liang.zhao
commited on
Commit
•
ab1dbe6
1
Parent(s):
fd43302
update model and config
Browse files- config.json +26 -26
- configuration_skywork.py +27 -14
- generation_config.json +1 -1
- modeling_skywork.py +79 -279
- pytorch_model-00032-of-00053.bin +2 -2
- tokenization_skywork.py +2 -19
config.json
CHANGED
@@ -1,27 +1,27 @@
|
|
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 |
-
}
|
|
|
1 |
{
|
2 |
+
"architectures": [
|
3 |
+
"SkyworkForCausalLM"
|
4 |
+
],
|
5 |
+
"auto_map": {
|
6 |
+
"AutoConfig": "configuration_skywork.SkyworkConfig",
|
7 |
+
"AutoModelForCausalLM": "modeling_skywork.SkyworkForCausalLM"
|
8 |
+
},
|
9 |
+
"bos_token_id": 1,
|
10 |
+
"eos_token_id": 2,
|
11 |
+
"pad_token_id": 0,
|
12 |
+
"hidden_act": "silu",
|
13 |
+
"hidden_size": 4608,
|
14 |
+
"initializer_range": 0.01,
|
15 |
+
"intermediate_size": 12288,
|
16 |
+
"max_position_embeddings": 131072,
|
17 |
+
"model_type": "skywork",
|
18 |
+
"num_attention_heads": 36,
|
19 |
+
"num_hidden_layers": 52,
|
20 |
+
"num_key_value_heads": 36,
|
21 |
+
"rms_norm_eps": 1e-06,
|
22 |
+
"tie_word_embeddings": false,
|
23 |
+
"torch_dtype": "bfloat16",
|
24 |
+
"transformers_version": "4.33.1",
|
25 |
+
"use_cache": true,
|
26 |
+
"vocab_size": 65519
|
27 |
+
}
|
configuration_skywork.py
CHANGED
@@ -1,13 +1,14 @@
|
|
1 |
# Copyright (c) SkyworkAI and the HuggingFace Inc. team. All rights reserved.
|
2 |
# This code is built upon Huggingface's transformers repository.
|
3 |
|
|
|
4 |
from transformers.configuration_utils import PretrainedConfig
|
5 |
from transformers.utils import logging
|
6 |
|
7 |
|
8 |
logger = logging.get_logger(__name__)
|
9 |
|
10 |
-
|
11 |
|
12 |
|
13 |
class SkyworkConfig(PretrainedConfig):
|
@@ -28,15 +29,13 @@ class SkyworkConfig(PretrainedConfig):
|
|
28 |
initializer_range=0.02,
|
29 |
rms_norm_eps=1e-6,
|
30 |
use_cache=True,
|
31 |
-
pad_token_id=
|
32 |
bos_token_id=1,
|
33 |
eos_token_id=2,
|
34 |
pretraining_tp=1,
|
35 |
tie_word_embeddings=False,
|
36 |
-
rope_scaling=None,
|
37 |
rope_theta=10000.0,
|
38 |
-
|
39 |
-
use_flash_attention=False,
|
40 |
**kwargs,
|
41 |
):
|
42 |
self.vocab_size = vocab_size
|
@@ -56,16 +55,9 @@ class SkyworkConfig(PretrainedConfig):
|
|
56 |
self.rms_norm_eps = rms_norm_eps
|
57 |
self.pretraining_tp = pretraining_tp
|
58 |
self.use_cache = use_cache
|
59 |
-
self.rope_scaling = rope_scaling
|
60 |
self.rope_theta = rope_theta
|
61 |
-
self.
|
62 |
-
self.
|
63 |
-
if self.use_flash_attention:
|
64 |
-
try:
|
65 |
-
from flash_attn.flash_attn_interface import flash_attn_varlen_func
|
66 |
-
from einops import rearrange
|
67 |
-
except:
|
68 |
-
raise ValueError("`use_flash_attention` requires Flash Attention 2+ and einops.\nTry `pip install einops` and installing Flash Attention from from https://github.com/Dao-AILab/flash-attention")
|
69 |
|
70 |
super().__init__(
|
71 |
pad_token_id=pad_token_id,
|
@@ -74,3 +66,24 @@ class SkyworkConfig(PretrainedConfig):
|
|
74 |
tie_word_embeddings=tie_word_embeddings,
|
75 |
**kwargs,
|
76 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
# Copyright (c) SkyworkAI and the HuggingFace Inc. team. All rights reserved.
|
2 |
# This code is built upon Huggingface's transformers repository.
|
3 |
|
4 |
+
|
5 |
from transformers.configuration_utils import PretrainedConfig
|
6 |
from transformers.utils import logging
|
7 |
|
8 |
|
9 |
logger = logging.get_logger(__name__)
|
10 |
|
11 |
+
LLAMA_PRETRAINED_CONFIG_ARCHIVE_MAP = {}
|
12 |
|
13 |
|
14 |
class SkyworkConfig(PretrainedConfig):
|
|
|
29 |
initializer_range=0.02,
|
30 |
rms_norm_eps=1e-6,
|
31 |
use_cache=True,
|
32 |
+
pad_token_id=None,
|
33 |
bos_token_id=1,
|
34 |
eos_token_id=2,
|
35 |
pretraining_tp=1,
|
36 |
tie_word_embeddings=False,
|
|
|
37 |
rope_theta=10000.0,
|
38 |
+
rope_scaling=None,
|
|
|
39 |
**kwargs,
|
40 |
):
|
41 |
self.vocab_size = vocab_size
|
|
|
55 |
self.rms_norm_eps = rms_norm_eps
|
56 |
self.pretraining_tp = pretraining_tp
|
57 |
self.use_cache = use_cache
|
|
|
58 |
self.rope_theta = rope_theta
|
59 |
+
self.rope_scaling = rope_scaling
|
60 |
+
self._rope_scaling_validation()
|
|
|
|
|
|
|
|
|
|
|
|
|
61 |
|
62 |
super().__init__(
|
63 |
pad_token_id=pad_token_id,
|
|
|
66 |
tie_word_embeddings=tie_word_embeddings,
|
67 |
**kwargs,
|
68 |
)
|
69 |
+
|
70 |
+
def _rope_scaling_validation(self):
|
71 |
+
"""
|
72 |
+
Validate the `rope_scaling` configuration.
|
73 |
+
"""
|
74 |
+
if self.rope_scaling is None:
|
75 |
+
return
|
76 |
+
|
77 |
+
if not isinstance(self.rope_scaling, dict) or len(self.rope_scaling) != 2:
|
78 |
+
raise ValueError(
|
79 |
+
"`rope_scaling` must be a dictionary with with two fields, `type` and `factor`, "
|
80 |
+
f"got {self.rope_scaling}"
|
81 |
+
)
|
82 |
+
rope_scaling_type = self.rope_scaling.get("type", None)
|
83 |
+
rope_scaling_factor = self.rope_scaling.get("factor", None)
|
84 |
+
if rope_scaling_type is None or rope_scaling_type not in ["linear", "dynamic", "ntk"]:
|
85 |
+
raise ValueError(
|
86 |
+
f"`rope_scaling`'s type field must be one of ['linear', 'dynamic'], got {rope_scaling_type}"
|
87 |
+
)
|
88 |
+
if rope_scaling_factor is None or not isinstance(rope_scaling_factor, float) or rope_scaling_factor <= 1.0:
|
89 |
+
raise ValueError(f"`rope_scaling`'s factor field must be an float > 1, got {rope_scaling_factor}")
|
generation_config.json
CHANGED
@@ -6,5 +6,5 @@
|
|
6 |
"pad_token_id": 0,
|
7 |
"temperature": 0.6,
|
8 |
"top_p": 0.9,
|
9 |
-
"transformers_version": "4.
|
10 |
}
|
|
|
6 |
"pad_token_id": 0,
|
7 |
"temperature": 0.6,
|
8 |
"top_p": 0.9,
|
9 |
+
"transformers_version": "4.33.1"
|
10 |
}
|
modeling_skywork.py
CHANGED
@@ -1,5 +1,6 @@
|
|
1 |
# Copyright (c) SkyworkAI and the HuggingFace Inc. team. All rights reserved.
|
2 |
# This code is built upon Huggingface's transformers repository.
|
|
|
3 |
import math
|
4 |
from typing import List, Optional, Tuple, Union
|
5 |
|
@@ -12,39 +13,15 @@ from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
|
|
12 |
from transformers.activations import ACT2FN
|
13 |
from transformers.modeling_outputs import BaseModelOutputWithPast, CausalLMOutputWithPast, SequenceClassifierOutputWithPast
|
14 |
from transformers.modeling_utils import PreTrainedModel
|
15 |
-
from transformers.
|
16 |
-
from transformers.utils import (
|
17 |
-
add_start_docstrings,
|
18 |
-
add_start_docstrings_to_model_forward,
|
19 |
-
is_flash_attn_available,
|
20 |
-
logging,
|
21 |
-
replace_return_docstrings,
|
22 |
-
)
|
23 |
from .configuration_skywork import SkyworkConfig
|
24 |
|
25 |
|
26 |
-
if is_flash_attn_available():
|
27 |
-
from flash_attn import flash_attn_func, flash_attn_varlen_func
|
28 |
-
from flash_attn.bert_padding import index_first_axis, pad_input, unpad_input # noqa
|
29 |
-
|
30 |
-
|
31 |
logger = logging.get_logger(__name__)
|
32 |
|
33 |
_CONFIG_FOR_DOC = "SkyworkConfig"
|
34 |
|
35 |
|
36 |
-
def _get_unpad_data(padding_mask):
|
37 |
-
seqlens_in_batch = padding_mask.sum(dim=-1, dtype=torch.int32)
|
38 |
-
indices = torch.nonzero(padding_mask.flatten(), as_tuple=False).flatten()
|
39 |
-
max_seqlen_in_batch = seqlens_in_batch.max().item()
|
40 |
-
cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.torch.int32), (1, 0))
|
41 |
-
return (
|
42 |
-
indices,
|
43 |
-
cu_seqlens,
|
44 |
-
max_seqlen_in_batch,
|
45 |
-
)
|
46 |
-
|
47 |
-
|
48 |
# Copied from transformers.models.bart.modeling_bart._make_causal_mask
|
49 |
def _make_causal_mask(
|
50 |
input_ids_shape: torch.Size, dtype: torch.dtype, device: torch.device, past_key_values_length: int = 0
|
@@ -95,10 +72,7 @@ class SkyworkRMSNorm(nn.Module):
|
|
95 |
return self.weight * hidden_states.to(input_dtype)
|
96 |
|
97 |
|
98 |
-
|
99 |
-
|
100 |
-
|
101 |
-
class SkyworkRotaryEmbedding(nn.Module):
|
102 |
def __init__(self, dim, max_position_embeddings=2048, base=10000, device=None):
|
103 |
super().__init__()
|
104 |
|
@@ -120,8 +94,8 @@ class SkyworkRotaryEmbedding(nn.Module):
|
|
120 |
freqs = torch.einsum("i,j->ij", t, self.inv_freq)
|
121 |
# Different from paper, but it uses a different permutation in order to obtain the same calculation
|
122 |
emb = torch.cat((freqs, freqs), dim=-1)
|
123 |
-
self.register_buffer("cos_cached", emb.cos().to(dtype), persistent=False)
|
124 |
-
self.register_buffer("sin_cached", emb.sin().to(dtype), persistent=False)
|
125 |
|
126 |
def forward(self, x, seq_len=None):
|
127 |
# x: [bs, num_attention_heads, seq_len, head_size]
|
@@ -129,8 +103,8 @@ class SkyworkRotaryEmbedding(nn.Module):
|
|
129 |
self._set_cos_sin_cache(seq_len=seq_len, device=x.device, dtype=x.dtype)
|
130 |
|
131 |
return (
|
132 |
-
self.cos_cached[:seq_len].to(dtype=x.dtype),
|
133 |
-
self.sin_cached[:seq_len].to(dtype=x.dtype),
|
134 |
)
|
135 |
|
136 |
|
@@ -149,8 +123,8 @@ class SkyworkLinearScalingRotaryEmbedding(SkyworkRotaryEmbedding):
|
|
149 |
freqs = torch.einsum("i,j->ij", t, self.inv_freq)
|
150 |
# Different from paper, but it uses a different permutation in order to obtain the same calculation
|
151 |
emb = torch.cat((freqs, freqs), dim=-1)
|
152 |
-
self.register_buffer("cos_cached", emb.cos().to(dtype), persistent=False)
|
153 |
-
self.register_buffer("sin_cached", emb.sin().to(dtype), persistent=False)
|
154 |
|
155 |
|
156 |
class SkyworkDynamicNTKScalingRotaryEmbedding(SkyworkRotaryEmbedding):
|
@@ -175,9 +149,42 @@ class SkyworkDynamicNTKScalingRotaryEmbedding(SkyworkRotaryEmbedding):
|
|
175 |
freqs = torch.einsum("i,j->ij", t, self.inv_freq)
|
176 |
# Different from paper, but it uses a different permutation in order to obtain the same calculation
|
177 |
emb = torch.cat((freqs, freqs), dim=-1)
|
178 |
-
self.register_buffer("cos_cached", emb.cos().to(dtype), persistent=False)
|
179 |
-
self.register_buffer("sin_cached", emb.sin().to(dtype), persistent=False)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
180 |
|
|
|
|
|
|
|
|
|
181 |
|
182 |
def rotate_half(x):
|
183 |
"""Rotates half the hidden dims of the input."""
|
@@ -186,10 +193,12 @@ def rotate_half(x):
|
|
186 |
return torch.cat((-x2, x1), dim=-1)
|
187 |
|
188 |
|
189 |
-
# Copied from transformers.models.gpt_neox.modeling_gpt_neox.apply_rotary_pos_emb
|
190 |
def apply_rotary_pos_emb(q, k, cos, sin, position_ids):
|
191 |
-
|
192 |
-
|
|
|
|
|
|
|
193 |
q_embed = (q * cos) + (rotate_half(q) * sin)
|
194 |
k_embed = (k * cos) + (rotate_half(k) * sin)
|
195 |
return q_embed, k_embed
|
@@ -260,10 +269,10 @@ class SkyworkAttention(nn.Module):
|
|
260 |
f"hidden_size must be divisible by num_heads (got `hidden_size`: {self.hidden_size}"
|
261 |
f" and `num_heads`: {self.num_heads})."
|
262 |
)
|
263 |
-
self.q_proj = nn.Linear(self.hidden_size, self.num_heads * self.head_dim, bias=
|
264 |
-
self.k_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=
|
265 |
-
self.v_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=
|
266 |
-
self.o_proj = nn.Linear(self.num_heads * self.head_dim, self.hidden_size, bias=
|
267 |
self._init_rope()
|
268 |
|
269 |
def _init_rope(self):
|
@@ -290,9 +299,18 @@ class SkyworkAttention(nn.Module):
|
|
290 |
scaling_factor=scaling_factor,
|
291 |
base=self.rope_theta,
|
292 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
293 |
else:
|
294 |
raise ValueError(f"Unknown RoPE scaling type {scaling_type}")
|
295 |
-
|
|
|
|
|
296 |
def _shape(self, tensor: torch.Tensor, seq_len: int, bsz: int):
|
297 |
return tensor.view(bsz, seq_len, self.num_heads, self.head_dim).transpose(1, 2).contiguous()
|
298 |
|
@@ -304,7 +322,6 @@ class SkyworkAttention(nn.Module):
|
|
304 |
past_key_value: Optional[Tuple[torch.Tensor]] = None,
|
305 |
output_attentions: bool = False,
|
306 |
use_cache: bool = False,
|
307 |
-
padding_mask: Optional[torch.LongTensor] = None,
|
308 |
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
|
309 |
bsz, q_len, _ = hidden_states.size()
|
310 |
|
@@ -347,6 +364,7 @@ class SkyworkAttention(nn.Module):
|
|
347 |
|
348 |
past_key_value = (key_states, value_states) if use_cache else None
|
349 |
|
|
|
350 |
key_states = repeat_kv(key_states, self.num_key_value_groups)
|
351 |
value_states = repeat_kv(value_states, self.num_key_value_groups)
|
352 |
|
@@ -376,7 +394,6 @@ class SkyworkAttention(nn.Module):
|
|
376 |
)
|
377 |
|
378 |
attn_output = attn_output.transpose(1, 2).contiguous()
|
379 |
-
|
380 |
attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
|
381 |
|
382 |
if self.config.pretraining_tp > 1:
|
@@ -392,193 +409,11 @@ class SkyworkAttention(nn.Module):
|
|
392 |
return attn_output, attn_weights, past_key_value
|
393 |
|
394 |
|
395 |
-
class SkyworkFlashAttention2(SkyworkAttention):
|
396 |
-
"""
|
397 |
-
Skywork flash attention module. This module inherits from `SkyworkAttention` as the weights of the module stays
|
398 |
-
untouched. The only required change would be on the forward pass where it needs to correctly call the public API of
|
399 |
-
flash attention and deal with padding tokens in case the input contains any of them.
|
400 |
-
"""
|
401 |
-
|
402 |
-
def forward(
|
403 |
-
self,
|
404 |
-
hidden_states: torch.Tensor,
|
405 |
-
attention_mask: Optional[torch.Tensor] = None,
|
406 |
-
position_ids: Optional[torch.LongTensor] = None,
|
407 |
-
past_key_value: Optional[Tuple[torch.Tensor]] = None,
|
408 |
-
output_attentions: bool = False,
|
409 |
-
use_cache: bool = False,
|
410 |
-
padding_mask: Optional[torch.LongTensor] = None,
|
411 |
-
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
|
412 |
-
# SkyworkFlashAttention2 attention does not support output_attentions
|
413 |
-
output_attentions = False
|
414 |
-
|
415 |
-
bsz, q_len, _ = hidden_states.size()
|
416 |
-
|
417 |
-
query_states = self.q_proj(hidden_states)
|
418 |
-
key_states = self.k_proj(hidden_states)
|
419 |
-
value_states = self.v_proj(hidden_states)
|
420 |
-
|
421 |
-
# Flash attention requires the input to have the shape
|
422 |
-
# batch_size x seq_length x head_dime x hidden_dim
|
423 |
-
# therefore we just need to keep the original shape
|
424 |
-
query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
|
425 |
-
key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
|
426 |
-
value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
|
427 |
-
|
428 |
-
kv_seq_len = key_states.shape[-2]
|
429 |
-
if past_key_value is not None:
|
430 |
-
kv_seq_len += past_key_value[0].shape[-2]
|
431 |
-
|
432 |
-
cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
|
433 |
-
|
434 |
-
query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids)
|
435 |
-
|
436 |
-
if past_key_value is not None:
|
437 |
-
# reuse k, v, self_attention
|
438 |
-
key_states = torch.cat([past_key_value[0], key_states], dim=2)
|
439 |
-
value_states = torch.cat([past_key_value[1], value_states], dim=2)
|
440 |
-
|
441 |
-
past_key_value = (key_states, value_states) if use_cache else None
|
442 |
-
|
443 |
-
query_states = query_states.transpose(1, 2)
|
444 |
-
key_states = key_states.transpose(1, 2)
|
445 |
-
value_states = value_states.transpose(1, 2)
|
446 |
-
|
447 |
-
# TODO: skywork does not have dropout in the config??
|
448 |
-
# It is recommended to use dropout with FA according to the docs
|
449 |
-
# when training.
|
450 |
-
dropout_rate = 0.0 # if not self.training else self.attn_dropout
|
451 |
-
|
452 |
-
# In PEFT, usually we cast the layer norms in float32 for training stability reasons
|
453 |
-
# therefore the input hidden states gets silently casted in float32. Hence, we need
|
454 |
-
# cast them back in float16 just to be sure everything works as expected.
|
455 |
-
# This might slowdown training & inference so it is recommended to not cast the LayerNorms
|
456 |
-
# in fp32. (SkyworkRMSNorm handles it correctly)
|
457 |
-
input_dtype = query_states.dtype
|
458 |
-
if input_dtype == torch.float32:
|
459 |
-
logger.warning_once(
|
460 |
-
"The input hidden states seems to be silently casted in float32, this might be related to"
|
461 |
-
" the fact you have upcasted embedding or layer norm layers in float32. We will cast back the input in"
|
462 |
-
" float16."
|
463 |
-
)
|
464 |
-
|
465 |
-
query_states = query_states.to(torch.float16)
|
466 |
-
key_states = key_states.to(torch.float16)
|
467 |
-
value_states = value_states.to(torch.float16)
|
468 |
-
|
469 |
-
attn_output = self._flash_attention_forward(
|
470 |
-
query_states, key_states, value_states, padding_mask, q_len, dropout=dropout_rate
|
471 |
-
)
|
472 |
-
|
473 |
-
attn_output = attn_output.reshape(bsz, q_len, self.hidden_size).contiguous()
|
474 |
-
attn_output = self.o_proj(attn_output)
|
475 |
-
|
476 |
-
if not output_attentions:
|
477 |
-
attn_weights = None
|
478 |
-
|
479 |
-
return attn_output, attn_weights, past_key_value
|
480 |
-
|
481 |
-
def _flash_attention_forward(
|
482 |
-
self, query_states, key_states, value_states, padding_mask, query_length, dropout=0.0, softmax_scale=None
|
483 |
-
):
|
484 |
-
"""
|
485 |
-
Calls the forward method of Flash Attention - if the input hidden states contain at least one padding token
|
486 |
-
first unpad the input, then computes the attention scores and pad the final attention scores.
|
487 |
-
|
488 |
-
Args:
|
489 |
-
query_states (`torch.Tensor`):
|
490 |
-
Input query states to be passed to Flash Attention API
|
491 |
-
key_states (`torch.Tensor`):
|
492 |
-
Input key states to be passed to Flash Attention API
|
493 |
-
value_states (`torch.Tensor`):
|
494 |
-
Input value states to be passed to Flash Attention API
|
495 |
-
padding_mask (`torch.Tensor`):
|
496 |
-
The padding mask - corresponds to a tensor of size `(batch_size, seq_len)` where 0 stands for the
|
497 |
-
position of padding tokens and 1 for the position of non-padding tokens.
|
498 |
-
dropout (`int`, *optional*):
|
499 |
-
Attention dropout
|
500 |
-
softmax_scale (`float`, *optional*):
|
501 |
-
The scaling of QK^T before applying softmax. Default to 1 / sqrt(head_dim)
|
502 |
-
"""
|
503 |
-
# Contains at least one padding token in the sequence
|
504 |
-
if padding_mask is not None:
|
505 |
-
batch_size = query_states.shape[0]
|
506 |
-
query_states, key_states, value_states, indices_q, cu_seq_lens, max_seq_lens = self._upad_input(
|
507 |
-
query_states, key_states, value_states, padding_mask, query_length
|
508 |
-
)
|
509 |
-
|
510 |
-
cu_seqlens_q, cu_seqlens_k = cu_seq_lens
|
511 |
-
max_seqlen_in_batch_q, max_seqlen_in_batch_k = max_seq_lens
|
512 |
-
|
513 |
-
attn_output_unpad = flash_attn_varlen_func(
|
514 |
-
query_states,
|
515 |
-
key_states,
|
516 |
-
value_states,
|
517 |
-
cu_seqlens_q=cu_seqlens_q,
|
518 |
-
cu_seqlens_k=cu_seqlens_k,
|
519 |
-
max_seqlen_q=max_seqlen_in_batch_q,
|
520 |
-
max_seqlen_k=max_seqlen_in_batch_k,
|
521 |
-
dropout_p=dropout,
|
522 |
-
softmax_scale=softmax_scale,
|
523 |
-
causal=True,
|
524 |
-
)
|
525 |
-
|
526 |
-
attn_output = pad_input(attn_output_unpad, indices_q, batch_size, query_length)
|
527 |
-
else:
|
528 |
-
attn_output = flash_attn_func(
|
529 |
-
query_states, key_states, value_states, dropout, softmax_scale=softmax_scale, causal=True
|
530 |
-
)
|
531 |
-
|
532 |
-
return attn_output
|
533 |
-
|
534 |
-
def _upad_input(self, query_layer, key_layer, value_layer, padding_mask, query_length):
|
535 |
-
indices_k, cu_seqlens_k, max_seqlen_in_batch_k = _get_unpad_data(padding_mask)
|
536 |
-
batch_size, kv_seq_len, num_key_value_heads, head_dim = key_layer.shape
|
537 |
-
|
538 |
-
key_layer = index_first_axis(
|
539 |
-
key_layer.reshape(batch_size * kv_seq_len, num_key_value_heads, head_dim), indices_k
|
540 |
-
)
|
541 |
-
value_layer = index_first_axis(
|
542 |
-
value_layer.reshape(batch_size * kv_seq_len, num_key_value_heads, head_dim), indices_k
|
543 |
-
)
|
544 |
-
if query_length == kv_seq_len:
|
545 |
-
query_layer = index_first_axis(
|
546 |
-
query_layer.reshape(batch_size * kv_seq_len, self.num_heads, head_dim), indices_k
|
547 |
-
)
|
548 |
-
cu_seqlens_q = cu_seqlens_k
|
549 |
-
max_seqlen_in_batch_q = max_seqlen_in_batch_k
|
550 |
-
indices_q = indices_k
|
551 |
-
elif query_length == 1:
|
552 |
-
max_seqlen_in_batch_q = 1
|
553 |
-
cu_seqlens_q = torch.arange(
|
554 |
-
batch_size + 1, dtype=torch.int32, device=query_layer.device
|
555 |
-
) # There is a memcpy here, that is very bad.
|
556 |
-
indices_q = cu_seqlens_q[:-1]
|
557 |
-
query_layer = query_layer.squeeze(1)
|
558 |
-
else:
|
559 |
-
# The -q_len: slice assumes left padding.
|
560 |
-
padding_mask = padding_mask[:, -query_length:]
|
561 |
-
query_layer, indices_q, cu_seqlens_q, max_seqlen_in_batch_q = unpad_input(query_layer, padding_mask)
|
562 |
-
|
563 |
-
return (
|
564 |
-
query_layer,
|
565 |
-
key_layer,
|
566 |
-
value_layer,
|
567 |
-
indices_q,
|
568 |
-
(cu_seqlens_q, cu_seqlens_k),
|
569 |
-
(max_seqlen_in_batch_q, max_seqlen_in_batch_k),
|
570 |
-
)
|
571 |
-
|
572 |
-
|
573 |
class SkyworkDecoderLayer(nn.Module):
|
574 |
def __init__(self, config: SkyworkConfig):
|
575 |
super().__init__()
|
576 |
self.hidden_size = config.hidden_size
|
577 |
-
self.self_attn = (
|
578 |
-
SkyworkAttention(config=config)
|
579 |
-
if not getattr(config, "_flash_attn_2_enabled", False)
|
580 |
-
else SkyworkFlashAttention2(config=config)
|
581 |
-
)
|
582 |
self.mlp = SkyworkMLP(config)
|
583 |
self.input_layernorm = SkyworkRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
|
584 |
self.post_attention_layernorm = SkyworkRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
|
@@ -591,7 +426,6 @@ class SkyworkDecoderLayer(nn.Module):
|
|
591 |
past_key_value: Optional[Tuple[torch.Tensor]] = None,
|
592 |
output_attentions: Optional[bool] = False,
|
593 |
use_cache: Optional[bool] = False,
|
594 |
-
padding_mask: Optional[torch.LongTensor] = None,
|
595 |
) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]:
|
596 |
"""
|
597 |
Args:
|
@@ -619,7 +453,6 @@ class SkyworkDecoderLayer(nn.Module):
|
|
619 |
past_key_value=past_key_value,
|
620 |
output_attentions=output_attentions,
|
621 |
use_cache=use_cache,
|
622 |
-
padding_mask=padding_mask,
|
623 |
)
|
624 |
hidden_states = residual + hidden_states
|
625 |
|
@@ -645,7 +478,6 @@ class SkyworkPreTrainedModel(PreTrainedModel):
|
|
645 |
supports_gradient_checkpointing = True
|
646 |
_no_split_modules = ["SkyworkDecoderLayer"]
|
647 |
_skip_keys_device_placement = "past_key_values"
|
648 |
-
_supports_flash_attn_2 = True
|
649 |
|
650 |
def _init_weights(self, module):
|
651 |
std = self.config.initializer_range
|
@@ -735,13 +567,13 @@ class SkyworkModel(SkyworkPreTrainedModel):
|
|
735 |
|
736 |
# retrieve input_ids and inputs_embeds
|
737 |
if input_ids is not None and inputs_embeds is not None:
|
738 |
-
raise ValueError("You cannot specify both
|
739 |
elif input_ids is not None:
|
740 |
batch_size, seq_length = input_ids.shape
|
741 |
elif inputs_embeds is not None:
|
742 |
batch_size, seq_length, _ = inputs_embeds.shape
|
743 |
else:
|
744 |
-
raise ValueError("You have to specify either
|
745 |
|
746 |
seq_length_with_past = seq_length
|
747 |
past_key_values_length = 0
|
@@ -755,7 +587,9 @@ class SkyworkModel(SkyworkPreTrainedModel):
|
|
755 |
position_ids = torch.arange(
|
756 |
past_key_values_length, seq_length + past_key_values_length, dtype=torch.long, device=device
|
757 |
)
|
758 |
-
position_ids = position_ids.unsqueeze(0)
|
|
|
|
|
759 |
|
760 |
if inputs_embeds is None:
|
761 |
inputs_embeds = self.embed_tokens(input_ids)
|
@@ -764,13 +598,6 @@ class SkyworkModel(SkyworkPreTrainedModel):
|
|
764 |
attention_mask = torch.ones(
|
765 |
(batch_size, seq_length_with_past), dtype=torch.bool, device=inputs_embeds.device
|
766 |
)
|
767 |
-
padding_mask = None
|
768 |
-
else:
|
769 |
-
if 0 in attention_mask:
|
770 |
-
padding_mask = attention_mask
|
771 |
-
else:
|
772 |
-
padding_mask = None
|
773 |
-
|
774 |
attention_mask = self._prepare_decoder_attention_mask(
|
775 |
attention_mask, (batch_size, seq_length), inputs_embeds, past_key_values_length
|
776 |
)
|
@@ -800,12 +627,15 @@ class SkyworkModel(SkyworkPreTrainedModel):
|
|
800 |
def create_custom_forward(module):
|
801 |
def custom_forward(*inputs):
|
802 |
# None for past_key_value
|
803 |
-
return module(*inputs, past_key_value, output_attentions
|
804 |
|
805 |
return custom_forward
|
806 |
|
807 |
layer_outputs = torch.utils.checkpoint.checkpoint(
|
808 |
-
create_custom_forward(decoder_layer),
|
|
|
|
|
|
|
809 |
)
|
810 |
else:
|
811 |
layer_outputs = decoder_layer(
|
@@ -815,7 +645,6 @@ class SkyworkModel(SkyworkPreTrainedModel):
|
|
815 |
past_key_value=past_key_value,
|
816 |
output_attentions=output_attentions,
|
817 |
use_cache=use_cache,
|
818 |
-
padding_mask=padding_mask,
|
819 |
)
|
820 |
|
821 |
hidden_states = layer_outputs[0]
|
@@ -873,7 +702,6 @@ class SkyworkForCausalLM(SkyworkPreTrainedModel):
|
|
873 |
def get_decoder(self):
|
874 |
return self.model
|
875 |
|
876 |
-
@replace_return_docstrings(output_type=CausalLMOutputWithPast, config_class=_CONFIG_FOR_DOC)
|
877 |
def forward(
|
878 |
self,
|
879 |
input_ids: torch.LongTensor = None,
|
@@ -887,31 +715,6 @@ class SkyworkForCausalLM(SkyworkPreTrainedModel):
|
|
887 |
output_hidden_states: Optional[bool] = None,
|
888 |
return_dict: Optional[bool] = None,
|
889 |
) -> Union[Tuple, CausalLMOutputWithPast]:
|
890 |
-
r"""
|
891 |
-
Args:
|
892 |
-
labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
|
893 |
-
Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
|
894 |
-
config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
|
895 |
-
(masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`.
|
896 |
-
|
897 |
-
Returns:
|
898 |
-
|
899 |
-
Example:
|
900 |
-
|
901 |
-
```python
|
902 |
-
>>> from transformers import AutoTokenizer, SkyworkForCausalLM
|
903 |
-
|
904 |
-
>>> model = SkyworkForCausalLM.from_pretrained(PATH_TO_CONVERTED_WEIGHTS)
|
905 |
-
>>> tokenizer = AutoTokenizer.from_pretrained(PATH_TO_CONVERTED_TOKENIZER)
|
906 |
-
|
907 |
-
>>> prompt = "Hey, are you conscious? Can you talk to me?"
|
908 |
-
>>> inputs = tokenizer(prompt, return_tensors="pt")
|
909 |
-
|
910 |
-
>>> # Generate
|
911 |
-
>>> generate_ids = model.generate(inputs.input_ids, max_length=30)
|
912 |
-
>>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
|
913 |
-
"Hey, are you conscious? Can you talk to me?\nI'm not conscious, but I can talk to you."
|
914 |
-
```"""
|
915 |
|
916 |
output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
|
917 |
output_hidden_states = (
|
@@ -1005,6 +808,7 @@ class SkyworkForCausalLM(SkyworkPreTrainedModel):
|
|
1005 |
)
|
1006 |
return reordered_past
|
1007 |
|
|
|
1008 |
class SkyworkForSequenceClassification(SkyworkPreTrainedModel):
|
1009 |
def __init__(self, config):
|
1010 |
super().__init__(config)
|
@@ -1034,12 +838,8 @@ class SkyworkForSequenceClassification(SkyworkPreTrainedModel):
|
|
1034 |
output_hidden_states: Optional[bool] = None,
|
1035 |
return_dict: Optional[bool] = None,
|
1036 |
) -> Union[Tuple, SequenceClassifierOutputWithPast]:
|
1037 |
-
|
1038 |
-
|
1039 |
-
Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
|
1040 |
-
config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
|
1041 |
-
`config.num_labels > 1` a classification loss is computed (Cross-Entropy).
|
1042 |
-
"""
|
1043 |
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
|
1044 |
|
1045 |
transformer_outputs = self.model(
|
@@ -1108,4 +908,4 @@ class SkyworkForSequenceClassification(SkyworkPreTrainedModel):
|
|
1108 |
past_key_values=transformer_outputs.past_key_values,
|
1109 |
hidden_states=transformer_outputs.hidden_states,
|
1110 |
attentions=transformer_outputs.attentions,
|
1111 |
-
)
|
|
|
1 |
# Copyright (c) SkyworkAI and the HuggingFace Inc. team. All rights reserved.
|
2 |
# This code is built upon Huggingface's transformers repository.
|
3 |
+
|
4 |
import math
|
5 |
from typing import List, Optional, Tuple, Union
|
6 |
|
|
|
13 |
from transformers.activations import ACT2FN
|
14 |
from transformers.modeling_outputs import BaseModelOutputWithPast, CausalLMOutputWithPast, SequenceClassifierOutputWithPast
|
15 |
from transformers.modeling_utils import PreTrainedModel
|
16 |
+
from transformers.utils import logging
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
17 |
from .configuration_skywork import SkyworkConfig
|
18 |
|
19 |
|
|
|
|
|
|
|
|
|
|
|
20 |
logger = logging.get_logger(__name__)
|
21 |
|
22 |
_CONFIG_FOR_DOC = "SkyworkConfig"
|
23 |
|
24 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
25 |
# Copied from transformers.models.bart.modeling_bart._make_causal_mask
|
26 |
def _make_causal_mask(
|
27 |
input_ids_shape: torch.Size, dtype: torch.dtype, device: torch.device, past_key_values_length: int = 0
|
|
|
72 |
return self.weight * hidden_states.to(input_dtype)
|
73 |
|
74 |
|
75 |
+
class SkyworkRotaryEmbedding(torch.nn.Module):
|
|
|
|
|
|
|
76 |
def __init__(self, dim, max_position_embeddings=2048, base=10000, device=None):
|
77 |
super().__init__()
|
78 |
|
|
|
94 |
freqs = torch.einsum("i,j->ij", t, self.inv_freq)
|
95 |
# Different from paper, but it uses a different permutation in order to obtain the same calculation
|
96 |
emb = torch.cat((freqs, freqs), dim=-1)
|
97 |
+
self.register_buffer("cos_cached", emb.cos()[None, None, :, :].to(dtype), persistent=False)
|
98 |
+
self.register_buffer("sin_cached", emb.sin()[None, None, :, :].to(dtype), persistent=False)
|
99 |
|
100 |
def forward(self, x, seq_len=None):
|
101 |
# x: [bs, num_attention_heads, seq_len, head_size]
|
|
|
103 |
self._set_cos_sin_cache(seq_len=seq_len, device=x.device, dtype=x.dtype)
|
104 |
|
105 |
return (
|
106 |
+
self.cos_cached[:, :, :seq_len, ...].to(dtype=x.dtype),
|
107 |
+
self.sin_cached[:, :, :seq_len, ...].to(dtype=x.dtype),
|
108 |
)
|
109 |
|
110 |
|
|
|
123 |
freqs = torch.einsum("i,j->ij", t, self.inv_freq)
|
124 |
# Different from paper, but it uses a different permutation in order to obtain the same calculation
|
125 |
emb = torch.cat((freqs, freqs), dim=-1)
|
126 |
+
self.register_buffer("cos_cached", emb.cos()[None, None, :, :].to(dtype), persistent=False)
|
127 |
+
self.register_buffer("sin_cached", emb.sin()[None, None, :, :].to(dtype), persistent=False)
|
128 |
|
129 |
|
130 |
class SkyworkDynamicNTKScalingRotaryEmbedding(SkyworkRotaryEmbedding):
|
|
|
149 |
freqs = torch.einsum("i,j->ij", t, self.inv_freq)
|
150 |
# Different from paper, but it uses a different permutation in order to obtain the same calculation
|
151 |
emb = torch.cat((freqs, freqs), dim=-1)
|
152 |
+
self.register_buffer("cos_cached", emb.cos()[None, None, :, :].to(dtype), persistent=False)
|
153 |
+
self.register_buffer("sin_cached", emb.sin()[None, None, :, :].to(dtype), persistent=False)
|
154 |
+
|
155 |
+
|
156 |
+
|
157 |
+
class SkyworkNTKScalingRotaryEmbedding(torch.nn.Module):
|
158 |
+
def __init__(self, dim, max_position_embeddings=2048, base=10000, scaling_factor=100, device=None):
|
159 |
+
super().__init__()
|
160 |
+
|
161 |
+
self.dim = dim
|
162 |
+
self.max_position_embeddings = max_position_embeddings
|
163 |
+
self.base = base * scaling_factor
|
164 |
+
inv_freq = 1.0 / (self.base ** (torch.arange(0, self.dim, 2).float().to(device) / self.dim))
|
165 |
+
self.register_buffer("inv_freq", inv_freq, persistent=False)
|
166 |
+
|
167 |
+
# Build here to make `torch.jit.trace` work.
|
168 |
+
self._set_cos_sin_cache(
|
169 |
+
seq_len=max_position_embeddings, device=self.inv_freq.device, dtype=torch.get_default_dtype()
|
170 |
+
)
|
171 |
+
|
172 |
+
def _set_cos_sin_cache(self, seq_len, device, dtype):
|
173 |
+
self.max_seq_len_cached = seq_len
|
174 |
+
t = torch.arange(self.max_seq_len_cached, device=device, dtype=self.inv_freq.dtype)
|
175 |
+
freqs = torch.einsum("i,j->ij", t, self.inv_freq)
|
176 |
+
emb = torch.cat((freqs, freqs), dim=-1)
|
177 |
+
self.register_buffer("cos_cached", emb.cos()[None, None, :, :].to(dtype), persistent=False)
|
178 |
+
self.register_buffer("sin_cached", emb.sin()[None, None, :, :].to(dtype), persistent=False)
|
179 |
+
|
180 |
+
def forward(self, x, seq_len=None):
|
181 |
+
if seq_len > self.max_seq_len_cached:
|
182 |
+
self._set_cos_sin_cache(seq_len=seq_len, device=x.device, dtype=x.dtype)
|
183 |
|
184 |
+
return (
|
185 |
+
self.cos_cached[:, :, :seq_len, ...].to(dtype=x.dtype),
|
186 |
+
self.sin_cached[:, :, :seq_len, ...].to(dtype=x.dtype),
|
187 |
+
)
|
188 |
|
189 |
def rotate_half(x):
|
190 |
"""Rotates half the hidden dims of the input."""
|
|
|
193 |
return torch.cat((-x2, x1), dim=-1)
|
194 |
|
195 |
|
|
|
196 |
def apply_rotary_pos_emb(q, k, cos, sin, position_ids):
|
197 |
+
# The first two dimensions of cos and sin are always 1, so we can `squeeze` them.
|
198 |
+
cos = cos.squeeze(1).squeeze(0) # [seq_len, dim]
|
199 |
+
sin = sin.squeeze(1).squeeze(0) # [seq_len, dim]
|
200 |
+
cos = cos[position_ids].unsqueeze(1) # [bs, 1, seq_len, dim]
|
201 |
+
sin = sin[position_ids].unsqueeze(1) # [bs, 1, seq_len, dim]
|
202 |
q_embed = (q * cos) + (rotate_half(q) * sin)
|
203 |
k_embed = (k * cos) + (rotate_half(k) * sin)
|
204 |
return q_embed, k_embed
|
|
|
269 |
f"hidden_size must be divisible by num_heads (got `hidden_size`: {self.hidden_size}"
|
270 |
f" and `num_heads`: {self.num_heads})."
|
271 |
)
|
272 |
+
self.q_proj = nn.Linear(self.hidden_size, self.num_heads * self.head_dim, bias=False)
|
273 |
+
self.k_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False)
|
274 |
+
self.v_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False)
|
275 |
+
self.o_proj = nn.Linear(self.num_heads * self.head_dim, self.hidden_size, bias=False)
|
276 |
self._init_rope()
|
277 |
|
278 |
def _init_rope(self):
|
|
|
299 |
scaling_factor=scaling_factor,
|
300 |
base=self.rope_theta,
|
301 |
)
|
302 |
+
elif scaling_type == "ntk":
|
303 |
+
self.rotary_emb = SkyworkNTKScalingRotaryEmbedding(
|
304 |
+
self.head_dim,
|
305 |
+
max_position_embeddings=self.max_position_embeddings,
|
306 |
+
scaling_factor=scaling_factor,
|
307 |
+
base=self.rope_theta,
|
308 |
+
)
|
309 |
else:
|
310 |
raise ValueError(f"Unknown RoPE scaling type {scaling_type}")
|
311 |
+
print('-'*80)
|
312 |
+
print(f"USING COSTOM MODELING, scaling_type is {scaling_type}, scaling_factor is {scaling_factor}")
|
313 |
+
|
314 |
def _shape(self, tensor: torch.Tensor, seq_len: int, bsz: int):
|
315 |
return tensor.view(bsz, seq_len, self.num_heads, self.head_dim).transpose(1, 2).contiguous()
|
316 |
|
|
|
322 |
past_key_value: Optional[Tuple[torch.Tensor]] = None,
|
323 |
output_attentions: bool = False,
|
324 |
use_cache: bool = False,
|
|
|
325 |
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
|
326 |
bsz, q_len, _ = hidden_states.size()
|
327 |
|
|
|
364 |
|
365 |
past_key_value = (key_states, value_states) if use_cache else None
|
366 |
|
367 |
+
# repeat k/v heads if n_kv_heads < n_heads
|
368 |
key_states = repeat_kv(key_states, self.num_key_value_groups)
|
369 |
value_states = repeat_kv(value_states, self.num_key_value_groups)
|
370 |
|
|
|
394 |
)
|
395 |
|
396 |
attn_output = attn_output.transpose(1, 2).contiguous()
|
|
|
397 |
attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
|
398 |
|
399 |
if self.config.pretraining_tp > 1:
|
|
|
409 |
return attn_output, attn_weights, past_key_value
|
410 |
|
411 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
412 |
class SkyworkDecoderLayer(nn.Module):
|
413 |
def __init__(self, config: SkyworkConfig):
|
414 |
super().__init__()
|
415 |
self.hidden_size = config.hidden_size
|
416 |
+
self.self_attn = SkyworkAttention(config=config)
|
|
|
|
|
|
|
|
|
417 |
self.mlp = SkyworkMLP(config)
|
418 |
self.input_layernorm = SkyworkRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
|
419 |
self.post_attention_layernorm = SkyworkRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
|
|
|
426 |
past_key_value: Optional[Tuple[torch.Tensor]] = None,
|
427 |
output_attentions: Optional[bool] = False,
|
428 |
use_cache: Optional[bool] = False,
|
|
|
429 |
) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]:
|
430 |
"""
|
431 |
Args:
|
|
|
453 |
past_key_value=past_key_value,
|
454 |
output_attentions=output_attentions,
|
455 |
use_cache=use_cache,
|
|
|
456 |
)
|
457 |
hidden_states = residual + hidden_states
|
458 |
|
|
|
478 |
supports_gradient_checkpointing = True
|
479 |
_no_split_modules = ["SkyworkDecoderLayer"]
|
480 |
_skip_keys_device_placement = "past_key_values"
|
|
|
481 |
|
482 |
def _init_weights(self, module):
|
483 |
std = self.config.initializer_range
|
|
|
567 |
|
568 |
# retrieve input_ids and inputs_embeds
|
569 |
if input_ids is not None and inputs_embeds is not None:
|
570 |
+
raise ValueError("You cannot specify both decoder_input_ids and decoder_inputs_embeds at the same time")
|
571 |
elif input_ids is not None:
|
572 |
batch_size, seq_length = input_ids.shape
|
573 |
elif inputs_embeds is not None:
|
574 |
batch_size, seq_length, _ = inputs_embeds.shape
|
575 |
else:
|
576 |
+
raise ValueError("You have to specify either decoder_input_ids or decoder_inputs_embeds")
|
577 |
|
578 |
seq_length_with_past = seq_length
|
579 |
past_key_values_length = 0
|
|
|
587 |
position_ids = torch.arange(
|
588 |
past_key_values_length, seq_length + past_key_values_length, dtype=torch.long, device=device
|
589 |
)
|
590 |
+
position_ids = position_ids.unsqueeze(0).view(-1, seq_length)
|
591 |
+
else:
|
592 |
+
position_ids = position_ids.view(-1, seq_length).long()
|
593 |
|
594 |
if inputs_embeds is None:
|
595 |
inputs_embeds = self.embed_tokens(input_ids)
|
|
|
598 |
attention_mask = torch.ones(
|
599 |
(batch_size, seq_length_with_past), dtype=torch.bool, device=inputs_embeds.device
|
600 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
601 |
attention_mask = self._prepare_decoder_attention_mask(
|
602 |
attention_mask, (batch_size, seq_length), inputs_embeds, past_key_values_length
|
603 |
)
|
|
|
627 |
def create_custom_forward(module):
|
628 |
def custom_forward(*inputs):
|
629 |
# None for past_key_value
|
630 |
+
return module(*inputs, past_key_value, output_attentions)
|
631 |
|
632 |
return custom_forward
|
633 |
|
634 |
layer_outputs = torch.utils.checkpoint.checkpoint(
|
635 |
+
create_custom_forward(decoder_layer),
|
636 |
+
hidden_states,
|
637 |
+
attention_mask,
|
638 |
+
position_ids,
|
639 |
)
|
640 |
else:
|
641 |
layer_outputs = decoder_layer(
|
|
|
645 |
past_key_value=past_key_value,
|
646 |
output_attentions=output_attentions,
|
647 |
use_cache=use_cache,
|
|
|
648 |
)
|
649 |
|
650 |
hidden_states = layer_outputs[0]
|
|
|
702 |
def get_decoder(self):
|
703 |
return self.model
|
704 |
|
|
|
705 |
def forward(
|
706 |
self,
|
707 |
input_ids: torch.LongTensor = None,
|
|
|
715 |
output_hidden_states: Optional[bool] = None,
|
716 |
return_dict: Optional[bool] = None,
|
717 |
) -> Union[Tuple, CausalLMOutputWithPast]:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
718 |
|
719 |
output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
|
720 |
output_hidden_states = (
|
|
|
808 |
)
|
809 |
return reordered_past
|
810 |
|
811 |
+
|
812 |
class SkyworkForSequenceClassification(SkyworkPreTrainedModel):
|
813 |
def __init__(self, config):
|
814 |
super().__init__(config)
|
|
|
838 |
output_hidden_states: Optional[bool] = None,
|
839 |
return_dict: Optional[bool] = None,
|
840 |
) -> Union[Tuple, SequenceClassifierOutputWithPast]:
|
841 |
+
|
842 |
+
|
|
|
|
|
|
|
|
|
843 |
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
|
844 |
|
845 |
transformer_outputs = self.model(
|
|
|
908 |
past_key_values=transformer_outputs.past_key_values,
|
909 |
hidden_states=transformer_outputs.hidden_states,
|
910 |
attentions=transformer_outputs.attentions,
|
911 |
+
)
|
pytorch_model-00032-of-00053.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 sha256:94d2a7a13be4bb785cdf29f0654109ff58604608141b126a4c3fd2dc40c60188
|
3 |
+
size 301957120
|
tokenization_skywork.py
CHANGED
@@ -1,22 +1,5 @@
|
|
1 |
-
#
|
2 |
-
#
|
3 |
-
#
|
4 |
-
# This code is based on EleutherAI's GPT-NeoX library and the GPT-NeoX
|
5 |
-
# and OPT implementations in this library. It has been modified from its
|
6 |
-
# original forms to accommodate minor architectural differences compared
|
7 |
-
# to GPT-NeoX and OPT used by the Meta AI team that trained the model.
|
8 |
-
#
|
9 |
-
# Licensed under the Apache License, Version 2.0 (the "License");
|
10 |
-
# you may not use this file except in compliance with the License.
|
11 |
-
# You may obtain a copy of the License at
|
12 |
-
#
|
13 |
-
# http://www.apache.org/licenses/LICENSE-2.0
|
14 |
-
#
|
15 |
-
# Unless required by applicable law or agreed to in writing, software
|
16 |
-
# distributed under the License is distributed on an "AS IS" BASIS,
|
17 |
-
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
18 |
-
# See the License for the specific language governing permissions and
|
19 |
-
# limitations under the License.
|
20 |
|
21 |
"""Tokenization classes for Skywork."""
|
22 |
import os
|
|
|
1 |
+
# Copyright (c) SkyworkAI and the HuggingFace Inc. team. All rights reserved.
|
2 |
+
# This code is built upon Huggingface's transformers repository.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
|
4 |
"""Tokenization classes for Skywork."""
|
5 |
import os
|