Efficiently get the length of the tokenized docs (#1063)
Browse files* Efficiently get the length of the tokenized docs
* chore: lint
---------
Co-authored-by: Wing Lian <wing.lian@gmail.com>
src/axolotl/core/trainer_builder.py
CHANGED
@@ -37,7 +37,7 @@ from axolotl.utils.collators import (
|
|
37 |
DataCollatorForSeq2Seq,
|
38 |
MambaDataCollator,
|
39 |
)
|
40 |
-
from axolotl.utils.samplers import MultipackBatchSampler
|
41 |
from axolotl.utils.schedulers import get_cosine_schedule_with_quadratic_warmup
|
42 |
|
43 |
try:
|
@@ -170,12 +170,7 @@ class AxolotlTrainer(Trainer):
|
|
170 |
self.args.train_batch_size,
|
171 |
drop_last=True,
|
172 |
batch_max_len=self._train_batch_size * self.args.max_seq_length,
|
173 |
-
lengths=(
|
174 |
-
self.train_dataset.data.column("position_ids")
|
175 |
-
.to_pandas()
|
176 |
-
.apply(lambda x: x[-1] + 1)
|
177 |
-
.values
|
178 |
-
),
|
179 |
packing_efficiency_estimate=self.args.sample_packing_efficiency,
|
180 |
)
|
181 |
return super()._get_train_sampler()
|
@@ -189,12 +184,7 @@ class AxolotlTrainer(Trainer):
|
|
189 |
self.args.per_device_eval_batch_size,
|
190 |
drop_last=True,
|
191 |
batch_max_len=self.args.eval_batch_size * self.args.max_seq_length,
|
192 |
-
lengths=(
|
193 |
-
eval_dataset.data.column("position_ids")
|
194 |
-
.to_pandas()
|
195 |
-
.apply(lambda x: x[-1] + 1)
|
196 |
-
.values
|
197 |
-
),
|
198 |
packing_efficiency_estimate=self.args.sample_packing_efficiency,
|
199 |
)
|
200 |
return super()._get_eval_sampler(eval_dataset)
|
|
|
37 |
DataCollatorForSeq2Seq,
|
38 |
MambaDataCollator,
|
39 |
)
|
40 |
+
from axolotl.utils.samplers import MultipackBatchSampler, get_dataset_lengths
|
41 |
from axolotl.utils.schedulers import get_cosine_schedule_with_quadratic_warmup
|
42 |
|
43 |
try:
|
|
|
170 |
self.args.train_batch_size,
|
171 |
drop_last=True,
|
172 |
batch_max_len=self._train_batch_size * self.args.max_seq_length,
|
173 |
+
lengths=get_dataset_lengths(self.train_dataset),
|
|
|
|
|
|
|
|
|
|
|
174 |
packing_efficiency_estimate=self.args.sample_packing_efficiency,
|
175 |
)
|
176 |
return super()._get_train_sampler()
|
|
|
184 |
self.args.per_device_eval_batch_size,
|
185 |
drop_last=True,
|
186 |
batch_max_len=self.args.eval_batch_size * self.args.max_seq_length,
|
187 |
+
lengths=get_dataset_lengths(eval_dataset),
|
|
|
|
|
|
|
|
|
|
|
188 |
packing_efficiency_estimate=self.args.sample_packing_efficiency,
|
189 |
)
|
190 |
return super()._get_eval_sampler(eval_dataset)
|
src/axolotl/utils/data.py
CHANGED
@@ -44,7 +44,7 @@ from axolotl.prompters import (
|
|
44 |
from axolotl.utils.collators import PretrainingBatchSamplerDataCollatorForSeq2Seq
|
45 |
from axolotl.utils.dict import DictDefault
|
46 |
from axolotl.utils.distributed import is_main_process, zero_first
|
47 |
-
from axolotl.utils.samplers
|
48 |
from axolotl.utils.trainer import (
|
49 |
calculate_total_num_steps,
|
50 |
process_datasets_for_packing,
|
@@ -889,12 +889,7 @@ def encode_packed_pretraining(
|
|
889 |
batch_size=batch_size,
|
890 |
drop_last=True,
|
891 |
batch_max_len=batch_size * max_seq_length,
|
892 |
-
lengths=(
|
893 |
-
train_dataset.data.column("position_ids")
|
894 |
-
.to_pandas()
|
895 |
-
.apply(lambda x: x[-1] + 1)
|
896 |
-
.values
|
897 |
-
),
|
898 |
)
|
899 |
|
900 |
chunked_data = defaultdict(list)
|
|
|
44 |
from axolotl.utils.collators import PretrainingBatchSamplerDataCollatorForSeq2Seq
|
45 |
from axolotl.utils.dict import DictDefault
|
46 |
from axolotl.utils.distributed import is_main_process, zero_first
|
47 |
+
from axolotl.utils.samplers import MultipackBatchSampler, get_dataset_lengths
|
48 |
from axolotl.utils.trainer import (
|
49 |
calculate_total_num_steps,
|
50 |
process_datasets_for_packing,
|
|
|
889 |
batch_size=batch_size,
|
890 |
drop_last=True,
|
891 |
batch_max_len=batch_size * max_seq_length,
|
892 |
+
lengths=get_dataset_lengths(train_dataset),
|
|
|
|
|
|
|
|
|
|
|
893 |
)
|
894 |
|
895 |
chunked_data = defaultdict(list)
|
src/axolotl/utils/samplers/__init__.py
CHANGED
@@ -2,3 +2,4 @@
|
|
2 |
axolotl samplers module
|
3 |
"""
|
4 |
from .multipack import MultipackBatchSampler # noqa: F401
|
|
|
|
2 |
axolotl samplers module
|
3 |
"""
|
4 |
from .multipack import MultipackBatchSampler # noqa: F401
|
5 |
+
from .utils import get_dataset_lengths # noqa: F401
|
src/axolotl/utils/samplers/utils.py
ADDED
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
"""
|
2 |
+
helper util to calculate dataset lengths
|
3 |
+
"""
|
4 |
+
import numpy as np
|
5 |
+
|
6 |
+
|
7 |
+
def get_dataset_lengths(dataset):
|
8 |
+
if "length" in dataset.data.column_names:
|
9 |
+
lengths = np.array(dataset.data.column("length"))
|
10 |
+
else:
|
11 |
+
lengths = (
|
12 |
+
dataset.data.column("position_ids")
|
13 |
+
.to_pandas()
|
14 |
+
.apply(lambda x: x[-1] + 1)
|
15 |
+
.values
|
16 |
+
)
|
17 |
+
return lengths
|
src/axolotl/utils/trainer.py
CHANGED
@@ -14,7 +14,7 @@ from torch.utils.data import DataLoader, RandomSampler
|
|
14 |
|
15 |
from axolotl.core.trainer_builder import HFCausalTrainerBuilder, HFDPOTrainerBuilder
|
16 |
from axolotl.utils.distributed import is_main_process, reduce_and_broadcast, zero_first
|
17 |
-
from axolotl.utils.samplers import MultipackBatchSampler
|
18 |
|
19 |
LOG = get_logger("axolotl")
|
20 |
|
@@ -212,12 +212,7 @@ def calculate_total_num_steps(cfg, train_dataset, update=True):
|
|
212 |
drop_last=True,
|
213 |
batch_max_len=cfg.micro_batch_size
|
214 |
* (cfg.max_packed_sequence_len or cfg.sequence_len),
|
215 |
-
lengths=(
|
216 |
-
train_dataset.data.column("position_ids")
|
217 |
-
.to_pandas()
|
218 |
-
.apply(lambda x: x[-1] + 1)
|
219 |
-
.values
|
220 |
-
),
|
221 |
)
|
222 |
|
223 |
data_loader = DataLoader(
|
|
|
14 |
|
15 |
from axolotl.core.trainer_builder import HFCausalTrainerBuilder, HFDPOTrainerBuilder
|
16 |
from axolotl.utils.distributed import is_main_process, reduce_and_broadcast, zero_first
|
17 |
+
from axolotl.utils.samplers import MultipackBatchSampler, get_dataset_lengths
|
18 |
|
19 |
LOG = get_logger("axolotl")
|
20 |
|
|
|
212 |
drop_last=True,
|
213 |
batch_max_len=cfg.micro_batch_size
|
214 |
* (cfg.max_packed_sequence_len or cfg.sequence_len),
|
215 |
+
lengths=get_dataset_lengths(train_dataset),
|
|
|
|
|
|
|
|
|
|
|
216 |
)
|
217 |
|
218 |
data_loader = DataLoader(
|