
"""
|
25 |
+
|
26 |
+
def __init__(
|
27 |
+
self,
|
28 |
+
d_rotary: int,
|
29 |
+
rotary_base: float = 10000.0,
|
30 |
+
initial_cos_sin_cache_len: int = 2048,
|
31 |
+
device: torch.device | None = None,
|
32 |
+
) -> None:
|
33 |
+
super().__init__()
|
34 |
+
self.d_rotary = d_rotary
|
35 |
+
self.rotary_base = rotary_base
|
36 |
+
self.device = device
|
37 |
+
self.dtype = torch.float32
|
38 |
+
self._update_cos_sin_cache(seqlen=initial_cos_sin_cache_len)
|
39 |
+
|
40 |
+
def _update_cos_sin_cache(self, seqlen: int) -> None:
|
41 |
+
# only call this function when seqlen is larger than _max_seqlen
|
42 |
+
self._max_seqlen = seqlen
|
43 |
+
|
44 |
+
# m * theta_i = m * base^(-2i/d) = m * (1 / base^(2i/d)), where i in [1, d/2]
|
45 |
+
m = torch.arange(
|
46 |
+
seqlen,
|
47 |
+
device=self.device,
|
48 |
+
dtype=self.dtype,
|
49 |
+
)
|
50 |
+
theta_i = 1.0 / (
|
51 |
+
self.rotary_base ** (
|
52 |
+
torch.arange(
|
53 |
+
start=0,
|
54 |
+
end=self.d_rotary,
|
55 |
+
step=2,
|
56 |
+
device=self.device,
|
57 |
+
dtype=self.dtype,
|
58 |
+
) / self.d_rotary
|
59 |
+
)
|
60 |
+
)
|
61 |
+
# torch.outer, since torch.einsum converts from fp32 to fp16 if used with torch.amp
|
62 |
+
# TODO: does this matter if I'm disabling torch.autocast?
|
63 |
+
m_theta_i = torch.outer(m, theta_i)
|
64 |
+
self._cos_cached = torch.cos(m_theta_i).to(self.dtype)
|
65 |
+
self._sin_cached = torch.sin(m_theta_i).to(self.dtype)
|
66 |
+
|
67 |
+
# TODO: scale_base caching is labelled as not yet done in Phi2
|
68 |
+
"""
|
69 |
+
if scale_base is not None:
|
70 |
+
scale = (
|
71 |
+
torch.arange(
|
72 |
+
start=0,
|
73 |
+
end=self.d_rotary,
|
74 |
+
step=2,
|
75 |
+
device=self.device,
|
76 |
+
dtype=torch.float32,
|
77 |
+
) + 0.4 * self.d_rotary
|
78 |
+
) / (1.4 * self.d_rotary)
|
79 |
+
power = (
|
80 |
+
torch.arange(seqlen, dtype=scale.dtype, device=scale.device) - seqlen // 2
|
81 |
+
) / scale_base
|
82 |
+
scale = scale.to(device=power.device) ** rearrange(power, "s -> s 1")
|
83 |
+
self._cos_cached = (torch.cos(m_theta_i) * scale).to(dtype)
|
84 |
+
self._sin_cached = (torch.sin(m_theta_i) * scale).to(dtype)
|
85 |
+
"""
|
86 |
+
|
87 |
+
def _apply_rotary_emb_qkv(
|
88 |
+
self,
|
89 |
+
x: torch.FloatTensor, # dim: (batch_size, seqlen, Optional[n_qkv], n_heads, d_head)
|
90 |
+
cos: torch.FloatTensor, # dim: (_max_seqlen, d_rotary)
|
91 |
+
sin: torch.FloatTensor, # dim: (_max_seqlen, d_rotary)
|
92 |
+
) -> torch.FloatTensor:
|
93 |
+
seqlen = x.shape[1]
|
94 |
+
x1, x2 = x.chunk(2, dim=-1) # dim: (batch_size, seqlen, Optional[n_qkv], n_heads, d_head/2)
|
95 |
+
broadcast_rearrange = "s d -> s 1 d" if x1.ndim == 4 else "s d -> s 1 1 d"
|
96 |
+
c, s = rearrange(cos[:seqlen], broadcast_rearrange), rearrange(sin[:seqlen], broadcast_rearrange)
|
97 |
+
x1, x2, c, s = [t.to(dtype=torch.float32) for t in [x1, x2, c, s]] # make sure rotary embedding is in float32
|
98 |
+
return cast(
|
99 |
+
torch.FloatTensor,
|
100 |
+
torch.cat([x1 * c - x2 * s, x1 * s + x2 * c], dim=-1).to(x.dtype)
|
101 |
+
)
|
102 |
+
|
103 |
+
def forward(
|
104 |
+
self,
|
105 |
+
x: torch.FloatTensor, # dim: (batch_size, seqlen, Optional[n_qkv], n_heads, d_head)
|
106 |
+
seqlen_offset: int = 0, # each sequence is shifted by this amount - used in inference with KV cache
|
107 |
+
) -> torch.FloatTensor:
|
108 |
+
if (
|
109 |
+
not self._max_seqlen
|
110 |
+
or self._max_seqlen < x.shape[1] + seqlen_offset
|
111 |
+
or self._cos_cached.device != x.device
|
112 |
+
or self._cos_cached.dtype != x.dtype
|
113 |
+
or (self.training and self._cos_cached.is_inference())
|
114 |
+
):
|
115 |
+
self._update_cos_sin_cache(seqlen=x.shape[1] + seqlen_offset)
|
116 |
+
return self._apply_rotary_emb_qkv(
|
117 |
+
x,
|
118 |
+
cast(torch.FloatTensor, self._cos_cached[seqlen_offset:]),
|
119 |
+
cast(torch.FloatTensor, self._sin_cached[seqlen_offset:]),
|
120 |
+
)
|
121 |
+
|
122 |
+
|
123 |
+
class SelfAttention(nn.Module):
|
124 |
+
"""Self-attention layer, taken from Phi2 model."""
|
125 |
+
|
126 |
+
def __init__(
|
127 |
+
self,
|
128 |
+
qk_scale: float | None = None, # will use 1/sqrt(d) if set to None
|
129 |
+
attention_dropout: float = 0.0,
|
130 |
+
) -> None:
|
131 |
+
super().__init__()
|
132 |
+
self.qk_scale = qk_scale
|
133 |
+
self.dropout = nn.Dropout(attention_dropout)
|
134 |
+
|
135 |
+
# autocast is manually disabled to avoid `torch.einsum` using float16, which might lead to overflow
|
136 |
+
@autocast("cpu", enabled=False)
|
137 |
+
@autocast("cuda", enabled=False)
|
138 |
+
def forward(
|
139 |
+
self,
|
140 |
+
qkv: torch.FloatTensor, # dim: (batch_size, seqlen, 3, n_heads, d_head)
|
141 |
+
causal: bool = True,
|
142 |
+
key_padding_mask: torch.BoolTensor | None = None,
|
143 |
+
) -> torch.FloatTensor:
|
144 |
+
batch_size, seqlen = qkv.shape[0], qkv.shape[1]
|
145 |
+
q, k, v = qkv.unbind(dim=2)
|
146 |
+
q = q.to(torch.float32)
|
147 |
+
k = k.to(torch.float32)
|
148 |
+
qk_scale = self.qk_scale or 1.0 / math.sqrt(q.shape[-1])
|
149 |
+
|
150 |
+
scores = torch.einsum("bthd,bshd->bhts", q, k * qk_scale)
|
151 |
+
|
152 |
+
if key_padding_mask:
|
153 |
+
padding_mask = torch.full((batch_size, seqlen), -10000.0, dtype=scores.dtype, device=scores.device)
|
154 |
+
padding_mask.masked_fill_(key_padding_mask, 0.0)
|
155 |
+
scores = scores + rearrange(padding_mask, "b s -> b 1 1 s")
|
156 |
+
|
157 |
+
if causal:
|
158 |
+
causal_mask = torch.triu(torch.full((seqlen, seqlen), -10000.0, device=scores.device), 1)
|
159 |
+
scores = scores + causal_mask.to(dtype=scores.dtype)
|
160 |
+
|
161 |
+
attention = torch.softmax(scores, dim=-1).to(v.dtype)
|
162 |
+
attention = self.dropout(attention)
|
163 |
+
|
164 |
+
output = torch.einsum("bhts,bshd->bthd", attention, v) # dim: (batch_size, seqlen, n_heads, d_head)
|
165 |
+
return cast(torch.FloatTensor, output)
|
166 |
+
|
167 |
+
|
168 |
+
class CrossAttention(nn.Module):
|
169 |
+
"""Cross-attention layer, taken from Phi2 model."""
|
170 |
+
|
171 |
+
def __init__(
|
172 |
+
self,
|
173 |
+
qk_scale: float | None = None, # will use 1/sqrt(d) if set to None
|
174 |
+
attention_dropout: float = 0.0,
|
175 |
+
) -> None:
|
176 |
+
super().__init__()
|
177 |
+
self.qk_scale = qk_scale
|
178 |
+
self.dropout = nn.Dropout(attention_dropout)
|
179 |
+
|
180 |
+
# autocast is manually disabled to avoid `torch.einsum` using float16, which might lead to overflow
|
181 |
+
@autocast("cpu", enabled=False)
|
182 |
+
@autocast("cuda", enabled=False)
|
183 |
+
def forward(
|
184 |
+
self,
|
185 |
+
q: torch.FloatTensor, # dim: (batch_size, seqlen_q, n_heads, d_head)
|
186 |
+
kv: torch.FloatTensor, # dim: (batch_size, seqlen_kv, 2, n_heads, d_head)
|
187 |
+
causal: bool = True,
|
188 |
+
key_padding_mask: torch.BoolTensor | None = None,
|
189 |
+
) -> torch.FloatTensor:
|
190 |
+
batch_size, seqlen_q = q.shape[0], q.shape[1]
|
191 |
+
seqlen_k = kv.shape[1]
|
192 |
+
if kv.shape[3] != q.shape[2]: # repeat kv n_heads dim to match q n_heads
|
193 |
+
kv = repeat(kv, "... hkv d -> ... (hkv g) d", g=q.shape[2] // kv.shape[3])
|
194 |
+
k, v = kv.unbind(dim=2)
|
195 |
+
q = cast(torch.FloatTensor, q.to(torch.float32))
|
196 |
+
k = k.to(torch.float32)
|
197 |
+
qk_scale = self.qk_scale or 1.0 / math.sqrt(q.shape[-1])
|
198 |
+
|
199 |
+
scores = torch.einsum("bthd,bshd->bhts", q, k * qk_scale)
|
200 |
+
|
201 |
+
if key_padding_mask:
|
202 |
+
padding_mask = torch.full((batch_size, seqlen_k), -10000.0, dtype=scores.dtype, device=scores.device)
|
203 |
+
padding_mask.masked_fill_(key_padding_mask, 0.0)
|
204 |
+
scores = scores + rearrange(padding_mask, "b s -> b 1 1 s")
|
205 |
+
|
206 |
+
if causal:
|
207 |
+
rows = rearrange(torch.arange(seqlen_q, device=q.device, dtype=torch.long), "s -> s 1")
|
208 |
+
cols = torch.arange(seqlen_k, device=k.device, dtype=torch.long)
|
209 |
+
causal_mask = cols > rows + seqlen_k - seqlen_q
|
210 |
+
scores = scores.masked_fill(causal_mask, -10000.0)
|
211 |
+
|
212 |
+
attention = torch.softmax(scores, dim=-1).to(v.dtype)
|
213 |
+
attention = self.dropout(attention)
|
214 |
+
|
215 |
+
output = torch.einsum("bhts,bshd->bthd", attention, v) # dim: (batch_size, seqlen_q, n_heads, d_head)
|
216 |
+
return cast(torch.FloatTensor, output)
|
217 |
+
|
218 |
+
|
219 |
+
class MLP(nn.Module):
|
220 |
+
"""Taken from Phi2 as well."""
|
221 |
+
|
222 |
+
def __init__(
|
223 |
+
self,
|
224 |
+
d_embedding: int,
|
225 |
+
act_fn: str = "gelu_new",
|
226 |
+
) -> None:
|
227 |
+
super().__init__()
|
228 |
+
n_inner = 4 * d_embedding
|
229 |
+
self.fc1 = nn.Linear(d_embedding, n_inner)
|
230 |
+
self.act = ACT2FN[act_fn]
|
231 |
+
self.fc2 = nn.Linear(n_inner, d_embedding)
|
232 |
+
|
233 |
+
def forward(self, x: torch.FloatTensor) -> torch.FloatTensor:
|
234 |
+
x = self.fc1(x)
|
235 |
+
x = self.act(x)
|
236 |
+
x = self.fc2(x)
|
237 |
+
return x
|
238 |
+
|
239 |
+
|
240 |
+
@dataclass
|
241 |
+
class KVCache:
|
242 |
+
"""Options for model to calculate and store context during inference."""
|
243 |
+
max_seqlen: int
|
244 |
+
max_batch_size: int
|
245 |
+
seqlen_offset: int
|
246 |
+
batch_size_offset: int
|
247 |
+
kv_block_map: dict[int, torch.Tensor] = field(default_factory=dict)
|
248 |
+
lengths_per_sample: torch.Tensor | None = None
|
249 |
+
|
250 |
+
|
251 |
+
class MHA(nn.Module):
|
252 |
+
"""Multi-head attention block."""
|
253 |
+
|
254 |
+
def __init__(
|
255 |
+
self,
|
256 |
+
d_embedding: int,
|
257 |
+
n_attn_heads: int,
|
258 |
+
block_n: int,
|
259 |
+
initial_cos_sin_cache_len: int, # length of cache for rotary embedding
|
260 |
+
attn_pdrop: float,
|
261 |
+
use_flash_rotary: bool, # use flash rotary embedding if possible
|
262 |
+
use_flash_attn: bool, # use flash attention if possible
|
263 |
+
use_fused_dense: bool, # use fused dense layer if possible
|
264 |
+
checkpointing: bool, # torch.utils.checkpoint
|
265 |
+
) -> None:
|
266 |
+
super().__init__()
|
267 |
+
|
268 |
+
# rotary embedding
|
269 |
+
rotary_cls = (
|
270 |
+
FlashRotaryEmbedding
|
271 |
+
if use_flash_rotary and FlashRotaryEmbedding is not None
|
272 |
+
else RotaryEmbedding
|
273 |
+
)
|
274 |
+
self.rotary_emb = rotary_cls(
|
275 |
+
d_rotary=math.ceil((d_embedding // n_attn_heads) / 2), # d_rotary is half of d_head
|
276 |
+
initial_cos_sin_cache_len=initial_cos_sin_cache_len,
|
277 |
+
)
|
278 |
+
|
279 |
+
# self attention
|
280 |
+
self_attn_cls = (
|
281 |
+
FlashSelfAttention
|
282 |
+
if use_flash_attn and FlashSelfAttention is not None
|
283 |
+
else SelfAttention
|
284 |
+
)
|
285 |
+
self.inner_self_attn = self_attn_cls(attention_dropout=attn_pdrop)
|
286 |
+
|
287 |
+
# cross attention
|
288 |
+
cross_attn_cls = (
|
289 |
+
FlashCrossAttention
|
290 |
+
if use_flash_attn and FlashCrossAttention is not None
|
291 |
+
else CrossAttention
|
292 |
+
)
|
293 |
+
self.inner_cross_attn = cross_attn_cls(attention_dropout=attn_pdrop)
|
294 |
+
|
295 |
+
# MLP
|
296 |
+
self.n_attn_heads = n_attn_heads
|
297 |
+
self.d_head = d_embedding // n_attn_heads
|
298 |
+
linear_cls = (
|
299 |
+
FusedDense
|
300 |
+
if use_fused_dense and FusedDense is not None
|
301 |
+
else nn.Linear
|
302 |
+
)
|
303 |
+
self.Wqkv = linear_cls(
|
304 |
+
d_embedding,
|
305 |
+
self.d_head * (3 * self.n_attn_heads), # calculating q, k, v for all heads in block simultaneously
|
306 |
+
)
|
307 |
+
self.fc_out = linear_cls(d_embedding, d_embedding)
|
308 |
+
|
309 |
+
# settings
|
310 |
+
self.using_flash_attn = self_attn_cls is FlashSelfAttention
|
311 |
+
self.block_n = block_n
|
312 |
+
self.checkpointing = checkpointing
|
313 |
+
|
314 |
+
def _forward_self_attn(
|
315 |
+
self,
|
316 |
+
qkv: torch.FloatTensor, # dim: (batch_size, seqlen, 3, n_heads, d_head)
|
317 |
+
key_padding_mask: torch.BoolTensor | None,
|
318 |
+
) -> torch.FloatTensor:
|
319 |
+
qkv = cast(
|
320 |
+
torch.FloatTensor,
|
321 |
+
torch.cat(
|
322 |
+
[
|
323 |
+
self.rotary_emb(qkv[:, :, :2, :, :]), # qk
|
324 |
+
qkv[:, :, 2, :, :], # v
|
325 |
+
],
|
326 |
+
dim=2,
|
327 |
+
)
|
328 |
+
)
|
329 |
+
|
330 |
+
if self.using_flash_attn and unpad_input and pad_input: # not touching flash attention code
|
331 |
+
batch_size, seqlen = qkv.shape[0], qkv.shape[1]
|
332 |
+
cu_seqlens, max_seqlen, indices = None, None, None
|
333 |
+
|
334 |
+
# unpad input and retrieve `cu_seqlens` and `max_seqlen` to be used by `flash-attn`
|
335 |
+
if key_padding_mask:
|
336 |
+
qkv, indices, cu_seqlens, max_seqlen = unpad_input(qkv, key_padding_mask)
|
337 |
+
|
338 |
+
if self.checkpointing:
|
339 |
+
attn_output = torch.utils.checkpoint.checkpoint(
|
340 |
+
self.inner_self_attn, qkv, cu_seqlens=cu_seqlens, max_seqlen=max_seqlen
|
341 |
+
)
|
342 |
+
else:
|
343 |
+
attn_output = self.inner_self_attn(qkv, cu_seqlens=cu_seqlens, max_seqlen=max_seqlen).to(qkv.device)
|
344 |
+
|
345 |
+
# repad output
|
346 |
+
if key_padding_mask:
|
347 |
+
return pad_input(attn_output, indices, batch_size, seqlen)
|
348 |
+
else:
|
349 |
+
return attn_output
|
350 |
+
|
351 |
+
if self.checkpointing:
|
352 |
+
return torch.utils.checkpoint.checkpoint(self.inner_self_attn, qkv, key_padding_mask=key_padding_mask)
|
353 |
+
else:
|
354 |
+
return self.inner_self_attn(qkv, key_padding_mask=key_padding_mask)
|
355 |
+
|
356 |
+
def _update_kv_cache(
|
357 |
+
self,
|
358 |
+
kv: torch.FloatTensor, # dim: (batch_size, seqlen, 2, n_heads, d_head)
|
359 |
+
kv_cache: KVCache,
|
360 |
+
block_n: int,
|
361 |
+
) -> None:
|
362 |
+
if block_n not in kv_cache.kv_block_map:
|
363 |
+
kv_cache.kv_block_map[block_n] = torch.empty(
|
364 |
+
kv_cache.max_batch_size,
|
365 |
+
kv_cache.max_seqlen,
|
366 |
+
2,
|
367 |
+
kv.shape[-2], # n_heads
|
368 |
+
kv.shape[-1], # d_head
|
369 |
+
dtype=kv.dtype,
|
370 |
+
device=kv.device,
|
371 |
+
)
|
372 |
+
kv_cache.kv_block_map[block_n][
|
373 |
+
kv_cache.batch_size_offset: kv_cache.batch_size_offset + kv.shape[0],
|
374 |
+
kv_cache.seqlen_offset: kv_cache.seqlen_offset + kv.shape[1],
|
375 |
+
...
|
376 |
+
] = kv
|
377 |
+
|
378 |
+
def _forward_cross_attn(
|
379 |
+
self,
|
380 |
+
qkv: torch.FloatTensor, # dim: (batch_size, seqlen, 3, n_heads, d_head)
|
381 |
+
kv_cache: KVCache,
|
382 |
+
key_padding_mask: torch.BoolTensor | None,
|
383 |
+
) -> torch.FloatTensor:
|
384 |
+
q = qkv[:, :, 0, :, :]
|
385 |
+
q = self.rotary_emb(
|
386 |
+
q,
|
387 |
+
seqlen_offset = 0 if kv_cache is None else kv_cache.seqlen_offset,
|
388 |
+
)
|
389 |
+
kv = cast(torch.FloatTensor, qkv[:, :, 1:, :, :])
|
390 |
+
self._update_kv_cache(kv, kv_cache, self.block_n)
|
391 |
+
causal = False # turning off causal mask for cross attention
|
392 |
+
|
393 |
+
if self.using_flash_attn and unpad_input and pad_input: # not touching flash attention code
|
394 |
+
batch_size, seqlen_q = q.shape[0], q.shape[1]
|
395 |
+
seqlen_k = kv.shape[1]
|
396 |
+
cu_seqlens_q, cu_seqlens_k, max_seqlen_q, max_seqlen_k, indices_q = (
|
397 |
+
None,
|
398 |
+
None,
|
399 |
+
None,
|
400 |
+
None,
|
401 |
+
None,
|
402 |
+
)
|
403 |
+
|
404 |
+
# unpad input and retrieve `cu_seqlens` and `max_seqlen` to be used by `flash-attn`
|
405 |
+
if key_padding_mask:
|
406 |
+
kv, _, cu_seqlens_k, max_seqlen_k = unpad_input(kv, key_padding_mask)
|
407 |
+
|
408 |
+
if seqlen_q == 1:
|
409 |
+
key_padding_mask = cast(torch.BoolTensor, torch.ones(batch_size, 1, device=q.device))
|
410 |
+
elif seqlen_q != seqlen_k:
|
411 |
+
key_padding_mask = cast(torch.BoolTensor, key_padding_mask[:, -seqlen_q:])
|
412 |
+
|
413 |
+
q, indices_q, cu_seqlens_q, max_seqlen_q = unpad_input(q, key_padding_mask)
|
414 |
+
|
415 |
+
if self.checkpointing:
|
416 |
+
attn_output = torch.utils.checkpoint.checkpoint(
|
417 |
+
self.inner_cross_attn,
|
418 |
+
q,
|
419 |
+
kv,
|
420 |
+
causal=causal,
|
421 |
+
cu_seqlens=cu_seqlens_q,
|
422 |
+
max_seqlen=max_seqlen_q,
|
423 |
+
cu_seqlens_k=cu_seqlens_k,
|
424 |
+
max_seqlen_k=max_seqlen_k,
|
425 |
+
)
|
426 |
+
else:
|
427 |
+
attn_output = self.inner_cross_attn(
|
428 |
+
q,
|
429 |
+
kv,
|
430 |
+
causal=causal,
|
431 |
+
cu_seqlens=cu_seqlens_q,
|
432 |
+
max_seqlen=max_seqlen_q,
|
433 |
+
cu_seqlens_k=cu_seqlens_k,
|
434 |
+
max_seqlen_k=max_seqlen_k,
|
435 |
+
)
|
436 |
+
|
437 |
+
if key_padding_mask:
|
438 |
+
return pad_input(attn_output, indices_q, batch_size, max_seqlen_q)
|
439 |
+
else:
|
440 |
+
return attn_output
|
441 |
+
|
442 |
+
if self.checkpointing:
|
443 |
+
return torch.utils.checkpoint.checkpoint(
|
444 |
+
self.inner_cross_attn,
|
445 |
+
q,
|
446 |
+
kv,
|
447 |
+
key_padding_mask=key_padding_mask,
|
448 |
+
causal=causal,
|
449 |
+
)
|
450 |
+
else:
|
451 |
+
return self.inner_cross_attn(q, kv, key_padding_mask=key_padding_mask, causal=causal)
|
452 |
+
|
453 |
+
def forward(
|
454 |
+
self,
|
455 |
+
x: torch.FloatTensor, # dim: (batch_size, seqlen, d_embedding)
|
456 |
+
kv_cache: KVCache | None = None,
|
457 |
+
key_padding_mask: torch.LongTensor | torch.BoolTensor | None = None,
|
458 |
+
) -> tuple[torch.FloatTensor, torch.FloatTensor]:
|
459 |
+
if key_padding_mask is not None:
|
460 |
+
key_padding_mask = cast(torch.BoolTensor, key_padding_mask.bool()) # make sure it's bool and not int
|
461 |
+
|
462 |
+
qkv = self.Wqkv(x) # dim: (batch_size, seqlen, 3*n_heads*d_head)
|
463 |
+
qkv = rearrange(qkv, "... (three h d) -> ... three h d", three=3, d=self.d_head) # dim: (batch_size, seqlen, 3, n_heads, d_head)
|
464 |
+
if kv_cache is None:
|
465 |
+
attn_output = self._forward_self_attn(qkv, key_padding_mask)
|
466 |
+
else:
|
467 |
+
attn_output = self._forward_cross_attn(qkv, kv_cache, key_padding_mask)
|
468 |
+
|
469 |
+
output = rearrange(attn_output, "... h d -> ... (h d)")
|
470 |
+
output = self.fc_out(output)
|
471 |
+
return output
|
472 |
+
|
473 |
+
|
474 |
+
class ParallelAttentionBlock(nn.Module):
|
475 |
+
"""From Phi2. Calculates attention and MLP in parallel. See 'Simplifying Transformer Blocks', Fig. 1 'Parallel'."""
|
476 |
+
|
477 |
+
def __init__(
|
478 |
+
self,
|
479 |
+
resid_pdrop: float, # a bit of a misnomer, right?
|
480 |
+
layer_norm_epsilon: float,
|
481 |
+
d_embedding: int,
|
482 |
+
n_attn_heads: int,
|
483 |
+
block_n: int,
|
484 |
+
initial_cos_sin_cache_len: int, # length of cache for rotary embedding
|
485 |
+
attn_pdrop: float,
|
486 |
+
use_flash_rotary: bool = True, # use flash rotary embedding if possible
|
487 |
+
use_flash_attn: bool = True, # use flash attention if possible
|
488 |
+
use_fused_dense: bool = True, # use fused dense layer if possible
|
489 |
+
checkpointing: bool = False, # torch.utils.checkpoint
|
490 |
+
) -> None:
|
491 |
+
super().__init__()
|
492 |
+
self.layer_norm = nn.LayerNorm(d_embedding, eps=layer_norm_epsilon)
|
493 |
+
self.block_n = block_n
|
494 |
+
self.multi_head_attention = MHA(
|
495 |
+
d_embedding=d_embedding,
|
496 |
+
n_attn_heads=n_attn_heads,
|
497 |
+
block_n=block_n,
|
498 |
+
initial_cos_sin_cache_len=initial_cos_sin_cache_len,
|
499 |
+
attn_pdrop=attn_pdrop,
|
500 |
+
use_flash_rotary=use_flash_rotary,
|
501 |
+
use_flash_attn=use_flash_attn,
|
502 |
+
use_fused_dense=use_fused_dense,
|
503 |
+
checkpointing=checkpointing,
|
504 |
+
)
|
505 |
+
self.mlp = MLP(d_embedding)
|
506 |
+
self.dropout = nn.Dropout(resid_pdrop)
|
507 |
+
|
508 |
+
def forward(
|
509 |
+
self,
|
510 |
+
x: torch.FloatTensor, # dim: (batch_size, seq_len, d_embedding)
|
511 |
+
kv_cache: KVCache | None = None,
|
512 |
+
key_padding_mask: torch.BoolTensor | None = None,
|
513 |
+
) -> torch.FloatTensor:
|
514 |
+
residual = x
|
515 |
+
x = self.layer_norm(x) # each token (dim: d_embedding) is normalized individually
|
516 |
+
attn_outputs = self.multi_head_attention(
|
517 |
+
x,
|
518 |
+
kv_cache=kv_cache,
|
519 |
+
key_padding_mask=key_padding_mask,
|
520 |
+
)
|
521 |
+
mlp_outputs = self.mlp(x)
|
522 |
+
return self.dropout(attn_outputs + mlp_outputs) + residual
|
config.json
CHANGED
@@ -1,29 +1,32 @@
|
|
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 |
}
|
|
|
1 |
{
|
2 |
+
"_name_or_path": "BucketOfFish/simplified_phi2",
|
3 |
+
"architectures": [
|
4 |
+
"Phi2Model",
|
5 |
+
"Phi2ModelForCausalLM"
|
6 |
+
],
|
7 |
+
"auto_map": {
|
8 |
+
"AutoConfig": "phi2_configuration.Phi2Config",
|
9 |
+
"AutoModel": "phi2_model.Phi2Model",
|
10 |
+
"AutoModelForCausalLM": "phi2_model.Phi2ModelForCausalLM"
|
11 |
+
},
|
12 |
+
"model_type": "phi2",
|
13 |
+
"torch_dtype": "float16",
|
14 |
+
"transformers_version": "4.29.0",
|
15 |
+
|
16 |
+
"vocab_size": 50304,
|
17 |
+
"vocab_chunk_for_gpu_efficiency": 64,
|
18 |
+
"initial_cos_sin_cache_len": 2048,
|
19 |
+
"d_embedding": 2560,
|
20 |
+
"n_blocks": 32,
|
21 |
+
"n_heads": 32,
|
22 |
+
"use_flash_attn": false,
|
23 |
+
"use_flash_rotary": false,
|
24 |
+
"use_fused_dense": false,
|
25 |
+
"attn_pdrop": 0.0,
|
26 |
+
"embd_pdrop": 0.0,
|
27 |
+
"resid_pdrop": 0.1,
|
28 |
+
"layer_norm_epsilon": 1e-05,
|
29 |
+
"weight_initialization_range": 0.02,
|
30 |
+
"tie_word_embeddings": false,
|
31 |
+
"checkpointing": false
|
32 |
}
|
configuration_phi.py
DELETED
@@ -1,56 +0,0 @@
|
|
1 |
-
# Copyright (c) Microsoft Corporation.
|
2 |
-
# Licensed under the MIT license.
|
3 |
-
|
4 |
-
import math
|
5 |
-
from typing import Optional
|
6 |
-
|
7 |
-
from transformers import PretrainedConfig
|
8 |
-
|
9 |
-
|
10 |
-
class PhiConfig(PretrainedConfig):
|
11 |
-
"""Phi configuration."""
|
12 |
-
|
13 |
-
model_type = "phi-msft"
|
14 |
-
attribute_map = {
|
15 |
-
"max_position_embeddings": "n_positions",
|
16 |
-
"hidden_size": "n_embd",
|
17 |
-
"num_attention_heads": "n_head",
|
18 |
-
"num_hidden_layers": "n_layer",
|
19 |
-
}
|
20 |
-
|
21 |
-
def __init__(
|
22 |
-
self,
|
23 |
-
vocab_size: int = 50304,
|
24 |
-
n_positions: int = 2048,
|
25 |
-
n_embd: int = 1024,
|
26 |
-
n_layer: int = 20,
|
27 |
-
n_inner: Optional[int] = None,
|
28 |
-
n_head: int = 16,
|
29 |
-
n_head_kv: Optional[int] = None,
|
30 |
-
rotary_dim: Optional[int] = 32,
|
31 |
-
activation_function: Optional[str] = "gelu_new",
|
32 |
-
attn_pdrop: float = 0.0,
|
33 |
-
embd_pdrop: float = 0.0,
|
34 |
-
resid_pdrop: float = 0.0,
|
35 |
-
layer_norm_epsilon: float = 1e-5,
|
36 |
-
initializer_range: float = 0.02,
|
37 |
-
tie_word_embeddings: bool = False,
|
38 |
-
pad_vocab_size_multiple: int = 64,
|
39 |
-
**kwargs
|
40 |
-
) -> None:
|
41 |
-
self.vocab_size = int(math.ceil(vocab_size / pad_vocab_size_multiple) * pad_vocab_size_multiple)
|
42 |
-
self.n_positions = n_positions
|
43 |
-
self.n_embd = n_embd
|
44 |
-
self.n_layer = n_layer
|
45 |
-
self.n_inner = n_inner
|
46 |
-
self.n_head = n_head
|
47 |
-
self.n_head_kv = n_head_kv
|
48 |
-
self.rotary_dim = min(rotary_dim, n_embd // n_head)
|
49 |
-
self.activation_function = activation_function
|
50 |
-
self.attn_pdrop = attn_pdrop
|
51 |
-
self.embd_pdrop = embd_pdrop
|
52 |
-
self.resid_pdrop = resid_pdrop
|
53 |
-
self.layer_norm_epsilon = layer_norm_epsilon
|
54 |
-
self.initializer_range = initializer_range
|
55 |
-
|
56 |
-
super().__init__(tie_word_embeddings=tie_word_embeddings, **kwargs)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
modeling_phi.py
DELETED
@@ -1,766 +0,0 @@
|
|
1 |
-
# Copyright (c) Microsoft Corporation.
|
2 |
-
# Licensed under the MIT license.
|
3 |
-
#
|
4 |
-
# Copyright (c) 2022, Tri Dao, [email protected].
|
5 |
-
# Licensed under the BSD 3-Clause License.
|
6 |
-
|
7 |
-
from __future__ import annotations
|
8 |
-
|
9 |
-
import math
|
10 |
-
from dataclasses import dataclass, field
|
11 |
-
from typing import Any, Dict, Optional, Tuple, Union
|
12 |
-
|
13 |
-
import torch
|
14 |
-
import torch.nn as nn
|
15 |
-
from einops import rearrange, repeat
|
16 |
-
from transformers import PretrainedConfig, PreTrainedModel
|
17 |
-
from transformers.activations import ACT2FN
|
18 |
-
from transformers.modeling_outputs import CausalLMOutputWithPast
|
19 |
-
|
20 |
-
from .configuration_phi import PhiConfig
|
21 |
-
|
22 |
-
|
23 |
-
@dataclass
|
24 |
-
class InferenceParams:
|
25 |
-
"""Inference parameters passed to model to efficiently calculate
|
26 |
-
and store context during inference.
|
27 |
-
|
28 |
-
Reference:
|
29 |
-
https://github.com/Dao-AILab/flash-attention/blob/main/flash_attn/utils/generation.py.
|
30 |
-
|
31 |
-
Args:
|
32 |
-
max_seqlen: Maximum sequence length.
|
33 |
-
max_batch_size: Maximum batch size.
|
34 |
-
seqlen_offset: Sequence length offset.
|
35 |
-
batch_size_offset: Batch size offset.
|
36 |
-
key_value_memory_dict: Key value memory dictionary.
|
37 |
-
lengths_per_sample: Lengths per sample.
|
38 |
-
|
39 |
-
"""
|
40 |
-
|
41 |
-
max_seqlen: int = field(metadata={"help": "Maximum sequence length."})
|
42 |
-
|
43 |
-
max_batch_size: int = field(metadata={"help": "Maximum batch size."})
|
44 |
-
|
45 |
-
seqlen_offset: int = field(default=0, metadata={"help": "Sequence length offset."})
|
46 |
-
|
47 |
-
batch_size_offset: int = field(default=0, metadata={"help": "Batch size offset."})
|
48 |
-
|
49 |
-
key_value_memory_dict: Dict[str, Any] = field(
|
50 |
-
default_factory=dict, metadata={"help": "Key value memory dictionary."}
|
51 |
-
)
|
52 |
-
|
53 |
-
lengths_per_sample: torch.Tensor = field(default=None, metadata={"help": "Lengths per sample."})
|
54 |
-
|
55 |
-
|
56 |
-
class Embedding(nn.Module):
|
57 |
-
"""Token embedding with dropout."""
|
58 |
-
|
59 |
-
def __init__(self, config: PretrainedConfig) -> None:
|
60 |
-
super().__init__()
|
61 |
-
|
62 |
-
self.wte = nn.Embedding(config.vocab_size, config.n_embd)
|
63 |
-
self.drop = nn.Dropout(config.embd_pdrop)
|
64 |
-
|
65 |
-
def forward(self, input_ids: torch.LongTensor) -> torch.FloatTensor:
|
66 |
-
input_shape = input_ids.size()
|
67 |
-
input_ids = input_ids.view(-1, input_shape[-1])
|
68 |
-
|
69 |
-
hidden_states = self.wte(input_ids)
|
70 |
-
hidden_states = self.drop(hidden_states)
|
71 |
-
|
72 |
-
return hidden_states
|
73 |
-
|
74 |
-
|
75 |
-
class RotaryEmbedding(nn.Module):
|
76 |
-
"""Rotary positional embedding (RoPE) from Phi2.
|
77 |
-
See https://www.youtube.com/watch?v=C6rV8BsrrCc
|
78 |
-
"""
|
79 |
-
|
80 |
-
def __init__(
|
81 |
-
self,
|
82 |
-
d_rotary: int,
|
83 |
-
rotary_base: float = 10000.0,
|
84 |
-
initial_cos_sin_cache_len: int = 2048,
|
85 |
-
device: torch.device | None = None,
|
86 |
-
) -> None:
|
87 |
-
super().__init__()
|
88 |
-
self.d_rotary = d_rotary
|
89 |
-
self.rotary_base = rotary_base
|
90 |
-
self.device = device
|
91 |
-
self.dtype = torch.float32
|
92 |
-
self._update_cos_sin_cache(seqlen=initial_cos_sin_cache_len)
|
93 |
-
|
94 |
-
def _update_cos_sin_cache(self, seqlen: int) -> None:
|
95 |
-
# only call this function when seqlen is larger than _max_seqlen
|
96 |
-
self._max_seqlen = seqlen
|
97 |
-
|
98 |
-
# m * theta_i = m * base^(-2i/d) = m * (1 / base^(2i/d)), where i in [1, d/2]
|
99 |
-
m = torch.arange(
|
100 |
-
seqlen,
|
101 |
-
device=self.device,
|
102 |
-
dtype=self.dtype,
|
103 |
-
)
|
104 |
-
theta_i = 1.0 / (
|
105 |
-
self.rotary_base ** (
|
106 |
-
torch.arange(
|
107 |
-
start=0,
|
108 |
-
end=self.d_rotary,
|
109 |
-
step=2,
|
110 |
-
device=self.device,
|
111 |
-
dtype=self.dtype,
|
112 |
-
) / self.d_rotary
|
113 |
-
)
|
114 |
-
)
|
115 |
-
# torch.outer, since torch.einsum converts from fp32 to fp16 if used with torch.amp
|
116 |
-
# TODO: does this matter if I'm disabling torch.autocast?
|
117 |
-
m_theta_i = torch.outer(m, theta_i)
|
118 |
-
self._cos_cached = torch.cos(m_theta_i).to(self.dtype)
|
119 |
-
self._sin_cached = torch.sin(m_theta_i).to(self.dtype)
|
120 |
-
|
121 |
-
# TODO: scale_base caching is labelled as not yet done in Phi2
|
122 |
-
"""
|
123 |
-
if scale_base is not None:
|
124 |
-
scale = (
|
125 |
-
torch.arange(
|
126 |
-
start=0,
|
127 |
-
end=self.d_rotary,
|
128 |
-
step=2,
|
129 |
-
device=self.device,
|
130 |
-
dtype=torch.float32,
|
131 |
-
) + 0.4 * self.d_rotary
|
132 |
-
) / (1.4 * self.d_rotary)
|
133 |
-
power = (
|
134 |
-
torch.arange(seqlen, dtype=scale.dtype, device=scale.device) - seqlen // 2
|
135 |
-
) / scale_base
|
136 |
-
scale = scale.to(device=power.device) ** rearrange(power, "s -> s 1")
|
137 |
-
self._cos_cached = (torch.cos(m_theta_i) * scale).to(dtype)
|
138 |
-
self._sin_cached = (torch.sin(m_theta_i) * scale).to(dtype)
|
139 |
-
"""
|
140 |
-
|
141 |
-
def _apply_rotary_emb_qkv(
|
142 |
-
self,
|
143 |
-
x: torch.FloatTensor, # dim: (batch_size, seqlen, Optional[n_qkv], n_heads, d_head)
|
144 |
-
cos: torch.FloatTensor, # dim: (_max_seqlen, d_rotary)
|
145 |
-
sin: torch.FloatTensor, # dim: (_max_seqlen, d_rotary)
|
146 |
-
) -> torch.FloatTensor:
|
147 |
-
seqlen = x.shape[1]
|
148 |
-
x1, x2 = x.chunk(2, dim=-1) # dim: (batch_size, seqlen, Optional[n_qkv], n_heads, d_head/2)
|
149 |
-
broadcast_rearrange = "s d -> s 1 d" if x1.ndim == 4 else "s d -> s 1 1 d"
|
150 |
-
c, s = rearrange(cos[:seqlen], broadcast_rearrange), rearrange(sin[:seqlen], broadcast_rearrange)
|
151 |
-
x1, x2, c, s = [t.to(dtype=torch.float32) for t in [x1, x2, c, s]] # make sure rotary embedding is in float32
|
152 |
-
return cast(
|
153 |
-
torch.FloatTensor,
|
154 |
-
torch.cat([x1 * c - x2 * s, x1 * s + x2 * c], dim=-1).to(x.dtype)
|
155 |
-
)
|
156 |
-
|
157 |
-
def forward(
|
158 |
-
self,
|
159 |
-
x: torch.FloatTensor, # dim: (batch_size, seqlen, Optional[n_qkv], n_heads, d_head)
|
160 |
-
seqlen_offset: int = 0, # each sequence is shifted by this amount - used in inference with KV cache
|
161 |
-
) -> torch.FloatTensor:
|
162 |
-
if (
|
163 |
-
not self._max_seqlen
|
164 |
-
or self._max_seqlen < x.shape[1] + seqlen_offset
|
165 |
-
or self._cos_cached.device != x.device
|
166 |
-
or self._cos_cached.dtype != x.dtype
|
167 |
-
or (self.training and self._cos_cached.is_inference())
|
168 |
-
):
|
169 |
-
self._update_cos_sin_cache(seqlen=x.shape[1] + seqlen_offset)
|
170 |
-
return self._apply_rotary_emb_qkv(
|
171 |
-
x,
|
172 |
-
cast(torch.FloatTensor, self._cos_cached[seqlen_offset:]),
|
173 |
-
cast(torch.FloatTensor, self._sin_cached[seqlen_offset:]),
|
174 |
-
)
|
175 |
-
|
176 |
-
|
177 |
-
class MLP(nn.Module):
|
178 |
-
"""Multi-Layer Perceptron.
|
179 |
-
|
180 |
-
Reference:
|
181 |
-
Attention Is All You Need.
|
182 |
-
https://arxiv.org/pdf/1706.03762.pdf.
|
183 |
-
|
184 |
-
"""
|
185 |
-
|
186 |
-
def __init__(
|
187 |
-
self,
|
188 |
-
config: PretrainedConfig,
|
189 |
-
n_inner: Optional[int] = None,
|
190 |
-
act_fn: Optional[str] = None,
|
191 |
-
) -> None:
|
192 |
-
super().__init__()
|
193 |
-
|
194 |
-
act_fn = config.activation_function if act_fn is None else act_fn
|
195 |
-
|
196 |
-
n_inner = getattr(config, "n_inner", None) if n_inner is None else n_inner
|
197 |
-
n_inner = n_inner if n_inner is not None else 4 * config.n_embd
|
198 |
-
|
199 |
-
self.fc1 = nn.Linear(config.n_embd, n_inner)
|
200 |
-
self.fc2 = nn.Linear(n_inner, config.n_embd)
|
201 |
-
self.act = ACT2FN[act_fn]
|
202 |
-
|
203 |
-
def forward(self, hidden_states: torch.FloatTensor) -> torch.FloatTensor:
|
204 |
-
hidden_states = self.fc1(hidden_states)
|
205 |
-
hidden_states = self.act(hidden_states)
|
206 |
-
hidden_states = self.fc2(hidden_states)
|
207 |
-
|
208 |
-
return hidden_states
|
209 |
-
|
210 |
-
|
211 |
-
class SelfAttention(nn.Module):
|
212 |
-
"""Self-attention layer (compatible with PyTorch).
|
213 |
-
|
214 |
-
Reference:
|
215 |
-
https://github.com/Dao-AILab/flash-attention/blob/main/flash_attn/modules/mha.py.
|
216 |
-
|
217 |
-
"""
|
218 |
-
|
219 |
-
def __init__(
|
220 |
-
self,
|
221 |
-
causal: bool = True,
|
222 |
-
softmax_scale: Optional[float] = None,
|
223 |
-
attention_dropout: float = 0.0,
|
224 |
-
) -> None:
|
225 |
-
super().__init__()
|
226 |
-
|
227 |
-
self.causal = causal
|
228 |
-
self.softmax_scale = softmax_scale
|
229 |
-
self.drop = nn.Dropout(attention_dropout)
|
230 |
-
|
231 |
-
@torch.autocast("cpu", enabled=False)
|
232 |
-
@torch.autocast("cuda", enabled=False)
|
233 |
-
def forward(
|
234 |
-
self,
|
235 |
-
qkv: torch.FloatTensor,
|
236 |
-
causal: bool = None,
|
237 |
-
key_padding_mask: Optional[torch.BoolTensor] = None,
|
238 |
-
**kwargs,
|
239 |
-
) -> torch.FloatTensor:
|
240 |
-
batch_size, seqlen = qkv.shape[0], qkv.shape[1]
|
241 |
-
q, k, v = qkv.unbind(dim=2)
|
242 |
-
|
243 |
-
q = q.to(torch.float32)
|
244 |
-
k = k.to(torch.float32)
|
245 |
-
|
246 |
-
causal = self.causal if causal is None else causal
|
247 |
-
softmax_scale = self.softmax_scale or 1.0 / math.sqrt(q.shape[-1])
|
248 |
-
|
249 |
-
# Autocast is manually disabled to avoid `torch.einsum` performing the operation
|
250 |
-
# using float16, which might lead to overflow
|
251 |
-
scores = torch.einsum("bthd,bshd->bhts", q, k * softmax_scale)
|
252 |
-
|
253 |
-
if key_padding_mask is not None:
|
254 |
-
padding_mask = torch.full((batch_size, seqlen), -10000.0, dtype=scores.dtype, device=scores.device)
|
255 |
-
padding_mask.masked_fill_(key_padding_mask, 0.0)
|
256 |
-
|
257 |
-
scores = scores + rearrange(padding_mask, "b s -> b 1 1 s")
|
258 |
-
|
259 |
-
if causal:
|
260 |
-
causal_mask = torch.triu(torch.full((seqlen, seqlen), -10000.0, device=scores.device), 1)
|
261 |
-
scores = scores + causal_mask.to(dtype=scores.dtype)
|
262 |
-
|
263 |
-
attention = torch.softmax(scores, dim=-1).to(v.dtype)
|
264 |
-
attention = self.drop(attention)
|
265 |
-
|
266 |
-
output = torch.einsum("bhts,bshd->bthd", attention, v)
|
267 |
-
|
268 |
-
return output
|
269 |
-
|
270 |
-
|
271 |
-
class CrossAttention(nn.Module):
|
272 |
-
"""Cross-attention layer (compatible with PyTorch).
|
273 |
-
|
274 |
-
Reference:
|
275 |
-
https://github.com/Dao-AILab/flash-attention/blob/main/flash_attn/modules/mha.py.
|
276 |
-
|
277 |
-
"""
|
278 |
-
|
279 |
-
def __init__(
|
280 |
-
self,
|
281 |
-
causal: bool = True,
|
282 |
-
softmax_scale: Optional[float] = None,
|
283 |
-
attention_dropout: float = 0.0,
|
284 |
-
) -> None:
|
285 |
-
super().__init__()
|
286 |
-
|
287 |
-
self.causal = causal
|
288 |
-
self.softmax_scale = softmax_scale
|
289 |
-
self.drop = nn.Dropout(attention_dropout)
|
290 |
-
|
291 |
-
@torch.autocast("cpu", enabled=False)
|
292 |
-
@torch.autocast("cuda", enabled=False)
|
293 |
-
def forward(
|
294 |
-
self,
|
295 |
-
q: torch.FloatTensor,
|
296 |
-
kv: torch.FloatTensor,
|
297 |
-
causal: bool = None,
|
298 |
-
key_padding_mask: Optional[torch.BoolTensor] = None,
|
299 |
-
**kwargs,
|
300 |
-
) -> torch.FloatTensor:
|
301 |
-
batch_size, seqlen_q = q.shape[0], q.shape[1]
|
302 |
-
seqlen_k = kv.shape[1]
|
303 |
-
|
304 |
-
if kv.shape[3] != q.shape[2]:
|
305 |
-
kv = repeat(kv, "... hkv d -> ... (hkv g) d", g=q.shape[2] // kv.shape[3])
|
306 |
-
k, v = kv.unbind(dim=2)
|
307 |
-
|
308 |
-
q = q.to(torch.float32)
|
309 |
-
k = k.to(torch.float32)
|
310 |
-
|
311 |
-
causal = self.causal if causal is None else causal
|
312 |
-
softmax_scale = self.softmax_scale or 1.0 / math.sqrt(q.shape[-1])
|
313 |
-
|
314 |
-
# Autocast is manually disabled to avoid `torch.einsum` performing the operation
|
315 |
-
# using float16, which might lead to overflow
|
316 |
-
scores = torch.einsum("bthd,bshd->bhts", q, k * softmax_scale)
|
317 |
-
|
318 |
-
if key_padding_mask is not None:
|
319 |
-
padding_mask = torch.full(
|
320 |
-
(batch_size, seqlen_k),
|
321 |
-
-10000.0,
|
322 |
-
dtype=scores.dtype,
|
323 |
-
device=scores.device,
|
324 |
-
)
|
325 |
-
padding_mask.masked_fill_(key_padding_mask, 0.0)
|
326 |
-
|
327 |
-
scores = scores + rearrange(padding_mask, "b s -> b 1 1 s")
|
328 |
-
|
329 |
-
if causal:
|
330 |
-
rows = rearrange(torch.arange(seqlen_q, device=q.device, dtype=torch.long), "s -> s 1")
|
331 |
-
cols = torch.arange(seqlen_k, device=k.device, dtype=torch.long)
|
332 |
-
causal_mask = cols > rows + seqlen_k - seqlen_q
|
333 |
-
|
334 |
-
scores = scores.masked_fill(causal_mask, -10000.0)
|
335 |
-
|
336 |
-
attention = torch.softmax(scores, dim=-1).to(v.dtype)
|
337 |
-
attention = self.drop(attention)
|
338 |
-
|
339 |
-
output = torch.einsum("bhts,bshd->bthd", attention, v)
|
340 |
-
|
341 |
-
return output
|
342 |
-
|
343 |
-
|
344 |
-
def _find_mha_dims(
|
345 |
-
config: PretrainedConfig,
|
346 |
-
n_head: Optional[int] = None,
|
347 |
-
n_head_kv: Optional[int] = None,
|
348 |
-
head_dim: Optional[int] = None,
|
349 |
-
) -> Tuple[int, int]:
|
350 |
-
if n_head is None and head_dim is None:
|
351 |
-
head_dim = config.n_embd // config.n_head
|
352 |
-
n_head = config.n_head
|
353 |
-
elif n_head is None or head_dim is None:
|
354 |
-
raise ValueError("`n_head` and `head_dim` must be both specified or `None`.")
|
355 |
-
|
356 |
-
if n_head_kv is None:
|
357 |
-
n_head_kv = getattr(config, "n_head_kv", None) or n_head
|
358 |
-
|
359 |
-
return n_head, n_head_kv, head_dim
|
360 |
-
|
361 |
-
|
362 |
-
def _update_kv_cache(kv: torch.FloatTensor, inference_params: InferenceParams, layer_idx: int) -> torch.FloatTensor:
|
363 |
-
num_heads, head_dim = kv.shape[-2:]
|
364 |
-
|
365 |
-
if layer_idx not in inference_params.key_value_memory_dict:
|
366 |
-
inference_params.key_value_memory_dict[layer_idx] = torch.empty(
|
367 |
-
inference_params.max_batch_size,
|
368 |
-
inference_params.max_seqlen,
|
369 |
-
2,
|
370 |
-
num_heads,
|
371 |
-
head_dim,
|
372 |
-
dtype=kv.dtype,
|
373 |
-
device=kv.device,
|
374 |
-
)
|
375 |
-
|
376 |
-
batch_start = inference_params.batch_size_offset
|
377 |
-
batch_end = batch_start + kv.shape[0]
|
378 |
-
|
379 |
-
sequence_start = inference_params.seqlen_offset
|
380 |
-
sequence_end = sequence_start + kv.shape[1]
|
381 |
-
|
382 |
-
# When the current sequence length is equal to or larger than the maximum sequence length,
|
383 |
-
# we need to concatenate the current `kv` with the cached `kv` to expand its length
|
384 |
-
if sequence_end >= inference_params.max_seqlen:
|
385 |
-
inference_params.key_value_memory_dict[layer_idx] = torch.concatenate((inference_params.key_value_memory_dict[layer_idx], kv), dim=1)
|
386 |
-
|
387 |
-
inference_params.key_value_memory_dict[layer_idx][batch_start:batch_end, sequence_start:sequence_end, ...] = kv
|
388 |
-
kv = inference_params.key_value_memory_dict[layer_idx][batch_start:batch_end, :sequence_end, ...]
|
389 |
-
|
390 |
-
return kv
|
391 |
-
|
392 |
-
|
393 |
-
class MHA(nn.Module):
|
394 |
-
"""Multi-head attention layer."""
|
395 |
-
|
396 |
-
def __init__(
|
397 |
-
self,
|
398 |
-
config: PretrainedConfig,
|
399 |
-
dtype: Optional[torch.dtype] = None,
|
400 |
-
device: Optional[str] = None,
|
401 |
-
rotary_dim: Optional[int] = None,
|
402 |
-
rotary_base: float = 10000.0,
|
403 |
-
rotary_scale_base: Optional[float] = None,
|
404 |
-
n_head: Optional[int] = None,
|
405 |
-
n_head_kv: Optional[int] = None,
|
406 |
-
head_dim: Optional[int] = None,
|
407 |
-
bias: bool = True,
|
408 |
-
causal: bool = True,
|
409 |
-
softmax_scale: Optional[float] = None,
|
410 |
-
layer_idx: Optional[int] = None,
|
411 |
-
return_residual: bool = False,
|
412 |
-
checkpointing: bool = False,
|
413 |
-
) -> None:
|
414 |
-
super().__init__()
|
415 |
-
|
416 |
-
# Rotary embedding
|
417 |
-
self.rotary_dim = rotary_dim if rotary_dim is not None else getattr(config, "rotary_dim", 0)
|
418 |
-
if self.rotary_dim > 0:
|
419 |
-
self.rotary_emb = RotaryEmbedding(
|
420 |
-
d_rotary=self.rotary_dim,
|
421 |
-
# d_rotary=math.ceil((rotary_dim // n_head) / 2), # d_rotary is half of d_head
|
422 |
-
initial_cos_sin_cache_len=config.n_positions,
|
423 |
-
)
|
424 |
-
|
425 |
-
# MLP
|
426 |
-
self.n_head, self.n_head_kv, self.head_dim = _find_mha_dims(
|
427 |
-
config, n_head=n_head, n_head_kv=n_head_kv, head_dim=head_dim
|
428 |
-
)
|
429 |
-
op_size = self.head_dim * (self.n_head + 2 * self.n_head_kv)
|
430 |
-
hidden_size = config.n_embd
|
431 |
-
|
432 |
-
linear_cls = nn.Linear
|
433 |
-
if linear_cls is None:
|
434 |
-
linear_cls = nn.Linear
|
435 |
-
|
436 |
-
self.Wqkv = linear_cls(hidden_size, op_size, bias=bias, device=device, dtype=dtype)
|
437 |
-
self.out_proj = linear_cls(hidden_size, hidden_size, bias=bias, device=device, dtype=dtype)
|
438 |
-
|
439 |
-
# Attention
|
440 |
-
attn_cls = SelfAttention
|
441 |
-
if attn_cls is None:
|
442 |
-
attn_cls = SelfAttention
|
443 |
-
|
444 |
-
cross_attn_cls = CrossAttention
|
445 |
-
if cross_attn_cls is None:
|
446 |
-
cross_attn_cls = CrossAttention
|
447 |
-
|
448 |
-
self.inner_attn = attn_cls(
|
449 |
-
causal=causal,
|
450 |
-
softmax_scale=softmax_scale,
|
451 |
-
attention_dropout=config.attn_pdrop,
|
452 |
-
)
|
453 |
-
self.inner_cross_attn = cross_attn_cls(
|
454 |
-
causal=causal,
|
455 |
-
softmax_scale=softmax_scale,
|
456 |
-
attention_dropout=config.attn_pdrop,
|
457 |
-
)
|
458 |
-
|
459 |
-
self.layer_idx = layer_idx
|
460 |
-
self.return_residual = return_residual
|
461 |
-
self.checkpointing = checkpointing
|
462 |
-
|
463 |
-
def _forward_self_attn(
|
464 |
-
self, x: torch.FloatTensor, key_padding_mask: Optional[torch.BoolTensor]
|
465 |
-
) -> torch.FloatTensor:
|
466 |
-
qkv = self.Wqkv(x)
|
467 |
-
qkv = rearrange(qkv, "... (three h d) -> ... three h d", three=3, d=self.head_dim)
|
468 |
-
|
469 |
-
if self.rotary_dim > 0:
|
470 |
-
qkv = self.rotary_emb(qkv)
|
471 |
-
|
472 |
-
if self.checkpointing:
|
473 |
-
return torch.utils.checkpoint.checkpoint(self.inner_attn, qkv, key_padding_mask=key_padding_mask)
|
474 |
-
|
475 |
-
return self.inner_attn(qkv, key_padding_mask=key_padding_mask)
|
476 |
-
|
477 |
-
def _forward_cross_attn(
|
478 |
-
self,
|
479 |
-
x: torch.FloatTensor,
|
480 |
-
past_key_values: Optional[InferenceParams],
|
481 |
-
key_padding_mask: Optional[torch.BoolTensor],
|
482 |
-
) -> torch.FloatTensor:
|
483 |
-
batch_size = x.shape[0]
|
484 |
-
|
485 |
-
qkv = self.Wqkv(x)
|
486 |
-
|
487 |
-
q = qkv[..., : self.n_head * self.head_dim]
|
488 |
-
q = rearrange(q, "... (h d) -> ... h d", d=self.head_dim)
|
489 |
-
|
490 |
-
kv = qkv[..., self.n_head * self.head_dim :]
|
491 |
-
kv = rearrange(kv, "... (two hkv d) -> ... two hkv d", two=2, d=self.head_dim)
|
492 |
-
|
493 |
-
seqlen_offset = past_key_values.seqlen_offset if past_key_values is not None else 0
|
494 |
-
causal = None if seqlen_offset == 0 else False
|
495 |
-
if self.rotary_dim > 0:
|
496 |
-
q, kv = self.rotary_emb(q, kv=kv, seqlen_offset=seqlen_offset)
|
497 |
-
|
498 |
-
if past_key_values is not None:
|
499 |
-
kv = _update_kv_cache(kv, past_key_values, self.layer_idx)
|
500 |
-
|
501 |
-
if self.checkpointing:
|
502 |
-
return torch.utils.checkpoint.checkpoint(
|
503 |
-
self.inner_cross_attn,
|
504 |
-
q,
|
505 |
-
kv,
|
506 |
-
key_padding_mask=key_padding_mask,
|
507 |
-
causal=causal,
|
508 |
-
)
|
509 |
-
|
510 |
-
return self.inner_cross_attn(q, kv, key_padding_mask=key_padding_mask, causal=causal)
|
511 |
-
|
512 |
-
def forward(
|
513 |
-
self,
|
514 |
-
x: torch.FloatTensor,
|
515 |
-
past_key_values: Optional[InferenceParams] = None,
|
516 |
-
attention_mask: Optional[Union[torch.LongTensor, torch.BoolTensor]] = None,
|
517 |
-
**kwargs,
|
518 |
-
) -> Tuple[torch.FloatTensor, torch.FloatTensor]:
|
519 |
-
if attention_mask is not None:
|
520 |
-
attention_mask = attention_mask.bool()
|
521 |
-
else:
|
522 |
-
attention_mask = None
|
523 |
-
|
524 |
-
# MHA
|
525 |
-
if self.n_head == self.n_head_kv:
|
526 |
-
if past_key_values is None:
|
527 |
-
# If `past_key_values` are not supplied, we run self-attention
|
528 |
-
attn_output = self._forward_self_attn(x, attention_mask)
|
529 |
-
else:
|
530 |
-
# If `past_key_values` are supplied, it means that we might have cached values and
|
531 |
-
# could take advantage of cross-attention
|
532 |
-
attn_output = self._forward_cross_attn(x, past_key_values, attention_mask)
|
533 |
-
# MQA / GQA
|
534 |
-
else:
|
535 |
-
# Regardless of `past_key_values` being supplied or not, it always use cross-attention
|
536 |
-
# because `q` and `kv` lengths might be different
|
537 |
-
attn_output = self._forward_cross_attn(x, past_key_values, attention_mask)
|
538 |
-
|
539 |
-
output = rearrange(attn_output, "... h d -> ... (h d)")
|
540 |
-
output = self.out_proj(output)
|
541 |
-
|
542 |
-
return output if not self.return_residual else (output, x)
|
543 |
-
|
544 |
-
|
545 |
-
class ParallelBlock(nn.Module):
|
546 |
-
"""Parallel block.
|
547 |
-
|
548 |
-
This block applies parallel mixer and MLP layers to the input (used in GPT-J and CodeGen).
|
549 |
-
|
550 |
-
"""
|
551 |
-
|
552 |
-
def __init__(
|
553 |
-
self,
|
554 |
-
config: PretrainedConfig,
|
555 |
-
block_idx: Optional[int] = None,
|
556 |
-
) -> None:
|
557 |
-
super().__init__()
|
558 |
-
|
559 |
-
self.ln = nn.LayerNorm(config.n_embd, eps=config.layer_norm_epsilon)
|
560 |
-
self.resid_dropout = nn.Dropout(config.resid_pdrop)
|
561 |
-
self.block_idx = block_idx
|
562 |
-
|
563 |
-
self.mixer = MHA(config, layer_idx=block_idx)
|
564 |
-
self.mlp = MLP(config)
|
565 |
-
|
566 |
-
def forward(
|
567 |
-
self,
|
568 |
-
hidden_states: torch.FloatTensor,
|
569 |
-
past_key_values: Optional[Union[torch.FloatTensor, InferenceParams]] = None,
|
570 |
-
attention_mask: Optional[torch.BoolTensor] = None,
|
571 |
-
**kwargs,
|
572 |
-
) -> torch.FloatTensor:
|
573 |
-
residual = hidden_states
|
574 |
-
hidden_states = self.ln(hidden_states)
|
575 |
-
|
576 |
-
attn_outputs = self.mixer(
|
577 |
-
hidden_states,
|
578 |
-
past_key_values=past_key_values,
|
579 |
-
attention_mask=attention_mask,
|
580 |
-
)
|
581 |
-
if isinstance(attn_outputs, tuple):
|
582 |
-
attn_outputs = attn_outputs[0]
|
583 |
-
|
584 |
-
attn_outputs = self.resid_dropout(attn_outputs)
|
585 |
-
feed_forward_hidden_states = self.resid_dropout(self.mlp(hidden_states))
|
586 |
-
|
587 |
-
hidden_states = attn_outputs + feed_forward_hidden_states + residual
|
588 |
-
|
589 |
-
return hidden_states
|
590 |
-
|
591 |
-
|
592 |
-
class CausalLMHead(nn.Module):
|
593 |
-
"""Causal Language Modeling head.
|
594 |
-
|
595 |
-
Reference:
|
596 |
-
Improving Language Understanding by Generative Pre-Training.
|
597 |
-
https://cdn.openai.com/research-covers/language-unsupervised/language_understanding_paper.pdf.
|
598 |
-
|
599 |
-
"""
|
600 |
-
|
601 |
-
def __init__(self, config: PretrainedConfig) -> None:
|
602 |
-
super().__init__()
|
603 |
-
|
604 |
-
self.ln = nn.LayerNorm(config.n_embd, eps=config.layer_norm_epsilon)
|
605 |
-
self.linear = nn.Linear(config.n_embd, config.vocab_size)
|
606 |
-
|
607 |
-
def forward(self, hidden_states: torch.FloatTensor) -> torch.FloatTensor:
|
608 |
-
hidden_states = self.ln(hidden_states)
|
609 |
-
logits = self.linear(hidden_states).to(torch.float32)
|
610 |
-
|
611 |
-
return logits
|
612 |
-
|
613 |
-
|
614 |
-
class CausalLMLoss(nn.Module):
|
615 |
-
"""Causal Language Modeling loss.
|
616 |
-
|
617 |
-
Reference:
|
618 |
-
Improving Language Understanding by Generative Pre-Training.
|
619 |
-
https://cdn.openai.com/research-covers/language-unsupervised/language_understanding_paper.pdf.
|
620 |
-
|
621 |
-
"""
|
622 |
-
|
623 |
-
def __init__(self, shift_labels: bool = True) -> None:
|
624 |
-
super().__init__()
|
625 |
-
|
626 |
-
self.shift_labels = shift_labels
|
627 |
-
self.loss_fct = nn.CrossEntropyLoss()
|
628 |
-
|
629 |
-
def forward(self, logits: torch.FloatTensor, labels: torch.LongTensor) -> torch.FloatTensor:
|
630 |
-
if self.shift_labels:
|
631 |
-
logits = logits[..., :-1, :].contiguous()
|
632 |
-
labels = labels[..., 1:].contiguous()
|
633 |
-
|
634 |
-
loss = self.loss_fct(logits.view(-1, logits.size(-1)), labels.view(-1))
|
635 |
-
|
636 |
-
return loss
|
637 |
-
|
638 |
-
|
639 |
-
class PhiPreTrainedModel(PreTrainedModel):
|
640 |
-
"""Phi pre-trained model."""
|
641 |
-
|
642 |
-
config_class = PhiConfig
|
643 |
-
base_model_prefix = "transformer"
|
644 |
-
supports_gradient_checkpointing = False
|
645 |
-
_no_split_modules = ["ParallelBlock"]
|
646 |
-
|
647 |
-
def __init__(self, *inputs, **kwargs) -> None:
|
648 |
-
super().__init__(*inputs, **kwargs)
|
649 |
-
|
650 |
-
def _init_weights(self, module: nn.Module) -> None:
|
651 |
-
if isinstance(module, (nn.Linear,)):
|
652 |
-
module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
|
653 |
-
if module.bias is not None:
|
654 |
-
module.bias.data.zero_()
|
655 |
-
elif isinstance(module, nn.Embedding):
|
656 |
-
module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
|
657 |
-
if module.padding_idx is not None:
|
658 |
-
module.weight.data[module.padding_idx].zero_()
|
659 |
-
elif isinstance(module, nn.LayerNorm):
|
660 |
-
if module.bias is not None:
|
661 |
-
module.bias.data.zero_()
|
662 |
-
module.weight.data.fill_(1.0)
|
663 |
-
|
664 |
-
def prepare_inputs_for_generation(
|
665 |
-
self,
|
666 |
-
input_ids: torch.LongTensor,
|
667 |
-
past_key_values: Optional[Union[torch.FloatTensor, InferenceParams]] = None,
|
668 |
-
attention_mask: Optional[Union[torch.LongTensor, torch.BoolTensor]] = None,
|
669 |
-
**kwargs,
|
670 |
-
) -> Dict[str, Any]:
|
671 |
-
if past_key_values is None or not (isinstance(past_key_values, InferenceParams)):
|
672 |
-
past_key_values = InferenceParams(
|
673 |
-
max_seqlen=self.config.n_positions,
|
674 |
-
max_batch_size=input_ids.shape[0],
|
675 |
-
seqlen_offset=0,
|
676 |
-
batch_size_offset=0,
|
677 |
-
key_value_memory_dict={},
|
678 |
-
lengths_per_sample=None,
|
679 |
-
)
|
680 |
-
else:
|
681 |
-
# Assume that `past_key_values` has cached all tokens up to the last token in `input_ids`
|
682 |
-
past_key_values.seqlen_offset = input_ids.shape[1] - 1
|
683 |
-
input_ids = input_ids[:, -1].unsqueeze(-1)
|
684 |
-
|
685 |
-
return {
|
686 |
-
"input_ids": input_ids,
|
687 |
-
"past_key_values": past_key_values,
|
688 |
-
"attention_mask": attention_mask,
|
689 |
-
}
|
690 |
-
|
691 |
-
|
692 |
-
class PhiModel(PhiPreTrainedModel):
|
693 |
-
"""Phi model."""
|
694 |
-
|
695 |
-
_keys_to_ignore_on_load_missing = [""]
|
696 |
-
_keys_to_ignore_on_load_unexpected = [r"h\.\d+\.mlp.(fc_in|fc_out)\.(weight|bias)"]
|
697 |
-
|
698 |
-
def __init__(self, config: PhiConfig) -> None:
|
699 |
-
super().__init__(config)
|
700 |
-
|
701 |
-
self.embd = Embedding(config)
|
702 |
-
self.h = nn.ModuleList([ParallelBlock(config, block_idx=i) for i in range(config.n_layer)])
|
703 |
-
self.gradient_checkpointing = False
|
704 |
-
self.post_init()
|
705 |
-
|
706 |
-
def get_input_embeddings(self) -> nn.Embedding:
|
707 |
-
return self.embd.wte
|
708 |
-
|
709 |
-
def set_input_embeddings(self, new_embeddings: nn.Embedding) -> None:
|
710 |
-
self.embd.wte = new_embeddings
|
711 |
-
|
712 |
-
def forward(
|
713 |
-
self,
|
714 |
-
input_ids: torch.LongTensor,
|
715 |
-
past_key_values: Optional[Union[torch.FloatTensor, InferenceParams]] = None,
|
716 |
-
attention_mask: Optional[torch.BoolTensor] = None,
|
717 |
-
) -> torch.FloatTensor:
|
718 |
-
hidden_states = self.embd(input_ids)
|
719 |
-
|
720 |
-
for layer in self.h:
|
721 |
-
hidden_states = layer(
|
722 |
-
hidden_states,
|
723 |
-
past_key_values=past_key_values,
|
724 |
-
attention_mask=attention_mask,
|
725 |
-
)
|
726 |
-
|
727 |
-
return hidden_states
|
728 |
-
|
729 |
-
|
730 |
-
class PhiForCausalLM(PhiPreTrainedModel):
|
731 |
-
"""Phi for Causal Language Modeling."""
|
732 |
-
|
733 |
-
_keys_to_ignore_on_load_missing = [""]
|
734 |
-
_keys_to_ignore_on_load_unexpected = [r"transformer\.h\.\d+\.mlp.(fc_in|fc_out)\.(weight|bias)"]
|
735 |
-
|
736 |
-
def __init__(self, config: PhiConfig) -> None:
|
737 |
-
super().__init__(config)
|
738 |
-
|
739 |
-
self.transformer = PhiModel(config)
|
740 |
-
self.lm_head = CausalLMHead(config)
|
741 |
-
self.loss = CausalLMLoss()
|
742 |
-
|
743 |
-
self.post_init()
|
744 |
-
|
745 |
-
def get_output_embeddings(self) -> nn.Linear:
|
746 |
-
return self.lm_head.linear
|
747 |
-
|
748 |
-
def set_output_embeddings(self, new_embeddings: nn.Linear) -> None:
|
749 |
-
self.lm_head.linear = new_embeddings
|
750 |
-
|
751 |
-
def forward(
|
752 |
-
self,
|
753 |
-
input_ids: torch.LongTensor,
|
754 |
-
past_key_values: Optional[Union[torch.FloatTensor, InferenceParams]] = None,
|
755 |
-
attention_mask: Optional[torch.BoolTensor] = None,
|
756 |
-
labels: Optional[torch.LongTensor] = None,
|
757 |
-
**kwargs,
|
758 |
-
) -> CausalLMOutputWithPast:
|
759 |
-
hidden_states = self.transformer(input_ids, past_key_values=past_key_values, attention_mask=attention_mask)
|
760 |
-
lm_logits = self.lm_head(hidden_states)
|
761 |
-
|
762 |
-
loss = None
|
763 |
-
if labels is not None:
|
764 |
-
loss = self.loss(lm_logits, labels)
|
765 |
-
|
766 |
-
return CausalLMOutputWithPast(loss=loss, logits=lm_logits, past_key_values=past_key_values)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
phi2_configuration.py
ADDED
@@ -0,0 +1,60 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import math
|
2 |
+
from transformers import PretrainedConfig
|
3 |
+
|
4 |
+
|
5 |
+
class Phi2Config(PretrainedConfig):
|
6 |
+
model_type = "phi2" # not necessary unless you want to register model with auto classes
|
7 |
+
attribute_map = {
|
8 |
+
"max_position_embeddings": "initial_cos_sin_cache_len",
|
9 |
+
"hidden_size": "d_embedding",
|
10 |
+
"num_attention_heads": "n_attn_heads",
|
11 |
+
"num_hidden_layers": "n_blocks",
|
12 |
+
}
|
13 |
+
|
14 |
+
def __init__(
|
15 |
+
self,
|
16 |
+
vocab_size: int = 50295, # this includes the extra tokens included by Phi2 in tokenizer_config.json
|
17 |
+
vocab_chunk_for_gpu_efficiency: int = 64,
|
18 |
+
initial_cos_sin_cache_len: int = 2048,
|
19 |
+
d_embedding: int = 1024, # 2560?
|
20 |
+
n_blocks: int = 20, # 32?
|
21 |
+
n_attn_heads: int = 16, # 32?
|
22 |
+
use_flash_attn: bool = False,
|
23 |
+
use_flash_rotary: bool = False,
|
24 |
+
use_fused_dense: bool = False,
|
25 |
+
attn_pdrop: float = 0.0,
|
26 |
+
embd_pdrop: float = 0.0,
|
27 |
+
resid_pdrop: float = 0.0,
|
28 |
+
layer_norm_epsilon: float = 1e-5,
|
29 |
+
weight_initialization_range: float = 0.02,
|
30 |
+
tie_word_embeddings: bool = False, # whether embedding weights are shared between the encoder and decoder
|
31 |
+
checkpointing: bool = False, # whether to use gradient checkpointing to reduce memory usage (I think)
|
32 |
+
**kwargs
|
33 |
+
) -> None:
|
34 |
+
self.vocab_size = (
|
35 |
+
math.ceil(
|
36 |
+
vocab_size / vocab_chunk_for_gpu_efficiency
|
37 |
+
) * vocab_chunk_for_gpu_efficiency
|
38 |
+
)
|
39 |
+
self.initial_cos_sin_cache_len = initial_cos_sin_cache_len
|
40 |
+
self.d_embedding = d_embedding
|
41 |
+
self.n_blocks = n_blocks
|
42 |
+
self.n_attn_heads = n_attn_heads
|
43 |
+
self.use_flash_attn = use_flash_attn
|
44 |
+
self.use_flash_rotary = use_flash_rotary
|
45 |
+
self.use_fused_dense = use_fused_dense
|
46 |
+
self.attn_pdrop = attn_pdrop
|
47 |
+
self.embd_pdrop = embd_pdrop
|
48 |
+
self.resid_pdrop = resid_pdrop
|
49 |
+
self.layer_norm_epsilon = layer_norm_epsilon
|
50 |
+
self.weight_initialization_range = weight_initialization_range
|
51 |
+
self.checkpointing = checkpointing
|
52 |
+
|
53 |
+
super().__init__(tie_word_embeddings=tie_word_embeddings, **kwargs)
|
54 |
+
|
55 |
+
|
56 |
+
if __name__ == "__main__":
|
57 |
+
phi2_config = Phi2Config()
|
58 |
+
# phi2_config.save_pretrained("phi2_config")
|
59 |
+
# phi2_config = Phi2Config.from_pretrained("phi2_config")
|
60 |
+
# phi2_config.push_to_hub("phi2_config")
|
phi2_model.py
ADDED
@@ -0,0 +1,166 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import torch
|
2 |
+
import torch.nn as nn
|
3 |
+
from transformers import PreTrainedModel
|
4 |
+
from transformers.modeling_outputs import CausalLMOutputWithPast
|
5 |
+
from typing import Any, cast
|
6 |
+
|
7 |
+
from .attention import ParallelAttentionBlock, KVCache
|
8 |
+
from .phi2_configuration import Phi2Config
|
9 |
+
|
10 |
+
|
11 |
+
class Phi2PreTrainedModel(PreTrainedModel):
|
12 |
+
config_class = Phi2Config # not necessary unless you want to register model with auto classes
|
13 |
+
supports_gradient_checkpointing = False
|
14 |
+
# _no_split_modules = ["ParallelAttentionBlock"]
|
15 |
+
|
16 |
+
# weight loading
|
17 |
+
# base_model_prefix = "transformer"
|
18 |
+
# _keys_to_ignore_on_load_missing = [""]
|
19 |
+
# _keys_to_ignore_on_load_unexpected = [r"h\.\d+\.mlp.(fc_in|fc_out)\.(weight|bias)"]
|
20 |
+
|
21 |
+
def __init__(self, config: Phi2Config):
|
22 |
+
super().__init__(config)
|
23 |
+
self.config = config
|
24 |
+
|
25 |
+
def _init_weights(self, module: nn.Module) -> None:
|
26 |
+
# initialize weights - will get overwritten by saved weights in from_pretrained() if they exist
|
27 |
+
if isinstance(module, (nn.Linear,)):
|
28 |
+
module.weight.data.normal_(mean=0.0, std=self.config.weight_initialization_range)
|
29 |
+
if module.bias is not None:
|
30 |
+
module.bias.data.zero_()
|
31 |
+
elif isinstance(module, nn.Embedding):
|
32 |
+
module.weight.data.normal_(mean=0.0, std=self.config.weight_initialization_range)
|
33 |
+
if module.padding_idx is not None:
|
34 |
+
module.weight.data[module.padding_idx].zero_()
|
35 |
+
elif isinstance(module, nn.LayerNorm):
|
36 |
+
if module.bias is not None:
|
37 |
+
module.bias.data.zero_()
|
38 |
+
module.weight.data.fill_(1.0)
|
39 |
+
|
40 |
+
def prepare_inputs_for_generation(
|
41 |
+
self,
|
42 |
+
input_ids: torch.LongTensor, # dim: (batch_size, seq_len)
|
43 |
+
kv_cache: KVCache | None = None,
|
44 |
+
key_padding_mask: torch.LongTensor | torch.BoolTensor | None = None,
|
45 |
+
) -> dict[str, Any]:
|
46 |
+
if not kv_cache:
|
47 |
+
kv_cache = KVCache(
|
48 |
+
max_seqlen=self.config.initial_cos_sin_cache_len,
|
49 |
+
max_batch_size=input_ids.shape[0],
|
50 |
+
seqlen_offset=0,
|
51 |
+
batch_size_offset=0,
|
52 |
+
kv_block_map={},
|
53 |
+
lengths_per_sample=None,
|
54 |
+
)
|
55 |
+
else:
|
56 |
+
# assume that `kv_cache` has cached all tokens up to the last token in `input_ids`
|
57 |
+
kv_cache.seqlen_offset = input_ids.shape[1] - 1
|
58 |
+
input_ids = cast(torch.LongTensor, input_ids[:, -1].unsqueeze(-1))
|
59 |
+
|
60 |
+
return { # to be passed to forward()
|
61 |
+
"input_ids": input_ids,
|
62 |
+
"kv_cache": kv_cache,
|
63 |
+
"key_padding_mask": key_padding_mask,
|
64 |
+
}
|
65 |
+
|
66 |
+
|
67 |
+
class Embedding(nn.Module):
|
68 |
+
"""Token embedding with dropout from Phi2."""
|
69 |
+
|
70 |
+
def __init__(
|
71 |
+
self,
|
72 |
+
vocab_size: int,
|
73 |
+
d_embedding: int,
|
74 |
+
embd_pdrop: float,
|
75 |
+
) -> None:
|
76 |
+
super().__init__()
|
77 |
+
self.embeddings = nn.Embedding(vocab_size, d_embedding)
|
78 |
+
self.dropout = nn.Dropout(embd_pdrop)
|
79 |
+
|
80 |
+
def forward(
|
81 |
+
self,
|
82 |
+
input_ids: torch.LongTensor, # dim: (batch_size, seq_len)
|
83 |
+
) -> torch.FloatTensor:
|
84 |
+
x = self.embeddings( # dim: (batch_size, seq_len, d_embedding)
|
85 |
+
input_ids.view(-1, input_ids.size()[-1])
|
86 |
+
)
|
87 |
+
x = self.dropout(x)
|
88 |
+
return x
|
89 |
+
|
90 |
+
|
91 |
+
class Phi2Model(Phi2PreTrainedModel):
|
92 |
+
def __init__(self, config: Phi2Config) -> None:
|
93 |
+
super().__init__(config)
|
94 |
+
self.embedding = Embedding(
|
95 |
+
vocab_size=config.vocab_size,
|
96 |
+
d_embedding=config.d_embedding,
|
97 |
+
embd_pdrop=config.embd_pdrop,
|
98 |
+
)
|
99 |
+
self.parallel_blocks = nn.ModuleList([
|
100 |
+
ParallelAttentionBlock(
|
101 |
+
resid_pdrop=config.resid_pdrop,
|
102 |
+
layer_norm_epsilon=config.layer_norm_epsilon,
|
103 |
+
d_embedding=config.d_embedding,
|
104 |
+
n_attn_heads=config.n_attn_heads,
|
105 |
+
block_n=i,
|
106 |
+
initial_cos_sin_cache_len=config.initial_cos_sin_cache_len,
|
107 |
+
attn_pdrop=config.attn_pdrop,
|
108 |
+
use_flash_rotary=config.use_flash_rotary,
|
109 |
+
use_flash_attn=config.use_flash_attn,
|
110 |
+
use_fused_dense=config.use_fused_dense,
|
111 |
+
checkpointing=config.checkpointing,
|
112 |
+
)
|
113 |
+
for i in range(config.n_blocks)
|
114 |
+
])
|
115 |
+
self.gradient_checkpointing_disable() # https://github.com/cybertronai/gradient-checkpointing - I think this is turned off due to flash attention?
|
116 |
+
self.post_init() # calls self._init_weights() for all modules
|
117 |
+
|
118 |
+
"""
|
119 |
+
def get_input_embeddings(self) -> nn.Embedding:
|
120 |
+
return self.embedding.embeddings
|
121 |
+
|
122 |
+
def set_input_embeddings(self, new_embeddings: nn.Embedding) -> None:
|
123 |
+
self.embedding.embeddings = new_embeddings
|
124 |
+
"""
|
125 |
+
|
126 |
+
def forward(
|
127 |
+
self,
|
128 |
+
input_ids: torch.LongTensor,
|
129 |
+
kv_cache: KVCache | None = None,
|
130 |
+
key_padding_mask: torch.BoolTensor | None = None,
|
131 |
+
) -> torch.FloatTensor:
|
132 |
+
x = self.embedding(input_ids)
|
133 |
+
for block in self.parallel_blocks:
|
134 |
+
x = block(
|
135 |
+
x,
|
136 |
+
kv_cache=kv_cache,
|
137 |
+
key_padding_mask=key_padding_mask,
|
138 |
+
)
|
139 |
+
return x
|
140 |
+
|
141 |
+
|
142 |
+
class Phi2ModelForCausalLM(Phi2PreTrainedModel):
|
143 |
+
def __init__(self, config: Phi2Config) -> None:
|
144 |
+
super().__init__(config)
|
145 |
+
self.pretrained_model = Phi2Model(config)
|
146 |
+
self.layer_norm = nn.LayerNorm(config.d_embedding, eps=config.layer_norm_epsilon)
|
147 |
+
self.linear = nn.Linear(config.d_embedding, config.vocab_size)
|
148 |
+
self.loss_fn = nn.CrossEntropyLoss()
|
149 |
+
self.post_init() # calls self._init_weights() for all modules
|
150 |
+
|
151 |
+
def forward(
|
152 |
+
self,
|
153 |
+
input_ids: torch.LongTensor,
|
154 |
+
kv_cache: KVCache | None = None,
|
155 |
+
key_padding_mask: torch.BoolTensor | None = None,
|
156 |
+
labels: torch.LongTensor | None = None,
|
157 |
+
) -> CausalLMOutputWithPast:
|
158 |
+
x = self.pretrained_model(input_ids, kv_cache=kv_cache, key_padding_mask=key_padding_mask)
|
159 |
+
x = self.layer_norm(x)
|
160 |
+
logits = self.linear(x).to(torch.float32)
|
161 |
+
loss = (
|
162 |
+
self.loss_fn(logits.view(-1, logits.size(-1)), labels.view(-1))
|
163 |
+
if labels is not None
|
164 |
+
else None
|
165 |
+
)
|
166 |
+
return CausalLMOutputWithPast(loss=loss, logits=logits)
|