feat: add flash attention v2
#9
by
jon-tow
- opened
- README.md +30 -1
- modeling_stablelm_epoch.py +246 -19
README.md
CHANGED
|
@@ -36,7 +36,7 @@ model = AutoModelForCausalLM.from_pretrained(
|
|
| 36 |
torch_dtype="auto",
|
| 37 |
)
|
| 38 |
model.cuda()
|
| 39 |
-
inputs = tokenizer("The weather is always wonderful", return_tensors="pt").to(
|
| 40 |
tokens = model.generate(
|
| 41 |
**inputs,
|
| 42 |
max_new_tokens=64,
|
|
@@ -47,6 +47,35 @@ tokens = model.generate(
|
|
| 47 |
print(tokenizer.decode(tokens[0], skip_special_tokens=True))
|
| 48 |
```
|
| 49 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 50 |
## Model Details
|
| 51 |
|
| 52 |
* **Developed by**: [Stability AI](https://stability.ai/)
|
|
|
|
| 36 |
torch_dtype="auto",
|
| 37 |
)
|
| 38 |
model.cuda()
|
| 39 |
+
inputs = tokenizer("The weather is always wonderful", return_tensors="pt").to(model.device)
|
| 40 |
tokens = model.generate(
|
| 41 |
**inputs,
|
| 42 |
max_new_tokens=64,
|
|
|
|
| 47 |
print(tokenizer.decode(tokens[0], skip_special_tokens=True))
|
| 48 |
```
|
| 49 |
|
| 50 |
+
### Run with Flash Attention 2 ⚡️
|
| 51 |
+
|
| 52 |
+
<details>
|
| 53 |
+
<summary> Click to expand </summary>
|
| 54 |
+
|
| 55 |
+
```python
|
| 56 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 57 |
+
tokenizer = AutoTokenizer.from_pretrained("stabilityai/stablelm-3b-4e1t")
|
| 58 |
+
model = AutoModelForCausalLM.from_pretrained(
|
| 59 |
+
"stabilityai/stablelm-3b-4e1t",
|
| 60 |
+
trust_remote_code=True,
|
| 61 |
+
torch_dtype="auto",
|
| 62 |
+
+ use_flash_attention_2=True,
|
| 63 |
+
)
|
| 64 |
+
model.cuda()
|
| 65 |
+
inputs = tokenizer("The weather is always wonderful", return_tensors="pt").to(model.device)
|
| 66 |
+
tokens = model.generate(
|
| 67 |
+
**inputs,
|
| 68 |
+
max_new_tokens=64,
|
| 69 |
+
temperature=0.75,
|
| 70 |
+
top_p=0.95,
|
| 71 |
+
do_sample=True,
|
| 72 |
+
)
|
| 73 |
+
print(tokenizer.decode(tokens[0], skip_special_tokens=True))
|
| 74 |
+
```
|
| 75 |
+
|
| 76 |
+
</details>
|
| 77 |
+
|
| 78 |
+
|
| 79 |
## Model Details
|
| 80 |
|
| 81 |
* **Developed by**: [Stability AI](https://stability.ai/)
|
modeling_stablelm_epoch.py
CHANGED
|
@@ -19,23 +19,46 @@
|
|
| 19 |
""" PyTorch StableLM Epoch model. """
|
| 20 |
from typing import Optional, Tuple, Union
|
| 21 |
import math
|
|
|
|
| 22 |
|
| 23 |
import torch
|
|
|
|
| 24 |
import torch.utils.checkpoint
|
| 25 |
from torch import nn
|
| 26 |
from torch.nn import CrossEntropyLoss
|
|
|
|
|
|
|
| 27 |
from transformers.modeling_outputs import (
|
| 28 |
BaseModelOutputWithPast,
|
| 29 |
CausalLMOutputWithPast,
|
| 30 |
)
|
| 31 |
from transformers.modeling_utils import PreTrainedModel
|
| 32 |
-
from transformers.utils import logging
|
|
|
|
| 33 |
from .configuration_stablelm_epoch import StableLMEpochConfig
|
| 34 |
|
| 35 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 36 |
logger = logging.get_logger(__name__)
|
| 37 |
|
| 38 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 39 |
# Copied from transformers.models.bart.modeling_bart._make_causal_mask
|
| 40 |
def _make_causal_mask(
|
| 41 |
input_ids_shape: torch.Size,
|
|
@@ -165,6 +188,7 @@ class Attention(nn.Module):
|
|
| 165 |
self.num_key_value_heads = config.num_key_value_heads
|
| 166 |
self.num_key_value_groups = self.num_heads // self.num_key_value_heads
|
| 167 |
self.max_position_embeddings = config.max_position_embeddings
|
|
|
|
| 168 |
|
| 169 |
if (self.head_dim * self.num_heads) != self.hidden_size:
|
| 170 |
raise ValueError(
|
|
@@ -269,10 +293,202 @@ class Attention(nn.Module):
|
|
| 269 |
return attn_output, attn_weights, past_key_value
|
| 270 |
|
| 271 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 272 |
class DecoderLayer(nn.Module):
|
| 273 |
def __init__(self, config: StableLMEpochConfig):
|
| 274 |
super().__init__()
|
| 275 |
-
self.self_attn =
|
| 276 |
self.mlp = MLP(config)
|
| 277 |
self.input_layernorm = nn.LayerNorm(config.hidden_size, eps=config.norm_eps)
|
| 278 |
self.post_attention_layernorm = nn.LayerNorm(config.hidden_size, eps=config.norm_eps)
|
|
@@ -328,6 +544,7 @@ class StableLMEpochPreTrainedModel(PreTrainedModel):
|
|
| 328 |
supports_gradient_checkpointing = True
|
| 329 |
_no_split_modules = ["DecoderLayer"]
|
| 330 |
_skip_keys_device_placement = "past_key_values"
|
|
|
|
| 331 |
|
| 332 |
def _init_weights(self, module: nn.Module):
|
| 333 |
"""Initialize the weights"""
|
|
@@ -355,6 +572,7 @@ class StableLMEpochModel(StableLMEpochPreTrainedModel):
|
|
| 355 |
self.layers = nn.ModuleList([DecoderLayer(config) for _ in range(config.num_hidden_layers)])
|
| 356 |
self.norm = nn.LayerNorm(config.hidden_size, eps=config.norm_eps)
|
| 357 |
|
|
|
|
| 358 |
self.gradient_checkpointing = False
|
| 359 |
# Initialize weights and apply final processing
|
| 360 |
self.post_init()
|
|
@@ -428,10 +646,6 @@ class StableLMEpochModel(StableLMEpochPreTrainedModel):
|
|
| 428 |
seq_length_with_past = seq_length
|
| 429 |
past_key_values_length = 0
|
| 430 |
|
| 431 |
-
if past_key_values is not None:
|
| 432 |
-
past_key_values_length = past_key_values[0][0].shape[2]
|
| 433 |
-
seq_length_with_past = seq_length_with_past + past_key_values_length
|
| 434 |
-
|
| 435 |
if position_ids is None:
|
| 436 |
device = input_ids.device if input_ids is not None else inputs_embeds.device
|
| 437 |
position_ids = torch.arange(
|
|
@@ -447,18 +661,22 @@ class StableLMEpochModel(StableLMEpochPreTrainedModel):
|
|
| 447 |
if inputs_embeds is None:
|
| 448 |
inputs_embeds = self.embed_tokens(input_ids)
|
| 449 |
# Embed positions
|
| 450 |
-
if
|
| 451 |
-
|
| 452 |
-
|
| 453 |
-
|
| 454 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 455 |
)
|
| 456 |
-
attention_mask = self._prepare_decoder_attention_mask(
|
| 457 |
-
attention_mask,
|
| 458 |
-
(batch_size, seq_length),
|
| 459 |
-
inputs_embeds,
|
| 460 |
-
past_key_values_length,
|
| 461 |
-
)
|
| 462 |
|
| 463 |
hidden_states = inputs_embeds
|
| 464 |
|
|
@@ -643,8 +861,17 @@ class StableLMEpochForCausalLM(StableLMEpochPreTrainedModel):
|
|
| 643 |
**kwargs,
|
| 644 |
):
|
| 645 |
# Trim decoder_input_ids if past is used
|
| 646 |
-
if past_key_values
|
| 647 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 648 |
|
| 649 |
position_ids = kwargs.get("position_ids", None)
|
| 650 |
if attention_mask is not None and position_ids is None:
|
|
|
|
| 19 |
""" PyTorch StableLM Epoch model. """
|
| 20 |
from typing import Optional, Tuple, Union
|
| 21 |
import math
|
| 22 |
+
import warnings
|
| 23 |
|
| 24 |
import torch
|
| 25 |
+
import torch.nn.functional as F
|
| 26 |
import torch.utils.checkpoint
|
| 27 |
from torch import nn
|
| 28 |
from torch.nn import CrossEntropyLoss
|
| 29 |
+
|
| 30 |
+
from transformers.cache_utils import Cache
|
| 31 |
from transformers.modeling_outputs import (
|
| 32 |
BaseModelOutputWithPast,
|
| 33 |
CausalLMOutputWithPast,
|
| 34 |
)
|
| 35 |
from transformers.modeling_utils import PreTrainedModel
|
| 36 |
+
from transformers.utils import logging, is_flash_attn_2_available, is_flash_attn_greater_or_equal_2_10
|
| 37 |
+
|
| 38 |
from .configuration_stablelm_epoch import StableLMEpochConfig
|
| 39 |
|
| 40 |
|
| 41 |
+
if is_flash_attn_2_available():
|
| 42 |
+
from flash_attn import flash_attn_func, flash_attn_varlen_func
|
| 43 |
+
from flash_attn.bert_padding import index_first_axis, pad_input, unpad_input # noqa
|
| 44 |
+
|
| 45 |
+
|
| 46 |
logger = logging.get_logger(__name__)
|
| 47 |
|
| 48 |
|
| 49 |
+
# Copied from transformers.models.llama.modeling_llama._get_unpad_data
|
| 50 |
+
def _get_unpad_data(attention_mask):
|
| 51 |
+
seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
|
| 52 |
+
indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
|
| 53 |
+
max_seqlen_in_batch = seqlens_in_batch.max().item()
|
| 54 |
+
cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.torch.int32), (1, 0))
|
| 55 |
+
return (
|
| 56 |
+
indices,
|
| 57 |
+
cu_seqlens,
|
| 58 |
+
max_seqlen_in_batch,
|
| 59 |
+
)
|
| 60 |
+
|
| 61 |
+
|
| 62 |
# Copied from transformers.models.bart.modeling_bart._make_causal_mask
|
| 63 |
def _make_causal_mask(
|
| 64 |
input_ids_shape: torch.Size,
|
|
|
|
| 188 |
self.num_key_value_heads = config.num_key_value_heads
|
| 189 |
self.num_key_value_groups = self.num_heads // self.num_key_value_heads
|
| 190 |
self.max_position_embeddings = config.max_position_embeddings
|
| 191 |
+
self.is_causal = True
|
| 192 |
|
| 193 |
if (self.head_dim * self.num_heads) != self.hidden_size:
|
| 194 |
raise ValueError(
|
|
|
|
| 293 |
return attn_output, attn_weights, past_key_value
|
| 294 |
|
| 295 |
|
| 296 |
+
class FlashAttention2(Attention):
|
| 297 |
+
"""
|
| 298 |
+
Reference: https://github.com/huggingface/transformers/blob/5d36025ca13d05151b7a0c761e90d429c4644a30/src/transformers/models/llama/modeling_llama.py#L456
|
| 299 |
+
"""
|
| 300 |
+
|
| 301 |
+
def __init__(self, *args, **kwargs):
|
| 302 |
+
super().__init__(*args, **kwargs)
|
| 303 |
+
|
| 304 |
+
# TODO: Should be removed once Flash Attention for RoCm is bumped to 2.1.
|
| 305 |
+
# flash_attn<2.1 generates top-left aligned causal mask, while what is needed here is bottom-right alignement, that was made default for flash_attn>=2.1. This attribute is used to handle this difference. Reference: https://github.com/Dao-AILab/flash-attention/releases/tag/v2.1.0.
|
| 306 |
+
# Beware that with flash_attn<2.1, using q_seqlen != k_seqlen (except for the case q_seqlen == 1) produces a wrong mask (top-left).
|
| 307 |
+
self._flash_attn_uses_top_left_mask = not is_flash_attn_greater_or_equal_2_10()
|
| 308 |
+
|
| 309 |
+
def forward(
|
| 310 |
+
self,
|
| 311 |
+
hidden_states: torch.Tensor,
|
| 312 |
+
attention_mask: Optional[torch.LongTensor] = None,
|
| 313 |
+
position_ids: Optional[torch.LongTensor] = None,
|
| 314 |
+
past_key_value: Optional[Cache] = None,
|
| 315 |
+
output_attentions: bool = False,
|
| 316 |
+
use_cache: bool = False,
|
| 317 |
+
**kwargs,
|
| 318 |
+
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
|
| 319 |
+
# FlashAttention2 attention does not support output_attentions
|
| 320 |
+
if "padding_mask" in kwargs:
|
| 321 |
+
warnings.warn(
|
| 322 |
+
"Passing `padding_mask` is deprecated and will be removed in v4.37. Please make sure use `attention_mask` instead.`"
|
| 323 |
+
)
|
| 324 |
+
|
| 325 |
+
# overwrite attention_mask with padding_mask
|
| 326 |
+
attention_mask = kwargs.pop("padding_mask")
|
| 327 |
+
|
| 328 |
+
output_attentions = False
|
| 329 |
+
|
| 330 |
+
bsz, q_len, _ = hidden_states.size()
|
| 331 |
+
|
| 332 |
+
query_states = self.q_proj(hidden_states)
|
| 333 |
+
key_states = self.k_proj(hidden_states)
|
| 334 |
+
value_states = self.v_proj(hidden_states)
|
| 335 |
+
|
| 336 |
+
# Flash attention requires the input to have the shape
|
| 337 |
+
# batch_size x seq_length x head_dim x hidden_dim
|
| 338 |
+
# therefore we just need to keep the original shape
|
| 339 |
+
query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
|
| 340 |
+
key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
|
| 341 |
+
value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
|
| 342 |
+
|
| 343 |
+
query_rot = query_states[..., : self.rotary_ndims]
|
| 344 |
+
query_pass = query_states[..., self.rotary_ndims :]
|
| 345 |
+
key_rot = key_states[..., : self.rotary_ndims]
|
| 346 |
+
key_pass = key_states[..., self.rotary_ndims :]
|
| 347 |
+
|
| 348 |
+
kv_seq_len = key_states.shape[-2]
|
| 349 |
+
if past_key_value is not None:
|
| 350 |
+
kv_seq_len += past_key_value[0].shape[-2]
|
| 351 |
+
cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
|
| 352 |
+
query_states, key_states = apply_rotary_pos_emb(query_rot, key_rot, cos, sin, position_ids)
|
| 353 |
+
|
| 354 |
+
# [batch_size, num_heads, seq_len, head_dim]
|
| 355 |
+
query_states = torch.cat((query_states, query_pass), dim=-1)
|
| 356 |
+
key_states = torch.cat((key_states, key_pass), dim=-1)
|
| 357 |
+
|
| 358 |
+
if past_key_value is not None:
|
| 359 |
+
# Reuse k, v, self_attention
|
| 360 |
+
key_states = torch.cat((past_key_value[0], key_states), dim=2)
|
| 361 |
+
value_states = torch.cat((past_key_value[1], value_states), dim=2)
|
| 362 |
+
|
| 363 |
+
past_key_value = (key_states, value_states) if use_cache else None
|
| 364 |
+
|
| 365 |
+
# TODO: These transpose are quite inefficient but Flash Attention requires the layout [batch_size, sequence_length, num_heads, head_dim]. We would need to refactor the KV cache
|
| 366 |
+
# to be able to avoid many of these transpose/reshape/view.
|
| 367 |
+
query_states = query_states.transpose(1, 2)
|
| 368 |
+
key_states = key_states.transpose(1, 2)
|
| 369 |
+
value_states = value_states.transpose(1, 2)
|
| 370 |
+
|
| 371 |
+
dropout_rate = self.attention_dropout if self.training else 0.0
|
| 372 |
+
|
| 373 |
+
attn_output = self._flash_attention_forward(
|
| 374 |
+
query_states, key_states, value_states, attention_mask, q_len, dropout=dropout_rate
|
| 375 |
+
)
|
| 376 |
+
attn_output = attn_output.reshape(bsz, q_len, self.hidden_size).contiguous()
|
| 377 |
+
attn_output = self.o_proj(attn_output)
|
| 378 |
+
|
| 379 |
+
if not output_attentions:
|
| 380 |
+
attn_weights = None
|
| 381 |
+
|
| 382 |
+
return attn_output, attn_weights, past_key_value
|
| 383 |
+
|
| 384 |
+
def _flash_attention_forward(
|
| 385 |
+
self, query_states, key_states, value_states, attention_mask, query_length, dropout=0.0, softmax_scale=None
|
| 386 |
+
):
|
| 387 |
+
"""
|
| 388 |
+
Calls the forward method of Flash Attention - if the input hidden states contain at least one padding token
|
| 389 |
+
first unpad the input, then computes the attention scores and pad the final attention scores.
|
| 390 |
+
|
| 391 |
+
Args:
|
| 392 |
+
query_states (`torch.Tensor`):
|
| 393 |
+
Input query states to be passed to Flash Attention API
|
| 394 |
+
key_states (`torch.Tensor`):
|
| 395 |
+
Input key states to be passed to Flash Attention API
|
| 396 |
+
value_states (`torch.Tensor`):
|
| 397 |
+
Input value states to be passed to Flash Attention API
|
| 398 |
+
attention_mask (`torch.Tensor`):
|
| 399 |
+
The padding mask - corresponds to a tensor of size `(batch_size, seq_len)` where 0 stands for the
|
| 400 |
+
position of padding tokens and 1 for the position of non-padding tokens.
|
| 401 |
+
dropout (`int`, *optional*):
|
| 402 |
+
Attention dropout
|
| 403 |
+
softmax_scale (`float`, *optional*):
|
| 404 |
+
The scaling of QK^T before applying softmax. Default to 1 / sqrt(head_dim)
|
| 405 |
+
"""
|
| 406 |
+
if not self._flash_attn_uses_top_left_mask:
|
| 407 |
+
causal = self.is_causal
|
| 408 |
+
else:
|
| 409 |
+
# TODO: Remove the `query_length != 1` check once Flash Attention for RoCm is bumped to 2.1. For details, please see the comment in FlashAttention2 __init__.
|
| 410 |
+
causal = self.is_causal and query_length != 1
|
| 411 |
+
|
| 412 |
+
# Contains at least one padding token in the sequence
|
| 413 |
+
if attention_mask is not None:
|
| 414 |
+
batch_size = query_states.shape[0]
|
| 415 |
+
query_states, key_states, value_states, indices_q, cu_seq_lens, max_seq_lens = self._upad_input(
|
| 416 |
+
query_states, key_states, value_states, attention_mask, query_length
|
| 417 |
+
)
|
| 418 |
+
|
| 419 |
+
cu_seqlens_q, cu_seqlens_k = cu_seq_lens
|
| 420 |
+
max_seqlen_in_batch_q, max_seqlen_in_batch_k = max_seq_lens
|
| 421 |
+
|
| 422 |
+
attn_output_unpad = flash_attn_varlen_func(
|
| 423 |
+
query_states,
|
| 424 |
+
key_states,
|
| 425 |
+
value_states,
|
| 426 |
+
cu_seqlens_q=cu_seqlens_q,
|
| 427 |
+
cu_seqlens_k=cu_seqlens_k,
|
| 428 |
+
max_seqlen_q=max_seqlen_in_batch_q,
|
| 429 |
+
max_seqlen_k=max_seqlen_in_batch_k,
|
| 430 |
+
dropout_p=dropout,
|
| 431 |
+
softmax_scale=softmax_scale,
|
| 432 |
+
causal=causal,
|
| 433 |
+
)
|
| 434 |
+
|
| 435 |
+
attn_output = pad_input(attn_output_unpad, indices_q, batch_size, query_length)
|
| 436 |
+
else:
|
| 437 |
+
attn_output = flash_attn_func(
|
| 438 |
+
query_states, key_states, value_states, dropout, softmax_scale=softmax_scale, causal=causal
|
| 439 |
+
)
|
| 440 |
+
|
| 441 |
+
return attn_output
|
| 442 |
+
|
| 443 |
+
def _upad_input(self, query_layer, key_layer, value_layer, attention_mask, query_length):
|
| 444 |
+
indices_k, cu_seqlens_k, max_seqlen_in_batch_k = _get_unpad_data(attention_mask)
|
| 445 |
+
batch_size, kv_seq_len, num_key_value_heads, head_dim = key_layer.shape
|
| 446 |
+
|
| 447 |
+
key_layer = index_first_axis(
|
| 448 |
+
key_layer.reshape(batch_size * kv_seq_len, num_key_value_heads, head_dim), indices_k
|
| 449 |
+
)
|
| 450 |
+
value_layer = index_first_axis(
|
| 451 |
+
value_layer.reshape(batch_size * kv_seq_len, num_key_value_heads, head_dim), indices_k
|
| 452 |
+
)
|
| 453 |
+
if query_length == kv_seq_len:
|
| 454 |
+
query_layer = index_first_axis(
|
| 455 |
+
query_layer.reshape(batch_size * kv_seq_len, self.num_heads, head_dim), indices_k
|
| 456 |
+
)
|
| 457 |
+
cu_seqlens_q = cu_seqlens_k
|
| 458 |
+
max_seqlen_in_batch_q = max_seqlen_in_batch_k
|
| 459 |
+
indices_q = indices_k
|
| 460 |
+
elif query_length == 1:
|
| 461 |
+
max_seqlen_in_batch_q = 1
|
| 462 |
+
cu_seqlens_q = torch.arange(
|
| 463 |
+
batch_size + 1, dtype=torch.int32, device=query_layer.device
|
| 464 |
+
) # There is a memcpy here, that is very bad.
|
| 465 |
+
indices_q = cu_seqlens_q[:-1]
|
| 466 |
+
query_layer = query_layer.squeeze(1)
|
| 467 |
+
else:
|
| 468 |
+
# The -q_len: slice assumes left padding.
|
| 469 |
+
attention_mask = attention_mask[:, -query_length:]
|
| 470 |
+
query_layer, indices_q, cu_seqlens_q, max_seqlen_in_batch_q = unpad_input(query_layer, attention_mask)
|
| 471 |
+
|
| 472 |
+
return (
|
| 473 |
+
query_layer,
|
| 474 |
+
key_layer,
|
| 475 |
+
value_layer,
|
| 476 |
+
indices_q,
|
| 477 |
+
(cu_seqlens_q, cu_seqlens_k),
|
| 478 |
+
(max_seqlen_in_batch_q, max_seqlen_in_batch_k),
|
| 479 |
+
)
|
| 480 |
+
|
| 481 |
+
|
| 482 |
+
ATTENTION_CLASSES = {
|
| 483 |
+
"eager": Attention,
|
| 484 |
+
"flash_attention_2": FlashAttention2,
|
| 485 |
+
}
|
| 486 |
+
|
| 487 |
+
|
| 488 |
class DecoderLayer(nn.Module):
|
| 489 |
def __init__(self, config: StableLMEpochConfig):
|
| 490 |
super().__init__()
|
| 491 |
+
self.self_attn = ATTENTION_CLASSES[config._attn_implementation](config=config)
|
| 492 |
self.mlp = MLP(config)
|
| 493 |
self.input_layernorm = nn.LayerNorm(config.hidden_size, eps=config.norm_eps)
|
| 494 |
self.post_attention_layernorm = nn.LayerNorm(config.hidden_size, eps=config.norm_eps)
|
|
|
|
| 544 |
supports_gradient_checkpointing = True
|
| 545 |
_no_split_modules = ["DecoderLayer"]
|
| 546 |
_skip_keys_device_placement = "past_key_values"
|
| 547 |
+
_supports_flash_attn_2 = True
|
| 548 |
|
| 549 |
def _init_weights(self, module: nn.Module):
|
| 550 |
"""Initialize the weights"""
|
|
|
|
| 572 |
self.layers = nn.ModuleList([DecoderLayer(config) for _ in range(config.num_hidden_layers)])
|
| 573 |
self.norm = nn.LayerNorm(config.hidden_size, eps=config.norm_eps)
|
| 574 |
|
| 575 |
+
self._use_flash_attention_2 = config._attn_implementation == "flash_attention_2"
|
| 576 |
self.gradient_checkpointing = False
|
| 577 |
# Initialize weights and apply final processing
|
| 578 |
self.post_init()
|
|
|
|
| 646 |
seq_length_with_past = seq_length
|
| 647 |
past_key_values_length = 0
|
| 648 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 649 |
if position_ids is None:
|
| 650 |
device = input_ids.device if input_ids is not None else inputs_embeds.device
|
| 651 |
position_ids = torch.arange(
|
|
|
|
| 661 |
if inputs_embeds is None:
|
| 662 |
inputs_embeds = self.embed_tokens(input_ids)
|
| 663 |
# Embed positions
|
| 664 |
+
if self._use_flash_attention_2:
|
| 665 |
+
# 2d mask is passed through the layers
|
| 666 |
+
attention_mask = attention_mask if (attention_mask is not None and 0 in attention_mask) else None
|
| 667 |
+
else:
|
| 668 |
+
if attention_mask is None:
|
| 669 |
+
attention_mask = torch.ones(
|
| 670 |
+
(batch_size, seq_length_with_past),
|
| 671 |
+
dtype=torch.bool,
|
| 672 |
+
device=inputs_embeds.device,
|
| 673 |
+
)
|
| 674 |
+
attention_mask = self._prepare_decoder_attention_mask(
|
| 675 |
+
attention_mask,
|
| 676 |
+
(batch_size, seq_length),
|
| 677 |
+
inputs_embeds,
|
| 678 |
+
past_key_values_length,
|
| 679 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 680 |
|
| 681 |
hidden_states = inputs_embeds
|
| 682 |
|
|
|
|
| 861 |
**kwargs,
|
| 862 |
):
|
| 863 |
# Trim decoder_input_ids if past is used
|
| 864 |
+
if past_key_values is not None:
|
| 865 |
+
past_length = past_key_values[0][0].shape[2]
|
| 866 |
+
|
| 867 |
+
# Some generation methods already pass only the last input ID
|
| 868 |
+
if input_ids.shape[1] > past_length:
|
| 869 |
+
remove_prefix_length = past_length
|
| 870 |
+
else:
|
| 871 |
+
# Default to old behavior: keep only final ID
|
| 872 |
+
remove_prefix_length = input_ids.shape[1] - 1
|
| 873 |
+
|
| 874 |
+
input_ids = input_ids[:, remove_prefix_length:]
|
| 875 |
|
| 876 |
position_ids = kwargs.get("position_ids", None)
|
| 877 |
if attention_mask is not None and position_ids is None:
|