reymondzzzz
commited on
Commit
•
680dd84
1
Parent(s):
1c551a8
Upload GPTRefactForCausalLM
Browse files- config.json +4 -6
- configuration_gpt_refact.py +5 -9
- generation_config.json +2 -2
- modeling_gpt_refact.py +53 -119
- pytorch_model.bin +2 -2
config.json
CHANGED
@@ -1,29 +1,27 @@
|
|
1 |
{
|
2 |
-
"activation_function": "gelu_new",
|
3 |
"architectures": [
|
4 |
"GPTRefactForCausalLM"
|
5 |
],
|
6 |
-
"attention_softmax_in_fp32":
|
7 |
"attn_pdrop": 0.1,
|
8 |
"auto_map": {
|
9 |
"AutoConfig": "configuration_gpt_refact.GPTRefactConfig",
|
10 |
"AutoModelForCausalLM": "modeling_gpt_refact.GPTRefactForCausalLM"
|
11 |
},
|
12 |
-
"bos_token_id":
|
13 |
"do_sample": true,
|
14 |
"embd_pdrop": 0.1,
|
15 |
-
"eos_token_id":
|
16 |
"initializer_range": 0.02,
|
17 |
"layer_norm_epsilon": 1e-05,
|
18 |
"model_type": "gpt_refact",
|
19 |
-
"multi_query": true,
|
20 |
"n_embd": 2048,
|
21 |
"n_head": 32,
|
22 |
"n_inner": null,
|
23 |
"n_layer": 32,
|
24 |
"n_positions": 1024,
|
25 |
"resid_pdrop": 0.1,
|
26 |
-
"scale_attention_softmax_in_fp32":
|
27 |
"scale_attn_weights": true,
|
28 |
"torch_dtype": "float32",
|
29 |
"transformers_version": "4.28.1",
|
|
|
1 |
{
|
|
|
2 |
"architectures": [
|
3 |
"GPTRefactForCausalLM"
|
4 |
],
|
5 |
+
"attention_softmax_in_fp32": false,
|
6 |
"attn_pdrop": 0.1,
|
7 |
"auto_map": {
|
8 |
"AutoConfig": "configuration_gpt_refact.GPTRefactConfig",
|
9 |
"AutoModelForCausalLM": "modeling_gpt_refact.GPTRefactForCausalLM"
|
10 |
},
|
11 |
+
"bos_token_id": 0,
|
12 |
"do_sample": true,
|
13 |
"embd_pdrop": 0.1,
|
14 |
+
"eos_token_id": 0,
|
15 |
"initializer_range": 0.02,
|
16 |
"layer_norm_epsilon": 1e-05,
|
17 |
"model_type": "gpt_refact",
|
|
|
18 |
"n_embd": 2048,
|
19 |
"n_head": 32,
|
20 |
"n_inner": null,
|
21 |
"n_layer": 32,
|
22 |
"n_positions": 1024,
|
23 |
"resid_pdrop": 0.1,
|
24 |
+
"scale_attention_softmax_in_fp32": false,
|
25 |
"scale_attn_weights": true,
|
26 |
"torch_dtype": "float32",
|
27 |
"transformers_version": "4.28.1",
|
configuration_gpt_refact.py
CHANGED
@@ -17,13 +17,12 @@ class GPTRefactConfig(PretrainedConfig):
|
|
17 |
|
18 |
def __init__(
|
19 |
self,
|
20 |
-
vocab_size=
|
21 |
n_positions=1024,
|
22 |
n_embd=768,
|
23 |
n_layer=12,
|
24 |
n_head=12,
|
25 |
n_inner=None,
|
26 |
-
activation_function="gelu_new",
|
27 |
resid_pdrop=0.1,
|
28 |
embd_pdrop=0.1,
|
29 |
attn_pdrop=0.1,
|
@@ -31,11 +30,10 @@ class GPTRefactConfig(PretrainedConfig):
|
|
31 |
initializer_range=0.02,
|
32 |
scale_attn_weights=True,
|
33 |
use_cache=True,
|
34 |
-
bos_token_id=
|
35 |
-
eos_token_id=
|
36 |
-
attention_softmax_in_fp32=
|
37 |
-
scale_attention_softmax_in_fp32=
|
38 |
-
multi_query=True,
|
39 |
**kwargs,
|
40 |
):
|
41 |
self.vocab_size = vocab_size
|
@@ -44,7 +42,6 @@ class GPTRefactConfig(PretrainedConfig):
|
|
44 |
self.n_layer = n_layer
|
45 |
self.n_head = n_head
|
46 |
self.n_inner = n_inner
|
47 |
-
self.activation_function = activation_function
|
48 |
self.resid_pdrop = resid_pdrop
|
49 |
self.embd_pdrop = embd_pdrop
|
50 |
self.attn_pdrop = attn_pdrop
|
@@ -54,7 +51,6 @@ class GPTRefactConfig(PretrainedConfig):
|
|
54 |
self.use_cache = use_cache
|
55 |
self.attention_softmax_in_fp32 = attention_softmax_in_fp32
|
56 |
self.scale_attention_softmax_in_fp32 = scale_attention_softmax_in_fp32
|
57 |
-
self.multi_query = multi_query
|
58 |
|
59 |
self.bos_token_id = bos_token_id
|
60 |
self.eos_token_id = eos_token_id
|
|
|
17 |
|
18 |
def __init__(
|
19 |
self,
|
20 |
+
vocab_size=49216,
|
21 |
n_positions=1024,
|
22 |
n_embd=768,
|
23 |
n_layer=12,
|
24 |
n_head=12,
|
25 |
n_inner=None,
|
|
|
26 |
resid_pdrop=0.1,
|
27 |
embd_pdrop=0.1,
|
28 |
attn_pdrop=0.1,
|
|
|
30 |
initializer_range=0.02,
|
31 |
scale_attn_weights=True,
|
32 |
use_cache=True,
|
33 |
+
bos_token_id=0,
|
34 |
+
eos_token_id=0,
|
35 |
+
attention_softmax_in_fp32=False,
|
36 |
+
scale_attention_softmax_in_fp32=False,
|
|
|
37 |
**kwargs,
|
38 |
):
|
39 |
self.vocab_size = vocab_size
|
|
|
42 |
self.n_layer = n_layer
|
43 |
self.n_head = n_head
|
44 |
self.n_inner = n_inner
|
|
|
45 |
self.resid_pdrop = resid_pdrop
|
46 |
self.embd_pdrop = embd_pdrop
|
47 |
self.attn_pdrop = attn_pdrop
|
|
|
51 |
self.use_cache = use_cache
|
52 |
self.attention_softmax_in_fp32 = attention_softmax_in_fp32
|
53 |
self.scale_attention_softmax_in_fp32 = scale_attention_softmax_in_fp32
|
|
|
54 |
|
55 |
self.bos_token_id = bos_token_id
|
56 |
self.eos_token_id = eos_token_id
|
generation_config.json
CHANGED
@@ -1,7 +1,7 @@
|
|
1 |
{
|
2 |
"_from_model_config": true,
|
3 |
-
"bos_token_id":
|
4 |
"do_sample": true,
|
5 |
-
"eos_token_id":
|
6 |
"transformers_version": "4.28.1"
|
7 |
}
|
|
|
1 |
{
|
2 |
"_from_model_config": true,
|
3 |
+
"bos_token_id": 0,
|
4 |
"do_sample": true,
|
5 |
+
"eos_token_id": 0,
|
6 |
"transformers_version": "4.28.1"
|
7 |
}
|
modeling_gpt_refact.py
CHANGED
@@ -1,39 +1,27 @@
|
|
1 |
import math
|
2 |
-
from typing import List, Optional, Tuple, Union
|
3 |
-
|
4 |
import torch
|
|
|
5 |
import torch.utils.checkpoint
|
6 |
from torch import nn
|
7 |
-
|
8 |
-
from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
|
9 |
-
|
10 |
-
from transformers.activations import ACT2FN
|
11 |
from transformers.modeling_outputs import (
|
12 |
BaseModelOutputWithPastAndCrossAttentions,
|
13 |
CausalLMOutputWithCrossAttentions,
|
14 |
-
SequenceClassifierOutputWithPast,
|
15 |
-
TokenClassifierOutput,
|
16 |
)
|
17 |
from transformers.modeling_utils import PreTrainedModel
|
18 |
from transformers.utils import (
|
19 |
-
add_code_sample_docstrings,
|
20 |
-
add_start_docstrings,
|
21 |
-
add_start_docstrings_to_model_forward,
|
22 |
logging,
|
23 |
)
|
24 |
-
from
|
25 |
|
|
|
26 |
|
27 |
logger = logging.get_logger(__name__)
|
28 |
|
29 |
|
30 |
-
# Fused kernels
|
31 |
-
# Use separate functions for each case because conditionals prevent kernel fusion.
|
32 |
-
# TODO: Could have better fused kernels depending on scaling, dropout and head mask.
|
33 |
-
# Is it doable without writing 32 functions?
|
34 |
@torch.jit.script
|
35 |
def upcast_masked_softmax(
|
36 |
-
|
37 |
):
|
38 |
input_dtype = x.dtype
|
39 |
x = x.to(softmax_dtype) * scale
|
@@ -56,7 +44,8 @@ def masked_softmax(x: torch.Tensor, mask: torch.Tensor, mask_value: torch.Tensor
|
|
56 |
x = torch.nn.functional.softmax(x, dim=-1)
|
57 |
return x
|
58 |
|
59 |
-
|
|
|
60 |
"""
|
61 |
## Get head-specific slope $m$ for each head
|
62 |
* `n_heads` is the number of heads in the attention layer $n$
|
@@ -70,7 +59,7 @@ def _get_slopes(attn_heads: int, dev: str) -> torch.Tensor:
|
|
70 |
# Get the closest power of 2 to `n_heads`.
|
71 |
# If `n_heads` is not a power of 2, then we first calculate slopes to the closest (smaller) power of 2,
|
72 |
# and then add the remaining slopes.
|
73 |
-
n = 2 ** math.floor(math.
|
74 |
# $2^{-\frac{8}{n}}$
|
75 |
m_0 = 2.0 ** (-8.0 / n)
|
76 |
# $2^{-1\frac{8}{n}}, 2^{-2 \frac{8}{n}}, 2^{-3 \frac{8}{n}}, \dots$
|
@@ -90,13 +79,13 @@ def _get_slopes(attn_heads: int, dev: str) -> torch.Tensor:
|
|
90 |
|
91 |
return m
|
92 |
|
93 |
-
|
94 |
def get_alibi_biases(
|
95 |
B: int,
|
96 |
T: int,
|
97 |
attn_heads: int,
|
98 |
-
dev:
|
99 |
-
dtype,
|
100 |
causal: bool = True) -> torch.Tensor:
|
101 |
"""
|
102 |
## Calculate the attention biases matrix
|
@@ -126,12 +115,12 @@ def get_alibi_biases(
|
|
126 |
biases = biases.repeat(B, 1, 1, 1)
|
127 |
return biases.to(dtype).contiguous()
|
128 |
|
|
|
129 |
class Attention(nn.Module):
|
130 |
def __init__(self, config, layer_idx=None):
|
131 |
super().__init__()
|
132 |
self.mask_value = None
|
133 |
|
134 |
-
self.multi_query = config.multi_query
|
135 |
self.embed_dim = config.hidden_size
|
136 |
self.num_heads = config.num_attention_heads
|
137 |
self.head_dim = self.embed_dim // self.num_heads
|
@@ -148,7 +137,7 @@ class Attention(nn.Module):
|
|
148 |
self.layer_idx = layer_idx
|
149 |
self.attention_softmax_in_fp32 = config.attention_softmax_in_fp32
|
150 |
self.scale_attention_softmax_in_fp32 = (
|
151 |
-
|
152 |
)
|
153 |
|
154 |
self.q = nn.Linear(self.embed_dim, self.embed_dim, bias=False)
|
@@ -162,13 +151,9 @@ class Attention(nn.Module):
|
|
162 |
upcast = dtype != softmax_dtype
|
163 |
unscale = self.layer_idx + 1 if self.scale_attention_softmax_in_fp32 and upcast else 1
|
164 |
|
165 |
-
# MQA models: (batch_size, query_length, num_heads * head_dim)
|
166 |
-
# MHA models: (batch_size, num_heads, query_length, head_dim)
|
167 |
attn_weights = alibi + torch.matmul(query * self.scale, key)
|
168 |
|
169 |
if upcast:
|
170 |
-
# Use a fused kernel to prevent a large overhead from casting and scaling.
|
171 |
-
# Sub-optimal when the key length is not a multiple of 8.
|
172 |
if attention_mask is None:
|
173 |
attn_weights = upcast_softmax(attn_weights, unscale, softmax_dtype)
|
174 |
else:
|
@@ -176,8 +161,6 @@ class Attention(nn.Module):
|
|
176 |
attn_weights = upcast_masked_softmax(attn_weights, attention_mask, mask_value, unscale, softmax_dtype)
|
177 |
else:
|
178 |
if attention_mask is not None:
|
179 |
-
|
180 |
-
# The fused kernel is very slow when the key length is not a multiple of 8, so we skip fusion.
|
181 |
attn_weights = torch.masked_fill(attn_weights, attention_mask, -10000)
|
182 |
|
183 |
attn_weights = torch.nn.functional.softmax(attn_weights, dim=-1)
|
@@ -192,15 +175,13 @@ class Attention(nn.Module):
|
|
192 |
return tensor.permute(0, 2, 1, 3)
|
193 |
|
194 |
def forward(
|
195 |
-
|
196 |
-
|
197 |
-
|
198 |
-
|
199 |
-
|
200 |
-
|
201 |
-
|
202 |
-
use_cache: Optional[bool] = False,
|
203 |
-
output_attentions: Optional[bool] = False,
|
204 |
) -> Union[
|
205 |
Tuple[torch.Tensor, Optional[torch.Tensor]],
|
206 |
Tuple[torch.Tensor, Optional[torch.Tensor], Tuple[torch.Tensor, ...]],
|
@@ -264,6 +245,7 @@ class LayerNormNoBias(nn.Module):
|
|
264 |
def forward(self, x: torch.Tensor) -> torch.Tensor:
|
265 |
return F.layer_norm(x, self.shape, self.weight, None, self.eps)
|
266 |
|
|
|
267 |
class GPTRefactBlock(nn.Module):
|
268 |
def __init__(self, config, layer_idx=None):
|
269 |
super().__init__()
|
@@ -277,15 +259,13 @@ class GPTRefactBlock(nn.Module):
|
|
277 |
self.mlp = MLP(self.inner_dim, config)
|
278 |
|
279 |
def forward(
|
280 |
-
|
281 |
-
|
282 |
-
|
283 |
-
|
284 |
-
|
285 |
-
|
286 |
-
|
287 |
-
use_cache: Optional[bool] = False,
|
288 |
-
output_attentions: Optional[bool] = False,
|
289 |
) -> Union[
|
290 |
Tuple[torch.Tensor], Tuple[torch.Tensor, torch.Tensor], Tuple[torch.Tensor, torch.Tensor, torch.Tensor]
|
291 |
]:
|
@@ -317,11 +297,6 @@ class GPTRefactBlock(nn.Module):
|
|
317 |
|
318 |
|
319 |
class GPTRefactPreTrainedModel(PreTrainedModel):
|
320 |
-
"""
|
321 |
-
An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained
|
322 |
-
models.
|
323 |
-
"""
|
324 |
-
|
325 |
config_class = GPTRefactConfig
|
326 |
base_model_prefix = "transformer"
|
327 |
supports_gradient_checkpointing = True
|
@@ -332,7 +307,6 @@ class GPTRefactPreTrainedModel(PreTrainedModel):
|
|
332 |
super().__init__(*inputs, **kwargs)
|
333 |
|
334 |
def _init_weights(self, module):
|
335 |
-
"""Initialize the weights."""
|
336 |
if isinstance(module, (MLP, Attention)):
|
337 |
# Reinitialize selected weights subject to the OpenAI GPT-2 Paper Scheme:
|
338 |
# > A modified initialization which accounts for the accumulation on the residual path with model depth. Scale
|
@@ -354,8 +328,7 @@ class GPTRefactPreTrainedModel(PreTrainedModel):
|
|
354 |
module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
|
355 |
if module.padding_idx is not None:
|
356 |
module.weight.data[module.padding_idx].zero_()
|
357 |
-
elif isinstance(module,
|
358 |
-
module.bias.data.zero_()
|
359 |
module.weight.data.fill_(1.0)
|
360 |
|
361 |
def _set_gradient_checkpointing(self, module, value=False):
|
@@ -394,20 +367,15 @@ class GPTRefactModel(GPTRefactPreTrainedModel):
|
|
394 |
return mask
|
395 |
|
396 |
def forward(
|
397 |
-
|
398 |
-
|
399 |
-
|
400 |
-
|
401 |
-
|
402 |
-
|
403 |
-
|
404 |
-
|
405 |
-
|
406 |
-
encoder_attention_mask: Optional[torch.Tensor] = None,
|
407 |
-
use_cache: Optional[bool] = None,
|
408 |
-
output_attentions: Optional[bool] = None,
|
409 |
-
output_hidden_states: Optional[bool] = None,
|
410 |
-
return_dict: Optional[bool] = None,
|
411 |
) -> Union[Tuple, BaseModelOutputWithPastAndCrossAttentions]:
|
412 |
output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
|
413 |
output_hidden_states = (
|
@@ -433,27 +401,12 @@ class GPTRefactModel(GPTRefactPreTrainedModel):
|
|
433 |
|
434 |
device = input_ids.device if input_ids is not None else inputs_embeds.device
|
435 |
|
436 |
-
if token_type_ids is not None:
|
437 |
-
token_type_ids = token_type_ids.view(-1, input_shape[-1])
|
438 |
-
if position_ids is not None:
|
439 |
-
position_ids = position_ids.view(-1, input_shape[-1])
|
440 |
-
|
441 |
if past_key_values is None:
|
442 |
past_length = 0
|
443 |
past_key_values = tuple([None] * len(self.h))
|
444 |
else:
|
445 |
past_length = past_key_values[0][0].size(-2)
|
446 |
|
447 |
-
if attention_mask is not None and len(attention_mask.shape) == 2 and position_ids is None:
|
448 |
-
# create position_ids on the fly for batch generation
|
449 |
-
position_ids = attention_mask.long().cumsum(-1) - 1
|
450 |
-
position_ids.masked_fill_(attention_mask == 0, 1)
|
451 |
-
if past_length > 0:
|
452 |
-
position_ids = position_ids[:, past_length : input_shape[-1] + past_length :]
|
453 |
-
elif position_ids is None:
|
454 |
-
position_ids = torch.arange(past_length, input_shape[-1] + past_length, dtype=torch.long, device=device)
|
455 |
-
position_ids = position_ids.unsqueeze(0).view(-1, input_shape[-1])
|
456 |
-
|
457 |
# Self-attention mask.
|
458 |
query_length = input_shape[-1]
|
459 |
|
@@ -468,10 +421,6 @@ class GPTRefactModel(GPTRefactPreTrainedModel):
|
|
468 |
alibi = get_alibi_biases(hidden_states.shape[0], seq_length_with_past,
|
469 |
self.num_heads, device, self.wte.weight.dtype)[:, :, -query_length:, :]
|
470 |
|
471 |
-
if token_type_ids is not None:
|
472 |
-
token_type_embeds = self.wte(token_type_ids)
|
473 |
-
hidden_states = hidden_states + token_type_embeds
|
474 |
-
|
475 |
output_shape = input_shape + (hidden_states.size(-1),)
|
476 |
|
477 |
presents = [] if use_cache else None
|
@@ -496,9 +445,7 @@ class GPTRefactModel(GPTRefactPreTrainedModel):
|
|
496 |
hidden_states,
|
497 |
None,
|
498 |
attention_mask,
|
499 |
-
|
500 |
-
encoder_hidden_states,
|
501 |
-
encoder_attention_mask,
|
502 |
)
|
503 |
else:
|
504 |
outputs = block(
|
@@ -506,8 +453,6 @@ class GPTRefactModel(GPTRefactPreTrainedModel):
|
|
506 |
layer_past=layer_past,
|
507 |
attention_mask=attention_mask,
|
508 |
alibi=alibi,
|
509 |
-
encoder_hidden_states=encoder_hidden_states,
|
510 |
-
encoder_attention_mask=encoder_attention_mask,
|
511 |
use_cache=use_cache,
|
512 |
output_attentions=output_attentions,
|
513 |
)
|
@@ -541,21 +486,20 @@ class GPTRefactModel(GPTRefactPreTrainedModel):
|
|
541 |
cross_attentions=all_cross_attentions,
|
542 |
)
|
543 |
|
|
|
544 |
class GPTRefactForCausalLM(GPTRefactPreTrainedModel):
|
545 |
-
_tied_weights_keys = ["lm_head.weight"]
|
546 |
|
547 |
def __init__(self, config):
|
548 |
super().__init__(config)
|
549 |
self.transformer = GPTRefactModel(config)
|
550 |
-
self.ln_f =
|
551 |
self.lm_head = nn.Linear(config.n_embd, config.vocab_size, bias=False)
|
552 |
|
553 |
# Initialize weights and apply final processing
|
554 |
self.post_init()
|
555 |
|
556 |
def prepare_inputs_for_generation(self, input_ids, past_key_values=None, inputs_embeds=None, **kwargs):
|
557 |
-
|
558 |
-
# if `inputs_embeds` are passed, we only want to use them in the 1st generation step
|
559 |
if inputs_embeds is not None and past_key_values is None:
|
560 |
model_inputs = {"inputs_embeds": inputs_embeds}
|
561 |
else:
|
@@ -573,21 +517,16 @@ class GPTRefactForCausalLM(GPTRefactPreTrainedModel):
|
|
573 |
return model_inputs
|
574 |
|
575 |
def forward(
|
576 |
-
|
577 |
-
|
578 |
-
|
579 |
-
|
580 |
-
|
581 |
-
|
582 |
-
|
583 |
-
|
584 |
-
|
585 |
-
|
586 |
-
labels: Optional[torch.Tensor] = None,
|
587 |
-
use_cache: Optional[bool] = None,
|
588 |
-
output_attentions: Optional[bool] = None,
|
589 |
-
output_hidden_states: Optional[bool] = None,
|
590 |
-
return_dict: Optional[bool] = None,
|
591 |
) -> Union[Tuple, CausalLMOutputWithCrossAttentions]:
|
592 |
r"""
|
593 |
labels (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
|
@@ -601,12 +540,7 @@ class GPTRefactForCausalLM(GPTRefactPreTrainedModel):
|
|
601 |
input_ids,
|
602 |
past_key_values=past_key_values,
|
603 |
attention_mask=attention_mask,
|
604 |
-
token_type_ids=token_type_ids,
|
605 |
-
position_ids=position_ids,
|
606 |
-
head_mask=head_mask,
|
607 |
inputs_embeds=inputs_embeds,
|
608 |
-
encoder_hidden_states=encoder_hidden_states,
|
609 |
-
encoder_attention_mask=encoder_attention_mask,
|
610 |
use_cache=use_cache,
|
611 |
output_attentions=output_attentions,
|
612 |
output_hidden_states=output_hidden_states,
|
@@ -641,7 +575,7 @@ class GPTRefactForCausalLM(GPTRefactPreTrainedModel):
|
|
641 |
|
642 |
@staticmethod
|
643 |
def _reorder_cache(
|
644 |
-
|
645 |
) -> Tuple[Tuple[torch.Tensor]]:
|
646 |
"""
|
647 |
This function is used to re-order the `past_key_values` cache if [`~PreTrainedModel.beam_search`] or
|
|
|
1 |
import math
|
|
|
|
|
2 |
import torch
|
3 |
+
import torch.nn.functional as F
|
4 |
import torch.utils.checkpoint
|
5 |
from torch import nn
|
6 |
+
from torch.nn import CrossEntropyLoss
|
|
|
|
|
|
|
7 |
from transformers.modeling_outputs import (
|
8 |
BaseModelOutputWithPastAndCrossAttentions,
|
9 |
CausalLMOutputWithCrossAttentions,
|
|
|
|
|
10 |
)
|
11 |
from transformers.modeling_utils import PreTrainedModel
|
12 |
from transformers.utils import (
|
|
|
|
|
|
|
13 |
logging,
|
14 |
)
|
15 |
+
from typing import List, Optional, Tuple, Union
|
16 |
|
17 |
+
from hf.configuration_gpt_refact import GPTRefactConfig
|
18 |
|
19 |
logger = logging.get_logger(__name__)
|
20 |
|
21 |
|
|
|
|
|
|
|
|
|
22 |
@torch.jit.script
|
23 |
def upcast_masked_softmax(
|
24 |
+
x: torch.Tensor, mask: torch.Tensor, mask_value: torch.Tensor, scale: float, softmax_dtype: torch.dtype
|
25 |
):
|
26 |
input_dtype = x.dtype
|
27 |
x = x.to(softmax_dtype) * scale
|
|
|
44 |
x = torch.nn.functional.softmax(x, dim=-1)
|
45 |
return x
|
46 |
|
47 |
+
@torch.jit.script
|
48 |
+
def _get_slopes(attn_heads: int, dev: torch.device) -> torch.Tensor:
|
49 |
"""
|
50 |
## Get head-specific slope $m$ for each head
|
51 |
* `n_heads` is the number of heads in the attention layer $n$
|
|
|
59 |
# Get the closest power of 2 to `n_heads`.
|
60 |
# If `n_heads` is not a power of 2, then we first calculate slopes to the closest (smaller) power of 2,
|
61 |
# and then add the remaining slopes.
|
62 |
+
n = 2 ** math.floor(math.log(attn_heads, 2))
|
63 |
# $2^{-\frac{8}{n}}$
|
64 |
m_0 = 2.0 ** (-8.0 / n)
|
65 |
# $2^{-1\frac{8}{n}}, 2^{-2 \frac{8}{n}}, 2^{-3 \frac{8}{n}}, \dots$
|
|
|
79 |
|
80 |
return m
|
81 |
|
82 |
+
@torch.jit.script
|
83 |
def get_alibi_biases(
|
84 |
B: int,
|
85 |
T: int,
|
86 |
attn_heads: int,
|
87 |
+
dev: torch.device,
|
88 |
+
dtype: torch.dtype,
|
89 |
causal: bool = True) -> torch.Tensor:
|
90 |
"""
|
91 |
## Calculate the attention biases matrix
|
|
|
115 |
biases = biases.repeat(B, 1, 1, 1)
|
116 |
return biases.to(dtype).contiguous()
|
117 |
|
118 |
+
|
119 |
class Attention(nn.Module):
|
120 |
def __init__(self, config, layer_idx=None):
|
121 |
super().__init__()
|
122 |
self.mask_value = None
|
123 |
|
|
|
124 |
self.embed_dim = config.hidden_size
|
125 |
self.num_heads = config.num_attention_heads
|
126 |
self.head_dim = self.embed_dim // self.num_heads
|
|
|
137 |
self.layer_idx = layer_idx
|
138 |
self.attention_softmax_in_fp32 = config.attention_softmax_in_fp32
|
139 |
self.scale_attention_softmax_in_fp32 = (
|
140 |
+
config.scale_attention_softmax_in_fp32 and config.attention_softmax_in_fp32
|
141 |
)
|
142 |
|
143 |
self.q = nn.Linear(self.embed_dim, self.embed_dim, bias=False)
|
|
|
151 |
upcast = dtype != softmax_dtype
|
152 |
unscale = self.layer_idx + 1 if self.scale_attention_softmax_in_fp32 and upcast else 1
|
153 |
|
|
|
|
|
154 |
attn_weights = alibi + torch.matmul(query * self.scale, key)
|
155 |
|
156 |
if upcast:
|
|
|
|
|
157 |
if attention_mask is None:
|
158 |
attn_weights = upcast_softmax(attn_weights, unscale, softmax_dtype)
|
159 |
else:
|
|
|
161 |
attn_weights = upcast_masked_softmax(attn_weights, attention_mask, mask_value, unscale, softmax_dtype)
|
162 |
else:
|
163 |
if attention_mask is not None:
|
|
|
|
|
164 |
attn_weights = torch.masked_fill(attn_weights, attention_mask, -10000)
|
165 |
|
166 |
attn_weights = torch.nn.functional.softmax(attn_weights, dim=-1)
|
|
|
175 |
return tensor.permute(0, 2, 1, 3)
|
176 |
|
177 |
def forward(
|
178 |
+
self,
|
179 |
+
hidden_states: torch.Tensor,
|
180 |
+
layer_past: Optional[torch.Tensor] = None,
|
181 |
+
attention_mask: Optional[torch.Tensor] = None,
|
182 |
+
alibi: Optional[torch.Tensor] = None,
|
183 |
+
use_cache: Optional[bool] = False,
|
184 |
+
output_attentions: Optional[bool] = False,
|
|
|
|
|
185 |
) -> Union[
|
186 |
Tuple[torch.Tensor, Optional[torch.Tensor]],
|
187 |
Tuple[torch.Tensor, Optional[torch.Tensor], Tuple[torch.Tensor, ...]],
|
|
|
245 |
def forward(self, x: torch.Tensor) -> torch.Tensor:
|
246 |
return F.layer_norm(x, self.shape, self.weight, None, self.eps)
|
247 |
|
248 |
+
|
249 |
class GPTRefactBlock(nn.Module):
|
250 |
def __init__(self, config, layer_idx=None):
|
251 |
super().__init__()
|
|
|
259 |
self.mlp = MLP(self.inner_dim, config)
|
260 |
|
261 |
def forward(
|
262 |
+
self,
|
263 |
+
hidden_states: Optional[Tuple[torch.Tensor]],
|
264 |
+
layer_past: Optional[torch.Tensor] = None,
|
265 |
+
attention_mask: Optional[torch.Tensor] = None,
|
266 |
+
alibi: Optional[torch.Tensor] = None,
|
267 |
+
use_cache: Optional[bool] = False,
|
268 |
+
output_attentions: Optional[bool] = False,
|
|
|
|
|
269 |
) -> Union[
|
270 |
Tuple[torch.Tensor], Tuple[torch.Tensor, torch.Tensor], Tuple[torch.Tensor, torch.Tensor, torch.Tensor]
|
271 |
]:
|
|
|
297 |
|
298 |
|
299 |
class GPTRefactPreTrainedModel(PreTrainedModel):
|
|
|
|
|
|
|
|
|
|
|
300 |
config_class = GPTRefactConfig
|
301 |
base_model_prefix = "transformer"
|
302 |
supports_gradient_checkpointing = True
|
|
|
307 |
super().__init__(*inputs, **kwargs)
|
308 |
|
309 |
def _init_weights(self, module):
|
|
|
310 |
if isinstance(module, (MLP, Attention)):
|
311 |
# Reinitialize selected weights subject to the OpenAI GPT-2 Paper Scheme:
|
312 |
# > A modified initialization which accounts for the accumulation on the residual path with model depth. Scale
|
|
|
328 |
module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
|
329 |
if module.padding_idx is not None:
|
330 |
module.weight.data[module.padding_idx].zero_()
|
331 |
+
elif isinstance(module, LayerNormNoBias):
|
|
|
332 |
module.weight.data.fill_(1.0)
|
333 |
|
334 |
def _set_gradient_checkpointing(self, module, value=False):
|
|
|
367 |
return mask
|
368 |
|
369 |
def forward(
|
370 |
+
self,
|
371 |
+
input_ids: Optional[torch.Tensor] = None,
|
372 |
+
past_key_values: Optional[List[torch.Tensor]] = None,
|
373 |
+
attention_mask: Optional[torch.Tensor] = None,
|
374 |
+
inputs_embeds: Optional[torch.Tensor] = None,
|
375 |
+
use_cache: Optional[bool] = None,
|
376 |
+
output_attentions: Optional[bool] = None,
|
377 |
+
output_hidden_states: Optional[bool] = None,
|
378 |
+
return_dict: Optional[bool] = None,
|
|
|
|
|
|
|
|
|
|
|
379 |
) -> Union[Tuple, BaseModelOutputWithPastAndCrossAttentions]:
|
380 |
output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
|
381 |
output_hidden_states = (
|
|
|
401 |
|
402 |
device = input_ids.device if input_ids is not None else inputs_embeds.device
|
403 |
|
|
|
|
|
|
|
|
|
|
|
404 |
if past_key_values is None:
|
405 |
past_length = 0
|
406 |
past_key_values = tuple([None] * len(self.h))
|
407 |
else:
|
408 |
past_length = past_key_values[0][0].size(-2)
|
409 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
410 |
# Self-attention mask.
|
411 |
query_length = input_shape[-1]
|
412 |
|
|
|
421 |
alibi = get_alibi_biases(hidden_states.shape[0], seq_length_with_past,
|
422 |
self.num_heads, device, self.wte.weight.dtype)[:, :, -query_length:, :]
|
423 |
|
|
|
|
|
|
|
|
|
424 |
output_shape = input_shape + (hidden_states.size(-1),)
|
425 |
|
426 |
presents = [] if use_cache else None
|
|
|
445 |
hidden_states,
|
446 |
None,
|
447 |
attention_mask,
|
448 |
+
alibi
|
|
|
|
|
449 |
)
|
450 |
else:
|
451 |
outputs = block(
|
|
|
453 |
layer_past=layer_past,
|
454 |
attention_mask=attention_mask,
|
455 |
alibi=alibi,
|
|
|
|
|
456 |
use_cache=use_cache,
|
457 |
output_attentions=output_attentions,
|
458 |
)
|
|
|
486 |
cross_attentions=all_cross_attentions,
|
487 |
)
|
488 |
|
489 |
+
|
490 |
class GPTRefactForCausalLM(GPTRefactPreTrainedModel):
|
491 |
+
_tied_weights_keys = ["lm_head.weight", "ln_f.weight"]
|
492 |
|
493 |
def __init__(self, config):
|
494 |
super().__init__(config)
|
495 |
self.transformer = GPTRefactModel(config)
|
496 |
+
self.ln_f = LayerNormNoBias(self.transformer.embed_dim, eps=config.layer_norm_epsilon)
|
497 |
self.lm_head = nn.Linear(config.n_embd, config.vocab_size, bias=False)
|
498 |
|
499 |
# Initialize weights and apply final processing
|
500 |
self.post_init()
|
501 |
|
502 |
def prepare_inputs_for_generation(self, input_ids, past_key_values=None, inputs_embeds=None, **kwargs):
|
|
|
|
|
503 |
if inputs_embeds is not None and past_key_values is None:
|
504 |
model_inputs = {"inputs_embeds": inputs_embeds}
|
505 |
else:
|
|
|
517 |
return model_inputs
|
518 |
|
519 |
def forward(
|
520 |
+
self,
|
521 |
+
input_ids: Optional[torch.Tensor] = None,
|
522 |
+
past_key_values: Optional[Tuple[Tuple[torch.Tensor]]] = None,
|
523 |
+
attention_mask: Optional[torch.Tensor] = None,
|
524 |
+
inputs_embeds: Optional[torch.Tensor] = None,
|
525 |
+
labels: Optional[torch.Tensor] = None,
|
526 |
+
use_cache: Optional[bool] = None,
|
527 |
+
output_attentions: Optional[bool] = None,
|
528 |
+
output_hidden_states: Optional[bool] = None,
|
529 |
+
return_dict: Optional[bool] = None,
|
|
|
|
|
|
|
|
|
|
|
530 |
) -> Union[Tuple, CausalLMOutputWithCrossAttentions]:
|
531 |
r"""
|
532 |
labels (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
|
|
|
540 |
input_ids,
|
541 |
past_key_values=past_key_values,
|
542 |
attention_mask=attention_mask,
|
|
|
|
|
|
|
543 |
inputs_embeds=inputs_embeds,
|
|
|
|
|
544 |
use_cache=use_cache,
|
545 |
output_attentions=output_attentions,
|
546 |
output_hidden_states=output_hidden_states,
|
|
|
575 |
|
576 |
@staticmethod
|
577 |
def _reorder_cache(
|
578 |
+
past_key_values: Tuple[Tuple[torch.Tensor]], beam_idx: torch.Tensor
|
579 |
) -> Tuple[Tuple[torch.Tensor]]:
|
580 |
"""
|
581 |
This function is used to re-order the `past_key_values` cache if [`~PreTrainedModel.beam_search`] or
|
pytorch_model.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:58d077cf9a7cf9fa4589e3adb03603dab48a47af9e9a9bf084add65fe7574811
|
3 |
+
size 6343461637
|