make sure padding is labeled as -100 for pretraining (#2227)
This commit is contained in:
@@ -28,8 +28,10 @@ def encode_pretraining(
|
|||||||
)
|
)
|
||||||
# Convert to PyTorch tensors
|
# Convert to PyTorch tensors
|
||||||
input_ids = [torch.tensor(seq) for seq in res["input_ids"]]
|
input_ids = [torch.tensor(seq) for seq in res["input_ids"]]
|
||||||
|
targets = [torch.tensor(seq) for seq in res["input_ids"]]
|
||||||
attention_mask = [torch.tensor(seq) for seq in res["attention_mask"]]
|
attention_mask = [torch.tensor(seq) for seq in res["attention_mask"]]
|
||||||
new_input_ids = []
|
new_input_ids = []
|
||||||
|
new_labels = []
|
||||||
new_attention_mask = []
|
new_attention_mask = []
|
||||||
# Append EOS and PAD tokens to input_ids, and correct attention_mask
|
# Append EOS and PAD tokens to input_ids, and correct attention_mask
|
||||||
for i, _ in enumerate(input_ids):
|
for i, _ in enumerate(input_ids):
|
||||||
@@ -40,22 +42,34 @@ def encode_pretraining(
|
|||||||
),
|
),
|
||||||
dim=0,
|
dim=0,
|
||||||
)
|
)
|
||||||
|
targets[i] = torch.cat(
|
||||||
|
(
|
||||||
|
targets[i],
|
||||||
|
torch.tensor([tokenizer.eos_token_id, -100]),
|
||||||
|
),
|
||||||
|
dim=0,
|
||||||
|
)
|
||||||
attention_mask[i] = torch.cat((attention_mask[i], torch.tensor([1, 0])), dim=0)
|
attention_mask[i] = torch.cat((attention_mask[i], torch.tensor([1, 0])), dim=0)
|
||||||
|
|
||||||
# Concatenate tokens so that their lengths are less than max_tokens
|
# Concatenate tokens so that their lengths are less than max_tokens
|
||||||
buffer_input_ids = torch.tensor([], dtype=torch.long)
|
buffer_input_ids = torch.tensor([], dtype=torch.long)
|
||||||
|
buffer_labels = torch.tensor([], dtype=torch.long)
|
||||||
buffer_attention_mask = torch.tensor([], dtype=torch.long)
|
buffer_attention_mask = torch.tensor([], dtype=torch.long)
|
||||||
|
|
||||||
for ids, mask in zip(input_ids, attention_mask):
|
for ids, labels, mask in zip(input_ids, targets, attention_mask):
|
||||||
if buffer_input_ids.numel() == max_tokens:
|
if buffer_input_ids.numel() == max_tokens:
|
||||||
new_input_ids.append(buffer_input_ids)
|
new_input_ids.append(buffer_input_ids)
|
||||||
|
new_labels.append(buffer_labels)
|
||||||
new_attention_mask.append(buffer_attention_mask)
|
new_attention_mask.append(buffer_attention_mask)
|
||||||
buffer_input_ids = torch.tensor([], dtype=torch.long)
|
buffer_input_ids = torch.tensor([], dtype=torch.long)
|
||||||
|
buffer_labels = torch.tensor([], dtype=torch.long)
|
||||||
buffer_attention_mask = torch.tensor([], dtype=torch.long)
|
buffer_attention_mask = torch.tensor([], dtype=torch.long)
|
||||||
buffer_input_ids = torch.cat((buffer_input_ids, ids), dim=0)
|
buffer_input_ids = torch.cat((buffer_input_ids, ids), dim=0)
|
||||||
|
buffer_labels = torch.cat((buffer_labels, labels), dim=0)
|
||||||
buffer_attention_mask = torch.cat((buffer_attention_mask, mask), dim=0)
|
buffer_attention_mask = torch.cat((buffer_attention_mask, mask), dim=0)
|
||||||
elif buffer_input_ids.numel() + ids.numel() <= max_tokens:
|
elif buffer_input_ids.numel() + ids.numel() <= max_tokens:
|
||||||
buffer_input_ids = torch.cat((buffer_input_ids, ids), dim=0)
|
buffer_input_ids = torch.cat((buffer_input_ids, ids), dim=0)
|
||||||
|
buffer_labels = torch.cat((buffer_labels, labels), dim=0)
|
||||||
buffer_attention_mask = torch.cat((buffer_attention_mask, mask), dim=0)
|
buffer_attention_mask = torch.cat((buffer_attention_mask, mask), dim=0)
|
||||||
else:
|
else:
|
||||||
buffer_input_ids = torch.cat(
|
buffer_input_ids = torch.cat(
|
||||||
@@ -69,6 +83,17 @@ def encode_pretraining(
|
|||||||
),
|
),
|
||||||
dim=0,
|
dim=0,
|
||||||
)
|
)
|
||||||
|
buffer_labels = torch.cat(
|
||||||
|
(
|
||||||
|
buffer_labels,
|
||||||
|
torch.full(
|
||||||
|
(max_tokens - buffer_labels.numel(),),
|
||||||
|
-100,
|
||||||
|
dtype=torch.long,
|
||||||
|
),
|
||||||
|
),
|
||||||
|
dim=0,
|
||||||
|
)
|
||||||
buffer_attention_mask = torch.cat(
|
buffer_attention_mask = torch.cat(
|
||||||
(
|
(
|
||||||
buffer_attention_mask,
|
buffer_attention_mask,
|
||||||
@@ -81,11 +106,14 @@ def encode_pretraining(
|
|||||||
dim=0,
|
dim=0,
|
||||||
)
|
)
|
||||||
new_input_ids.append(buffer_input_ids)
|
new_input_ids.append(buffer_input_ids)
|
||||||
|
new_labels.append(buffer_labels)
|
||||||
new_attention_mask.append(buffer_attention_mask)
|
new_attention_mask.append(buffer_attention_mask)
|
||||||
buffer_input_ids = torch.tensor([], dtype=torch.long)
|
buffer_input_ids = torch.tensor([], dtype=torch.long)
|
||||||
|
buffer_labels = torch.tensor([], dtype=torch.long)
|
||||||
buffer_attention_mask = torch.tensor([], dtype=torch.long)
|
buffer_attention_mask = torch.tensor([], dtype=torch.long)
|
||||||
|
|
||||||
buffer_input_ids = torch.cat((buffer_input_ids, ids), dim=0)
|
buffer_input_ids = torch.cat((buffer_input_ids, ids), dim=0)
|
||||||
|
buffer_labels = torch.cat((buffer_labels, labels), dim=0)
|
||||||
buffer_attention_mask = torch.cat((buffer_attention_mask, mask), dim=0)
|
buffer_attention_mask = torch.cat((buffer_attention_mask, mask), dim=0)
|
||||||
|
|
||||||
if buffer_input_ids.numel() > 0: # for any leftover tokens
|
if buffer_input_ids.numel() > 0: # for any leftover tokens
|
||||||
@@ -101,6 +129,17 @@ def encode_pretraining(
|
|||||||
),
|
),
|
||||||
dim=0,
|
dim=0,
|
||||||
)
|
)
|
||||||
|
buffer_labels = torch.cat(
|
||||||
|
(
|
||||||
|
buffer_labels,
|
||||||
|
torch.full(
|
||||||
|
(max_tokens - buffer_labels.numel(),),
|
||||||
|
-100,
|
||||||
|
dtype=torch.long,
|
||||||
|
),
|
||||||
|
),
|
||||||
|
dim=0,
|
||||||
|
)
|
||||||
buffer_attention_mask = torch.cat(
|
buffer_attention_mask = torch.cat(
|
||||||
(
|
(
|
||||||
buffer_attention_mask,
|
buffer_attention_mask,
|
||||||
@@ -113,11 +152,12 @@ def encode_pretraining(
|
|||||||
dim=0,
|
dim=0,
|
||||||
)
|
)
|
||||||
new_input_ids.append(buffer_input_ids)
|
new_input_ids.append(buffer_input_ids)
|
||||||
|
new_labels.append(buffer_labels)
|
||||||
new_attention_mask.append(buffer_attention_mask)
|
new_attention_mask.append(buffer_attention_mask)
|
||||||
|
|
||||||
ret = {
|
ret = {
|
||||||
"input_ids": [seq.tolist() for seq in new_input_ids],
|
"input_ids": [seq.tolist() for seq in new_input_ids],
|
||||||
"labels": [seq.tolist() for seq in new_input_ids],
|
"labels": [seq.tolist() for seq in new_labels],
|
||||||
"attention_mask": [seq.tolist() for seq in new_attention_mask],
|
"attention_mask": [seq.tolist() for seq in new_attention_mask],
|
||||||
}
|
}
|
||||||
|
|
||||||
|
|||||||
Reference in New Issue
Block a user