ros/LLaMA-Factory
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

merge data part to the text stream

Browse Source 
Former-commit-id: 7ee20286d9bcc2d5378bfd6bb02cd3648396d873
This commit is contained in:
BUAADreamer
2024-04-25 19:19:59 +08:00
parent 00e2a272ef
commit 3c792174db
13 changed files with 802 additions and 284 deletions
Download Patch File Download Diff File Expand all files Collapse all files

185
src/llmtuner/data/preprocess.py
View File

@@ -1,6 +1,6 @@
from functools import partial
from itertools import chain
from typing import TYPE_CHECKING, Any, Callable, Dict, List, Literal, Tuple
from typing import TYPE_CHECKING, Any, Callable, Dict, List, Literal, Tuple, Optional
from ..extras.constants import IGNORE_INDEX
from ..extras.logging import get_logger
@@ -9,7 +9,7 @@ from .utils import Role
if TYPE_CHECKING:
from transformers import Seq2SeqTrainingArguments
from transformers.tokenization_utils import PreTrainedTokenizer
from transformers.tokenization_utils import PreTrainedTokenizer, AutoProcessor
from ..hparams import DataArguments
from .template import Template
@@ -19,19 +19,27 @@ logger = get_logger(__name__)
def preprocess_pretrain_dataset(
examples: Dict[str, List[Any]], tokenizer: "PreTrainedTokenizer", data_args: "DataArguments"
examples: Dict[str, List[Any]],
tokenizer: "PreTrainedTokenizer",
data_args: "DataArguments",
) -> Dict[str, List[List[int]]]:
# build grouped texts with format `X1 X2 X3 ...` if packing is enabled
text_examples = [messages[0]["content"] + tokenizer.eos_token for messages in examples["prompt"]]
text_examples = [
messages[0]["content"] + tokenizer.eos_token for messages in examples["prompt"]
]
if not data_args.packing:
if data_args.template == "gemma":
text_examples = [tokenizer.bos_token + example for example in text_examples]
result = tokenizer(text_examples, add_special_tokens=False, max_length=data_args.cutoff_len)
result = tokenizer(
text_examples, add_special_tokens=False, max_length=data_args.cutoff_len
)
else:
tokenized_examples = tokenizer(text_examples, add_special_tokens=False)
concatenated_examples = {k: list(chain(*tokenized_examples[k])) for k in tokenized_examples.keys()}
concatenated_examples = {
k: list(chain(*tokenized_examples[k])) for k in tokenized_examples.keys()
}
total_length = len(concatenated_examples[list(concatenated_examples.keys())[0]])
block_size = data_args.cutoff_len
total_length = (total_length // block_size) * block_size
@@ -54,7 +62,11 @@ def preprocess_supervised_dataset(
) -> Dict[str, List[List[int]]]:
# build inputs with format `<bos> X Y <eos>` and labels with format `<ignore> ... <ignore> Y <eos>`
# for multiturn examples, we only mask the prompt part in each prompt-response pair.
model_inputs = {"input_ids": [], "attention_mask": [], "labels": []}
model_inputs = {
"input_ids": [],
"attention_mask": [],
"labels": [],
}
for i in range(len(examples["prompt"])):
if len(examples["prompt"][i]) % 2 != 1 or len(examples["response"][i]) != 1:
@@ -75,7 +87,9 @@ def preprocess_supervised_dataset(
if data_args.train_on_prompt:
source_mask = source_ids
elif turn_idx != 0 and template.efficient_eos:
source_mask = [tokenizer.eos_token_id] + [IGNORE_INDEX] * (len(source_ids) - 1)
source_mask = [tokenizer.eos_token_id] + [IGNORE_INDEX] * (
len(source_ids) - 1
)
else:
source_mask = [IGNORE_INDEX] * len(source_ids)
@@ -114,7 +128,9 @@ def preprocess_packed_supervised_dataset(
if data_args.train_on_prompt:
source_mask = source_ids
elif len(input_ids) != 0 and template.efficient_eos:
source_mask = [tokenizer.eos_token_id] + [IGNORE_INDEX] * (len(source_ids) - 1)
source_mask = [tokenizer.eos_token_id] + [IGNORE_INDEX] * (
len(source_ids) - 1
)
else:
source_mask = [IGNORE_INDEX] * len(source_ids)
@@ -139,6 +155,64 @@ def preprocess_packed_supervised_dataset(
return model_inputs
def preprocess_multimodal_supervised_dataset(
examples: Dict[str, List[Any]],
processor: "AutoProcessor",
template: "Template",
data_args: "DataArguments",
) -> Dict[str, List[List[int]]]:
# build inputs with format `<bos> X Y <eos>` and labels with format `<ignore> ... <ignore> Y <eos>`
# for multiturn examples, we only mask the prompt part in each prompt-response pair.
tokenizer = processor.tokenizer
model_inputs = {
"input_ids": [],
"attention_mask": [],
"labels": [],
"pixel_values": [],
}
for i in range(len(examples["prompt"])):
if len(examples["prompt"][i]) % 2 != 1 or len(examples["response"][i]) != 1:
continue
messages = examples["prompt"][i] + examples["response"][i]
input_ids, labels = [], []
for turn_idx, (source_ids, target_ids) in enumerate(
template.encode_multiturn(
tokenizer,
messages,
examples["system"][i],
examples["tools"][i],
data_args.cutoff_len,
data_args.reserved_label_len,
)
):
if data_args.train_on_prompt:
source_mask = source_ids
elif turn_idx != 0 and template.efficient_eos:
source_mask = [tokenizer.eos_token_id] + [IGNORE_INDEX] * (
len(source_ids) - 1
)
else:
source_mask = [IGNORE_INDEX] * len(source_ids)
input_ids += source_ids + target_ids
labels += source_mask + target_ids
if template.efficient_eos:
input_ids += [tokenizer.eos_token_id]
labels += [tokenizer.eos_token_id]
model_inputs["input_ids"].append(input_ids)
model_inputs["attention_mask"].append([1] * len(input_ids))
model_inputs["labels"].append(labels)
pixel_values = processor.image_processor(
examples["images"][0], return_tensors="pt"
)["pixel_values"][0]
model_inputs["pixel_values"].append(pixel_values)
return model_inputs
def preprocess_unsupervised_dataset(
examples: Dict[str, List[Any]],
tokenizer: "PreTrainedTokenizer",
@@ -155,7 +229,9 @@ def preprocess_unsupervised_dataset(
if len(examples["response"][i]) == 1:
messages = examples["prompt"][i] + examples["response"][i]
else:
messages = examples["prompt"][i] + [{"role": Role.ASSISTANT.value, "content": ""}]
messages = examples["prompt"][i] + [
{"role": Role.ASSISTANT.value, "content": ""}
]
input_ids, labels = template.encode_oneturn(
tokenizer,
@@ -218,29 +294,58 @@ def preprocess_pairwise_dataset(
return model_inputs
def print_supervised_dataset_example(example: Dict[str, List[int]], tokenizer: "PreTrainedTokenizer") -> None:
def print_supervised_dataset_example(
example: Dict[str, List[int]], tokenizer: "PreTrainedTokenizer"
) -> None:
print("input_ids:\n{}".format(example["input_ids"]))
print("inputs:\n{}".format(tokenizer.decode(example["input_ids"], skip_special_tokens=False)))
print(
"inputs:\n{}".format(
tokenizer.decode(example["input_ids"], skip_special_tokens=False)
)
)
print("label_ids:\n{}".format(example["labels"]))
print(
"labels:\n{}".format(
tokenizer.decode(list(filter(lambda x: x != IGNORE_INDEX, example["labels"])), skip_special_tokens=False)
tokenizer.decode(
list(filter(lambda x: x != IGNORE_INDEX, example["labels"])),
skip_special_tokens=False,
)
)
)
def print_pairwise_dataset_example(example: Dict[str, List[int]], tokenizer: "PreTrainedTokenizer") -> None:
def print_pairwise_dataset_example(
example: Dict[str, List[int]], tokenizer: "PreTrainedTokenizer"
) -> None:
print("prompt_ids:\n{}".format(example["prompt_ids"]))
print("prompt:\n{}".format(tokenizer.decode(example["prompt_ids"], skip_special_tokens=False)))
print(
"prompt:\n{}".format(
tokenizer.decode(example["prompt_ids"], skip_special_tokens=False)
)
)
print("chosen_ids:\n{}".format(example["chosen_ids"]))
print("chosen:\n{}".format(tokenizer.decode(example["chosen_ids"], skip_special_tokens=False)))
print(
"chosen:\n{}".format(
tokenizer.decode(example["chosen_ids"], skip_special_tokens=False)
)
)
print("rejected_ids:\n{}".format(example["rejected_ids"]))
print("rejected:\n{}".format(tokenizer.decode(example["rejected_ids"], skip_special_tokens=False)))
print(
"rejected:\n{}".format(
tokenizer.decode(example["rejected_ids"], skip_special_tokens=False)
)
)
def print_unsupervised_dataset_example(example: Dict[str, List[int]], tokenizer: "PreTrainedTokenizer") -> None:
def print_unsupervised_dataset_example(
example: Dict[str, List[int]], tokenizer: "PreTrainedTokenizer"
) -> None:
print("input_ids:\n{}".format(example["input_ids"]))
print("inputs:\n{}".format(tokenizer.decode(example["input_ids"], skip_special_tokens=False)))
print(
"inputs:\n{}".format(
tokenizer.decode(example["input_ids"], skip_special_tokens=False)
)
)
def get_preprocess_and_print_func(
@@ -249,30 +354,56 @@ def get_preprocess_and_print_func(
data_args: "DataArguments",
training_args: "Seq2SeqTrainingArguments",
stage: Literal["pt", "sft", "rm", "ppo"],
processor: Optional["AutoProcessor"] = None,
) -> Tuple[Callable, Callable]:
if stage == "pt":
preprocess_func = partial(preprocess_pretrain_dataset, tokenizer=tokenizer, data_args=data_args)
print_function = partial(print_unsupervised_dataset_example, tokenizer=tokenizer)
preprocess_func = partial(
preprocess_pretrain_dataset, tokenizer=tokenizer, data_args=data_args
)
print_function = partial(
print_unsupervised_dataset_example, tokenizer=tokenizer
)
elif stage == "sft" and not training_args.predict_with_generate:
if data_args.packing:
preprocess_func = partial(
preprocess_packed_supervised_dataset, tokenizer=tokenizer, template=template, data_args=data_args
preprocess_packed_supervised_dataset,
tokenizer=tokenizer,
template=template,
data_args=data_args,
)
elif processor is not None:
preprocess_func = partial(
preprocess_multimodal_supervised_dataset,
processor=processor,
template=template,
data_args=data_args,
)
else:
preprocess_func = partial(
preprocess_supervised_dataset, tokenizer=tokenizer, template=template, data_args=data_args
preprocess_supervised_dataset,
tokenizer=tokenizer,
template=template,
data_args=data_args,
)
print_function = partial(print_supervised_dataset_example, tokenizer=tokenizer)
elif stage == "rm":
preprocess_func = partial(
preprocess_pairwise_dataset, tokenizer=tokenizer, template=template, data_args=data_args
preprocess_pairwise_dataset,
tokenizer=tokenizer,
template=template,
data_args=data_args,
)
print_function = partial(print_pairwise_dataset_example, tokenizer=tokenizer)
else:
preprocess_func = partial(
preprocess_unsupervised_dataset, tokenizer=tokenizer, template=template, data_args=data_args
preprocess_unsupervised_dataset,
tokenizer=tokenizer,
template=template,
data_args=data_args,
)
print_function = partial(
print_unsupervised_dataset_example, tokenizer=tokenizer
)
print_function = partial(print_unsupervised_dataset_example, tokenizer=tokenizer)
return preprocess_func, print_function
return preprocess_func, print_function