puru22 commited on
Commit
1e9ac66
1 Parent(s): 1e7fdcc

Changes in modelling_RW.py to be able to handle past_key_values for faster model generations

Browse files

The current code has missed out passing past_key_values in every forward pass for fast generation of tokens. This results in lot of recompute. This "modelling_RW.py" I am uploading deals with this in the way pytorch huggingface transformers package generation/utils.py wants. All the changes are basically around including past_key_values everywhere. I think this will apply on all falcon models These are the changes specifically

Class RotaryEmbedding forward method
Include past_seq_length in forward pass and apply rotary embedding according to the position of the query token ---- if else condition added

_make_causal_mask function
to give masking according to the way F.scaled dot product attention behaves. F.scaled_dot_product attention treats the attention_mask matrix as receiving attentions. For example if attention_mask is
[[True, False], [True, True]]. It would mean the first token is "receiving" attentions from first token and not second token. This is unlike what we generally end up thinking which is first token is giving attention to itself and not to the second one. Due to reason the past_key_values attentions are all True in make_causal mask function. Also I have reversed the inequality above that due to the same reason.

Class Attention forward method
a) past_key_value length is passed in rotary function ---- if,else loop added
b) concatenation of past key and current key is done after permuting the past key shape to match the current key shape
c) to keep key_layer shape consistent with the output expectation which is (batch_size, head_dim, seq_length), another permutation done before creating "present" to return in the output

Class RWModel prepare_attn_mask method
Have removed src_length > 1 criteria for making causal mask

RW causal LM prepare inputs for generation
Read pastkey values from the input coming from huggingface generate method and dont call convert_to_rw_cache method

Files changed (1) hide show
  1. modelling_RW.py +42 -19
modelling_RW.py CHANGED
@@ -87,10 +87,18 @@ class RotaryEmbedding(torch.nn.Module):
87
 
88
  return self.cos_cached, self.sin_cached
89
 
90
- def forward(self, q, k):
91
- batch, seq_len, head_dim = q.shape
 
 
 
 
 
92
  cos, sin = self.cos_sin(seq_len, q.device, q.dtype)
93
- return (q * cos) + (rotate_half(q) * sin), (k * cos) + (rotate_half(k) * sin)
 
 
 
94
 
95
 
96
  def _make_causal_mask(
@@ -100,10 +108,10 @@ def _make_causal_mask(
100
  mask = torch.empty((target_length, target_length + past_key_values_length), dtype=torch.bool, device=device)
101
  # ONNX doesn't support `torch.Tensor.triu` properly, thus we use this workaround
102
  seq_ids = torch.arange(target_length, device=device)
103
- mask[:, past_key_values_length:] = seq_ids[:, None] < seq_ids[None, :]
104
 
105
  if past_key_values_length > 0:
106
- mask[:, :past_key_values_length] = False
107
 
108
  expanded_mask = mask[None, None, :, :].expand(batch_size, 1, target_length, target_length + past_key_values_length)
109
  return expanded_mask
@@ -264,20 +272,27 @@ class Attention(nn.Module):
264
  )
265
  value_layer = value_layer.transpose(1, 2).reshape(batch_size * self.num_heads, q_length, self.head_dim)
266
 
267
- query_layer, key_layer = self.maybe_rotary(query_layer, key_layer)
 
 
 
 
 
268
 
269
  if layer_past is not None:
270
  past_key, past_value = layer_past
271
  # concatenate along seq_length dimension:
272
  # - key: [batch_size * self.num_heads, head_dim, kv_length]
273
  # - value: [batch_size * self.num_heads, kv_length, head_dim]
 
274
  key_layer = torch.cat((past_key, key_layer), dim=1)
275
  value_layer = torch.cat((past_value, value_layer), dim=1)
276
 
277
  _, kv_length, _ = key_layer.shape
278
 
279
  if use_cache is True:
280
- present = (key_layer, value_layer)
 
281
  else:
282
  present = None
283
 
@@ -286,9 +301,14 @@ class Attention(nn.Module):
286
  key_layer_ = key_layer.reshape(batch_size, self.num_heads, -1, self.head_dim)
287
  value_layer_ = value_layer.reshape(batch_size, self.num_heads, -1, self.head_dim)
288
 
289
- attn_output = F.scaled_dot_product_attention(
290
- query_layer_, key_layer_, value_layer_, None, 0.0, is_causal=True
291
- )
 
 
 
 
 
292
 
293
  x = attn_output.view(batch_size, self.num_heads, q_length, self.head_dim)
294
  x = x.permute(0, 2, 1, 3)
@@ -528,10 +548,10 @@ class RWModel(RWPreTrainedModel):
528
  device = attention_mask.device
529
  _, src_length = input_shape
530
 
531
- if src_length > 1:
532
- combined_attention_mask = _make_causal_mask(
533
- input_shape, device=device, past_key_values_length=past_key_values_length
534
- )
535
 
536
  # [batch_size, seq_length] -> [batch_size, 1, tgt_length, src_length]
537
  expanded_attn_mask = _expand_mask(attention_mask, tgt_length=src_length)
@@ -710,16 +730,19 @@ class RWForCausalLM(RWPreTrainedModel):
710
  **kwargs,
711
  ) -> dict:
712
  # only last token for input_ids if past is not None
713
- if past:
714
  input_ids = input_ids[:, -1].unsqueeze(-1)
715
-
716
  # the cache may be in the stardard format (e.g. in contrastive search), convert to our's format if needed
717
- if past[0][0].shape[0] == input_ids.shape[0]:
718
- past = self._convert_to_rw_cache(past)
 
 
 
719
 
720
  return {
721
  "input_ids": input_ids,
722
- "past_key_values": past,
723
  "use_cache": kwargs.get("use_cache"),
724
  "attention_mask": attention_mask,
725
  }
 
87
 
88
  return self.cos_cached, self.sin_cached
89
 
90
+ def forward(self, q, k, past_seq_length=None):
91
+ if past_seq_length == None :
92
+ batch, seq_len, head_dim = q.shape
93
+ else :
94
+ # print("past_seq_length", past_seq_length)
95
+ batch, input_seq_len, head_dim = q.shape
96
+ seq_len = past_seq_length + input_seq_len
97
  cos, sin = self.cos_sin(seq_len, q.device, q.dtype)
98
+ if past_seq_length != None :
99
+ return (q * cos[:, past_seq_length:, :]) + (rotate_half(q) * sin[:, past_seq_length:, :]), (k * cos[:, past_seq_length:, :]) + (rotate_half(k) * sin[:, past_seq_length:, :])
100
+ else :
101
+ return (q * cos) + (rotate_half(q) * sin), (k * cos) + (rotate_half(k) * sin)
102
 
103
 
104
  def _make_causal_mask(
 
108
  mask = torch.empty((target_length, target_length + past_key_values_length), dtype=torch.bool, device=device)
109
  # ONNX doesn't support `torch.Tensor.triu` properly, thus we use this workaround
110
  seq_ids = torch.arange(target_length, device=device)
111
+ mask[:, past_key_values_length:] = seq_ids[:, None] >= seq_ids[None, :]
112
 
113
  if past_key_values_length > 0:
114
+ mask[:, :past_key_values_length] = True
115
 
116
  expanded_mask = mask[None, None, :, :].expand(batch_size, 1, target_length, target_length + past_key_values_length)
117
  return expanded_mask
 
272
  )
273
  value_layer = value_layer.transpose(1, 2).reshape(batch_size * self.num_heads, q_length, self.head_dim)
274
 
275
+ if layer_past is not None :
276
+ past_key, past_value = layer_past
277
+ past_kv_length = past_key.shape[2]
278
+ query_layer, key_layer = self.maybe_rotary(query_layer, key_layer, past_kv_length)
279
+ else :
280
+ query_layer, key_layer = self.maybe_rotary(query_layer, key_layer)
281
 
282
  if layer_past is not None:
283
  past_key, past_value = layer_past
284
  # concatenate along seq_length dimension:
285
  # - key: [batch_size * self.num_heads, head_dim, kv_length]
286
  # - value: [batch_size * self.num_heads, kv_length, head_dim]
287
+ past_key = past_key.permute(0, 2, 1)
288
  key_layer = torch.cat((past_key, key_layer), dim=1)
289
  value_layer = torch.cat((past_value, value_layer), dim=1)
290
 
291
  _, kv_length, _ = key_layer.shape
292
 
293
  if use_cache is True:
294
+ key_layer_permute = key_layer.permute(0, 2, 1)
295
+ present = (key_layer_permute, value_layer)
296
  else:
297
  present = None
298
 
 
301
  key_layer_ = key_layer.reshape(batch_size, self.num_heads, -1, self.head_dim)
302
  value_layer_ = value_layer.reshape(batch_size, self.num_heads, -1, self.head_dim)
303
 
304
+ if attention_mask is not None :
305
+ attn_output = F.scaled_dot_product_attention(
306
+ query_layer_, key_layer_, value_layer_, attention_mask, 0.0, is_causal=False
307
+ )
308
+ else :
309
+ attn_output = F.scaled_dot_product_attention(
310
+ query_layer_, key_layer_, value_layer_, None, 0.0, is_causal=True
311
+ )
312
 
313
  x = attn_output.view(batch_size, self.num_heads, q_length, self.head_dim)
314
  x = x.permute(0, 2, 1, 3)
 
548
  device = attention_mask.device
549
  _, src_length = input_shape
550
 
551
+ # if src_length > 1:
552
+ combined_attention_mask = _make_causal_mask(
553
+ input_shape, device=device, past_key_values_length=past_key_values_length
554
+ )
555
 
556
  # [batch_size, seq_length] -> [batch_size, 1, tgt_length, src_length]
557
  expanded_attn_mask = _expand_mask(attention_mask, tgt_length=src_length)
 
730
  **kwargs,
731
  ) -> dict:
732
  # only last token for input_ids if past is not None
733
+ if kwargs.get("past_key_values", None) :
734
  input_ids = input_ids[:, -1].unsqueeze(-1)
735
+ past_key_values = kwargs["past_key_values"]
736
  # the cache may be in the stardard format (e.g. in contrastive search), convert to our's format if needed
737
+ # if kwargs["past_key_values"][0][0].shape[0] == input_ids.shape[0]:
738
+ # past_key_values = self._convert_to_rw_cache(kwargs["past_key_values"])
739
+ # past_key_values = kwargs["past_key_values"]
740
+ else :
741
+ past_key_values = None
742
 
743
  return {
744
  "input_ids": input_ids,
745
+ "past_key_values": past_key_values,
746
  "use_cache": kwargs.get("use_cache"),
747
  "attention_mask": attention_mask,
748
  }