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

3
src/llmtuner/data/__init__.py
View File

@@ -1,12 +1,11 @@
from .collator import PairwiseDataCollatorWithPadding
from .loader import get_dataset, get_mm_dataset
from .loader import get_dataset
from .template import Template, get_template_and_fix_tokenizer, templates
from .utils import Role, split_dataset
__all__ = [
"PairwiseDataCollatorWithPadding",
"get_dataset",
"get_mm_dataset",
"Template",
"get_template_and_fix_tokenizer",
"templates",

118
src/llmtuner/data/aligner.py
View File

@@ -13,7 +13,9 @@ if TYPE_CHECKING:
from .parser import DatasetAttr
def convert_alpaca(examples: Dict[str, List[Any]], dataset_attr: "DatasetAttr") -> Dict[str, List[Any]]:
def convert_alpaca(
examples: Dict[str, List[Any]], dataset_attr: "DatasetAttr"
) -> Dict[str, List[Any]]:
outputs = {"prompt": [], "response": [], "system": [], "tools": []}
for i in range(len(examples[dataset_attr.prompt])):
prompt = []
@@ -31,24 +33,38 @@ def convert_alpaca(examples: Dict[str, List[Any]], dataset_attr: "DatasetAttr")
prompt.append({"role": Role.USER.value, "content": "\n".join(content)})
if dataset_attr.response and isinstance(examples[dataset_attr.response][i], list):
if dataset_attr.response and isinstance(
examples[dataset_attr.response][i], list
):
response = [
{"role": Role.ASSISTANT.value, "content": content} for content in examples[dataset_attr.response][i]
{"role": Role.ASSISTANT.value, "content": content}
for content in examples[dataset_attr.response][i]
]
elif dataset_attr.response and isinstance(
examples[dataset_attr.response][i], str
):
response = [
{
"role": Role.ASSISTANT.value,
"content": examples[dataset_attr.response][i],
}
]
elif dataset_attr.response and isinstance(examples[dataset_attr.response][i], str):
response = [{"role": Role.ASSISTANT.value, "content": examples[dataset_attr.response][i]}]
else:
response = []
outputs["prompt"].append(prompt)
outputs["response"].append(response)
outputs["system"].append(examples[dataset_attr.system][i] if dataset_attr.system else "")
outputs["system"].append(
examples[dataset_attr.system][i] if dataset_attr.system else ""
)
outputs["tools"].append("")
outputs["images"].append([])
return outputs
def convert_sharegpt(examples: Dict[str, List[Any]], dataset_attr: "DatasetAttr") -> Dict[str, List[Any]]:
def convert_sharegpt(
examples: Dict[str, List[Any]], dataset_attr: "DatasetAttr"
) -> Dict[str, List[Any]]:
outputs = {"prompt": [], "response": [], "system": [], "tools": []}
tag_mapping = {
dataset_attr.user_tag: Role.USER.value,
@@ -61,7 +77,10 @@ def convert_sharegpt(examples: Dict[str, List[Any]], dataset_attr: "DatasetAttr"
even_tags = (dataset_attr.assistant_tag, dataset_attr.function_tag)
accept_tags = (odd_tags, even_tags)
for i, messages in enumerate(examples[dataset_attr.messages]):
if dataset_attr.system_tag and messages[0][dataset_attr.role_tag] == dataset_attr.system_tag:
if (
dataset_attr.system_tag
and messages[0][dataset_attr.role_tag] == dataset_attr.system_tag
):
system = messages[0][dataset_attr.content_tag]
messages = messages[1:]
else:
@@ -77,19 +96,81 @@ def convert_sharegpt(examples: Dict[str, List[Any]], dataset_attr: "DatasetAttr"
raise ValueError("Invalid role tag in {}.".format(messages))
aligned_messages.append(
{"role": tag_mapping[message[dataset_attr.role_tag]], "content": message[dataset_attr.content_tag]}
{
"role": tag_mapping[message[dataset_attr.role_tag]],
"content": message[dataset_attr.content_tag],
}
)
outputs["prompt"].append(aligned_messages[:-1])
outputs["response"].append(aligned_messages[-1:])
outputs["system"].append(system)
outputs["tools"].append(examples[dataset_attr.tools][i] if dataset_attr.tools else "")
outputs["tools"].append(
examples[dataset_attr.tools][i] if dataset_attr.tools else ""
)
outputs["images"].append([])
return outputs
def convert_llava(
examples: Dict[str, List[Any]], dataset_attr: "DatasetAttr"
) -> Dict[str, List[Any]]:
outputs = {"prompt": [], "response": [], "system": [], "tools": [], "images": []}
tag_mapping = {
dataset_attr.user_tag: Role.USER.value,
dataset_attr.assistant_tag: Role.ASSISTANT.value,
dataset_attr.observation_tag: Role.OBSERVATION.value,
dataset_attr.function_tag: Role.FUNCTION.value,
dataset_attr.system_tag: Role.SYSTEM.value,
}
odd_tags = (dataset_attr.user_tag, dataset_attr.observation_tag)
even_tags = (dataset_attr.assistant_tag, dataset_attr.function_tag)
accept_tags = (odd_tags, even_tags)
for i, messages in enumerate(examples[dataset_attr.messages]):
if (
dataset_attr.system_tag
and messages[0][dataset_attr.role_tag] == dataset_attr.system_tag
):
system = messages[0][dataset_attr.content_tag]
messages = messages[1:]
else:
system = examples[dataset_attr.system][i] if dataset_attr.system else ""
messages = messages[: len(messages) // 2 * 2] # should be multiples of 2
if len(messages) == 0:
continue
aligned_messages = []
for turn_idx, message in enumerate(messages):
if message[dataset_attr.role_tag] not in accept_tags[turn_idx % 2]:
raise ValueError("Invalid role tag in {}.".format(messages))
aligned_messages.append(
{
"role": tag_mapping[message[dataset_attr.role_tag]],
"content": message[dataset_attr.content_tag],
}
)
outputs["prompt"].append(aligned_messages[:-1])
outputs["response"].append(aligned_messages[-1:])
outputs["system"].append(system)
outputs["tools"].append(
examples[dataset_attr.tools][i] if dataset_attr.tools else ""
)
print(examples[dataset_attr.images][i])
outputs["images"].append(
examples[dataset_attr.images][i] if dataset_attr.images else []
)
return outputs
def align_dataset(
dataset: Union["Dataset", "IterableDataset"], dataset_attr: "DatasetAttr", data_args: "DataArguments"
dataset: Union["Dataset", "IterableDataset"],
dataset_attr: "DatasetAttr",
data_args: "DataArguments",
) -> Union["Dataset", "IterableDataset"]:
r"""
Aligned dataset:
@@ -100,6 +181,8 @@ def align_dataset(
"""
if dataset_attr.formatting == "alpaca":
convert_func = partial(convert_alpaca, dataset_attr=dataset_attr)
elif dataset_attr.formatting == "llava":
convert_func = partial(convert_llava, dataset_attr=dataset_attr)
else:
convert_func = partial(convert_sharegpt, dataset_attr=dataset_attr)
@@ -107,13 +190,20 @@ def align_dataset(
features = Features.from_dict(
{
"prompt": [
{"role": {"dtype": "string", "_type": "Value"}, "content": {"dtype": "string", "_type": "Value"}}
{
"role": {"dtype": "string", "_type": "Value"},
"content": {"dtype": "string", "_type": "Value"},
}
],
"response": [
{"role": {"dtype": "string", "_type": "Value"}, "content": {"dtype": "string", "_type": "Value"}}
{
"role": {"dtype": "string", "_type": "Value"},
"content": {"dtype": "string", "_type": "Value"},
}
],
"system": {"dtype": "string", "_type": "Value"},
"tools": {"dtype": "string", "_type": "Value"},
"images": {"feature": {"_type": "Image"}, "_type": "Sequence"},
}
)
kwargs = {}

103
src/llmtuner/data/loader.py
View File

@@ -1,6 +1,6 @@
import inspect
import os
from typing import TYPE_CHECKING, Literal, Union
from typing import TYPE_CHECKING, Literal, Union, Optional
from datasets import load_dataset, load_from_disk
@@ -25,9 +25,9 @@ logger = get_logger(__name__)
def load_single_dataset(
dataset_attr: "DatasetAttr",
model_args: "ModelArguments",
data_args: "DataArguments",
dataset_attr: "DatasetAttr",
model_args: "ModelArguments",
data_args: "DataArguments",
) -> Union["Dataset", "IterableDataset"]:
logger.info("Loading dataset {}...".format(dataset_attr))
data_path, data_name, data_dir, data_files = None, None, None, None
@@ -78,14 +78,20 @@ def load_single_dataset(
split=data_args.split,
cache_dir=cache_dir,
token=model_args.ms_hub_token,
use_streaming=(data_args.streaming and (dataset_attr.load_from != "file")),
use_streaming=(
data_args.streaming and (dataset_attr.load_from != "file")
),
)
if isinstance(dataset, MsDataset):
dataset = dataset.to_hf_dataset()
except ImportError:
raise ImportError("Please install modelscope via `pip install modelscope -U`")
raise ImportError(
"Please install modelscope via `pip install modelscope -U`"
)
else:
if "trust_remote_code" in inspect.signature(load_dataset).parameters: # for datasets==2.16.0
if (
"trust_remote_code" in inspect.signature(load_dataset).parameters
): # for datasets==2.16.0
kwargs = {"trust_remote_code": True}
else:
kwargs = {}
@@ -102,7 +108,9 @@ def load_single_dataset(
**kwargs,
)
if data_args.streaming and (dataset_attr.load_from == "file"): # faster than specifying streaming=True
if data_args.streaming and (
dataset_attr.load_from == "file"
): # faster than specifying streaming=True
dataset = dataset.to_iterable_dataset() # TODO: add num shards parameter
if data_args.max_samples is not None: # truncate dataset
@@ -113,11 +121,12 @@ def load_single_dataset(
def get_dataset(
tokenizer: "PreTrainedTokenizer",
model_args: "ModelArguments",
data_args: "DataArguments",
training_args: "Seq2SeqTrainingArguments",
stage: Literal["pt", "sft", "rm", "ppo"],
tokenizer: "PreTrainedTokenizer",
model_args: "ModelArguments",
data_args: "DataArguments",
training_args: "Seq2SeqTrainingArguments",
stage: Literal["pt", "sft", "rm", "ppo"],
processor: Optional["AutoProcessor"] = None,
) -> Union["Dataset", "IterableDataset"]:
template = get_template_and_fix_tokenizer(tokenizer, data_args.template)
if data_args.train_on_prompt and template.efficient_eos:
@@ -126,9 +135,13 @@ def get_dataset(
# Load tokenized dataset
if data_args.tokenized_path is not None:
if has_tokenized_data(data_args.tokenized_path):
logger.warning("Loading dataset from disk will ignore other data arguments.")
logger.warning(
"Loading dataset from disk will ignore other data arguments."
)
dataset = load_from_disk(data_args.tokenized_path)
logger.info("Loaded tokenized dataset from {}.".format(data_args.tokenized_path))
logger.info(
"Loaded tokenized dataset from {}.".format(data_args.tokenized_path)
)
if data_args.streaming:
dataset = dataset.to_iterable_dataset()
return dataset
@@ -139,15 +152,21 @@ def get_dataset(
with training_args.main_process_first(desc="load dataset"):
all_datasets = []
for dataset_attr in get_dataset_list(data_args):
if (stage == "rm" and dataset_attr.ranking is False) or (stage != "rm" and dataset_attr.ranking is True):
raise ValueError("The dataset is not applicable in the current training stage.")
if (stage == "rm" and dataset_attr.ranking is False) or (
stage != "rm" and dataset_attr.ranking is True
):
raise ValueError(
"The dataset is not applicable in the current training stage."
)
all_datasets.append(load_single_dataset(dataset_attr, model_args, data_args))
all_datasets.append(
load_single_dataset(dataset_attr, model_args, data_args)
)
dataset = merge_dataset(all_datasets, data_args, training_args)
with training_args.main_process_first(desc="pre-process dataset"):
preprocess_func, print_function = get_preprocess_and_print_func(
tokenizer, template, data_args, training_args, stage
tokenizer, template, data_args, training_args, stage, processor
)
column_names = list(next(iter(dataset)).keys())
kwargs = {}
@@ -158,13 +177,21 @@ def get_dataset(
desc="Running tokenizer on dataset",
)
dataset = dataset.map(preprocess_func, batched=True, remove_columns=column_names, **kwargs)
dataset = dataset.map(
preprocess_func, batched=True, remove_columns=column_names, **kwargs
)
if data_args.tokenized_path is not None:
if training_args.should_save:
dataset.save_to_disk(data_args.tokenized_path)
logger.info("Tokenized dataset saved at {}.".format(data_args.tokenized_path))
logger.info("Please restart the training with `--tokenized_path {}`.".format(data_args.tokenized_path))
logger.info(
"Tokenized dataset saved at {}.".format(data_args.tokenized_path)
)
logger.info(
"Please restart the training with `--tokenized_path {}`.".format(
data_args.tokenized_path
)
)
exit(0)
@@ -172,34 +199,8 @@ def get_dataset(
try:
print_function(next(iter(dataset)))
except StopIteration:
raise RuntimeError("Cannot find valid samples, check `data/README.md` for the data format.")
raise RuntimeError(
"Cannot find valid samples, check `data/README.md` for the data format."
)
return dataset
def get_mm_dataset(
processor: "AutoProcessor",
model_args: "ModelArguments",
data_args: "DataArguments",
training_args: "Seq2SeqTrainingArguments",
stage: Literal["pt", "sft", "rm", "ppo"],
) -> Union["Dataset", "IterableDataset"]:
if data_args.tokenized_path is not None:
if has_tokenized_data(data_args.tokenized_path):
logger.warning("Loading dataset from disk will ignore other data arguments.")
dataset = load_from_disk(data_args.tokenized_path)
logger.info("Loaded tokenized dataset from {}.".format(data_args.tokenized_path))
if data_args.streaming:
dataset = dataset.to_iterable_dataset()
return dataset
if data_args.streaming:
raise ValueError("Turn off `streaming` when saving dataset to disk.")
with training_args.main_process_first(desc="load dataset"):
all_datasets = []
for dataset_attr in get_dataset_list(data_args):
all_datasets.append(load_dataset(dataset_attr.dataset_name)['train'])
dataset = merge_dataset(all_datasets, data_args, training_args)
return dataset

37
src/llmtuner/data/parser.py
View File

@@ -25,7 +25,7 @@ class DatasetAttr:
subset: Optional[str] = None
folder: Optional[str] = None
ranking: bool = False
formatting: Literal["alpaca", "sharegpt"] = "alpaca"
formatting: Literal["alpaca", "sharegpt", "llava"] = "alpaca"
""" columns """
system: Optional[str] = None
""" columns for the alpaca format """
@@ -44,11 +44,15 @@ class DatasetAttr:
observation_tag: Optional[str] = "observation"
function_tag: Optional[str] = "function_call"
system_tag: Optional[str] = "system"
""" columns for the mllm format """
images: Optional[str] = None
def __repr__(self) -> str:
return self.dataset_name
def set_attr(self, key: str, obj: Dict[str, Any], default: Optional[Any] = None) -> None:
def set_attr(
self, key: str, obj: Dict[str, Any], default: Optional[Any] = None
) -> None:
setattr(self, key, obj.get(key, default))
@@ -67,12 +71,16 @@ def get_dataset_list(data_args: "DataArguments") -> List["DatasetAttr"]:
except Exception as err:
if len(dataset_names) != 0:
raise ValueError(
"Cannot open {} due to {}.".format(os.path.join(data_args.dataset_dir, DATA_CONFIG), str(err))
"Cannot open {} due to {}.".format(
os.path.join(data_args.dataset_dir, DATA_CONFIG), str(err)
)
)
dataset_info = None
if data_args.interleave_probs is not None:
data_args.interleave_probs = [float(prob.strip()) for prob in data_args.interleave_probs.split(",")]
data_args.interleave_probs = [
float(prob.strip()) for prob in data_args.interleave_probs.split(",")
]
dataset_list: List[DatasetAttr] = []
for name in dataset_names:
@@ -90,31 +98,42 @@ def get_dataset_list(data_args: "DataArguments") -> List["DatasetAttr"]:
if has_hf_url or has_ms_url:
if (use_modelscope() and has_ms_url) or (not has_hf_url):
dataset_attr = DatasetAttr("ms_hub", dataset_name=dataset_info[name]["ms_hub_url"])
dataset_attr = DatasetAttr(
"ms_hub", dataset_name=dataset_info[name]["ms_hub_url"]
)
else:
dataset_attr = DatasetAttr("hf_hub", dataset_name=dataset_info[name]["hf_hub_url"])
dataset_attr = DatasetAttr(
"hf_hub", dataset_name=dataset_info[name]["hf_hub_url"]
)
elif "script_url" in dataset_info[name]:
dataset_attr = DatasetAttr("script", dataset_name=dataset_info[name]["script_url"])
dataset_attr = DatasetAttr(
"script", dataset_name=dataset_info[name]["script_url"]
)
else:
dataset_attr = DatasetAttr("file", dataset_name=dataset_info[name]["file_name"])
dataset_attr = DatasetAttr(
"file", dataset_name=dataset_info[name]["file_name"]
)
dataset_attr.set_attr("file_sha1", dataset_info[name])
dataset_attr.set_attr("subset", dataset_info[name])
dataset_attr.set_attr("folder", dataset_info[name])
dataset_attr.set_attr("ranking", dataset_info[name], default=False)
dataset_attr.set_attr("formatting", dataset_info[name], default="alpaca")
dataset_attr.set_attr("images", dataset_info[name], default="")
if "columns" in dataset_info[name]:
column_names = ["system"]
if dataset_attr.formatting == "alpaca":
column_names.extend(["prompt", "query", "response", "history"])
elif dataset_attr.formatting == "llava":
column_names.extend(["messages", "images"])
else:
column_names.extend(["messages", "tools"])
for column_name in column_names:
dataset_attr.set_attr(column_name, dataset_info[name]["columns"])
if dataset_attr.formatting == "sharegpt" and "tags" in dataset_info[name]:
if dataset_attr.formatting != "alpaca" and "tags" in dataset_info[name]:
tag_names = (
"role_tag",
"content_tag",

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

248
src/llmtuner/data/template.py
View File

@@ -42,7 +42,9 @@ class Template:
r"""
Returns a single pair of token ids representing prompt and response respectively.
"""
encoded_pairs = self._encode(tokenizer, messages, system, tools, cutoff_len, reserved_label_len)
encoded_pairs = self._encode(
tokenizer, messages, system, tools, cutoff_len, reserved_label_len
)
prompt_ids = []
for query_ids, resp_ids in encoded_pairs[:-1]:
prompt_ids += query_ids + resp_ids
@@ -62,7 +64,9 @@ class Template:
r"""
Returns multiple pairs of token ids representing prompts and responses respectively.
"""
return self._encode(tokenizer, messages, system, tools, cutoff_len, reserved_label_len)
return self._encode(
tokenizer, messages, system, tools, cutoff_len, reserved_label_len
)
def _encode(
self,
@@ -89,7 +93,9 @@ class Template:
elements += self.format_separator.apply()
if message["role"] == Role.USER.value:
elements += self.format_user.apply(content=message["content"], idx=str(i // 2))
elements += self.format_user.apply(
content=message["content"], idx=str(i // 2)
)
elif message["role"] == Role.ASSISTANT.value:
elements += self.format_assistant.apply(content=message["content"])
elif message["role"] == Role.OBSERVATION.value:
@@ -104,7 +110,9 @@ class Template:
return self._make_pairs(encoded_messages, cutoff_len, reserved_label_len)
def _convert_elements_to_ids(
self, tokenizer: "PreTrainedTokenizer", elements: List[Union[str, Dict[str, str]]]
self,
tokenizer: "PreTrainedTokenizer",
elements: List[Union[str, Dict[str, str]]],
) -> List[int]:
r"""
Converts elements to token ids.
@@ -122,7 +130,11 @@ class Template:
elif "eos_token" in elem and tokenizer.eos_token_id is not None:
token_ids += [tokenizer.eos_token_id]
else:
raise ValueError("Input must be string, set[str] or dict[str, str], got {}".format(type(elem)))
raise ValueError(
"Input must be string, set[str] or dict[str, str], got {}".format(
type(elem)
)
)
return token_ids
@@ -180,7 +192,9 @@ class Llama2Template(Template):
elements += self.format_separator.apply()
if message["role"] == Role.USER.value:
elements += self.format_user.apply(content=system_text + message["content"])
elements += self.format_user.apply(
content=system_text + message["content"]
)
elif message["role"] == Role.ASSISTANT.value:
elements += self.format_assistant.apply(content=message["content"])
elif message["role"] == Role.OBSERVATION.value:
@@ -243,7 +257,9 @@ def _register_template(
template_class = Llama2Template if name.startswith("llama2") else Template
default_user_formatter = StringFormatter(slots=["{{content}}"])
default_assistant_formatter = StringFormatter(slots=["{{content}}"] + eos_slots)
default_function_formatter = FunctionFormatter(slots=["Action: {{name}}\nAction Input: {{arguments}}"] + eos_slots)
default_function_formatter = FunctionFormatter(
slots=["Action: {{name}}\nAction Input: {{arguments}}"] + eos_slots
)
default_tool_formatter = ToolFormatter(tool_format="default")
default_separator_formatter = EmptyFormatter()
templates[name] = template_class(
@@ -279,7 +295,9 @@ def _jinja_escape(content: str) -> str:
return content.replace("\n", r"\n").replace("'", r"\'")
def _convert_slots_to_jinja(slots: "SLOTS", tokenizer: "PreTrainedTokenizer", placeholder: str = "content") -> str:
def _convert_slots_to_jinja(
slots: "SLOTS", tokenizer: "PreTrainedTokenizer", placeholder: str = "content"
) -> str:
slot_items = []
for slot in slots:
if isinstance(slot, str):
@@ -293,7 +311,9 @@ def _convert_slots_to_jinja(slots: "SLOTS", tokenizer: "PreTrainedTokenizer", pl
elif isinstance(slot, set):
if "bos_token" in slot:
slot_items.append("'" + tokenizer.bos_token + "'")
elif "eos_token" in slot: # do not use {{ eos_token }} since it may be replaced
elif (
"eos_token" in slot
): # do not use {{ eos_token }} since it may be replaced
slot_items.append("'" + tokenizer.eos_token + "'")
elif isinstance(slot, dict):
raise ValueError("Dict is not supported.")
@@ -305,25 +325,37 @@ def _get_jinja_template(template: "Template", tokenizer: "PreTrainedTokenizer")
jinja_template = ""
if template.default_system:
jinja_template += "{% set system_message = '" + _jinja_escape(template.default_system) + "' %}"
jinja_template += (
"{% set system_message = '"
+ _jinja_escape(template.default_system)
+ "' %}"
)
jinja_template += (
"{% if messages[0]['role'] == 'system' %}" "{% set system_message = messages[0]['content'] %}" "{% endif %}"
"{% if messages[0]['role'] == 'system' %}"
"{% set system_message = messages[0]['content'] %}"
"{% endif %}"
)
system_message = _convert_slots_to_jinja(template.format_system.apply(), tokenizer, placeholder="system_message")
system_message = _convert_slots_to_jinja(
template.format_system.apply(), tokenizer, placeholder="system_message"
)
if isinstance(template, Llama2Template):
pass
elif template.force_system:
jinja_template += "{{ " + system_message + " }}"
else:
jinja_template += "{% if system_message is defined %}{{ " + system_message + " }}{% endif %}"
jinja_template += (
"{% if system_message is defined %}{{ " + system_message + " }}{% endif %}"
)
jinja_template += "{% for message in messages %}"
jinja_template += "{% set content = message['content'] %}"
if isinstance(template, Llama2Template):
jinja_template += "{% if loop.index0 == 0 and system_message is defined %}"
jinja_template += "{% set content = " + system_message + " + message['content'] %}"
jinja_template += (
"{% set content = " + system_message + " + message['content'] %}"
)
jinja_template += "{% endif %}"
jinja_template += "{% if message['role'] == 'user' %}"
user_message = _convert_slots_to_jinja(template.format_user.apply(), tokenizer)
@@ -366,11 +398,14 @@ def get_template_and_fix_tokenizer(
if stop_words:
num_added_tokens = tokenizer.add_special_tokens(
dict(additional_special_tokens=stop_words), replace_additional_special_tokens=False
dict(additional_special_tokens=stop_words),
replace_additional_special_tokens=False,
)
logger.info("Add {} to stop words.".format(",".join(stop_words)))
if num_added_tokens > 0:
logger.warning("New tokens have been added, make sure `resize_vocab` is True.")
logger.warning(
"New tokens have been added, make sure `resize_vocab` is True."
)
try:
tokenizer.chat_template = _get_jinja_template(template, tokenizer)
@@ -382,7 +417,9 @@ def get_template_and_fix_tokenizer(
_register_template(
name="alpaca",
format_user=StringFormatter(slots=["### Instruction:\n{{content}}\n\n### Response:\n"]),
format_user=StringFormatter(
slots=["### Instruction:\n{{content}}\n\n### Response:\n"]
),
format_separator=EmptyFormatter(slots=["\n\n"]),
default_system=(
"Below is an instruction that describes a task. "
@@ -407,7 +444,13 @@ _register_template(
_register_template(
name="atom",
format_user=StringFormatter(
slots=[{"bos_token"}, "Human: {{content}}\n", {"eos_token"}, {"bos_token"}, "Assistant:"]
slots=[
{"bos_token"},
"Human: {{content}}\n",
{"eos_token"},
{"bos_token"},
"Assistant:",
]
),
format_assistant=StringFormatter(slots=["{{content}}\n", {"eos_token"}]),
)
@@ -415,7 +458,9 @@ _register_template(
_register_template(
name="baichuan",
format_user=StringFormatter(slots=[{"token": "<reserved_102>"}, "{{content}}", {"token": "<reserved_103>"}]),
format_user=StringFormatter(
slots=[{"token": "<reserved_102>"}, "{{content}}", {"token": "<reserved_103>"}]
),
efficient_eos=True,
)
@@ -438,7 +483,9 @@ _register_template(
_register_template(
name="bluelm",
format_user=StringFormatter(slots=[{"token": "[|Human|]:"}, "{{content}}", {"token": "[|AI|]:"}]),
format_user=StringFormatter(
slots=[{"token": "[|Human|]:"}, "{{content}}", {"token": "[|AI|]:"}]
),
)
@@ -457,7 +504,9 @@ _register_template(
_register_template(
name="chatglm2",
format_user=StringFormatter(slots=["[Round {{idx}}]\n\n问：{{content}}\n\n答："]),
format_system=StringFormatter(slots=[{"token": "[gMASK]"}, {"token": "sop"}, "{{content}}"]),
format_system=StringFormatter(
slots=[{"token": "[gMASK]"}, {"token": "sop"}, "{{content}}"]
),
format_separator=EmptyFormatter(slots=["\n\n"]),
efficient_eos=True,
force_system=True,
@@ -466,12 +515,21 @@ _register_template(
_register_template(
name="chatglm3",
format_user=StringFormatter(slots=[{"token": "<|user|>"}, "\n", "{{content}}", {"token": "<|assistant|>"}]),
format_user=StringFormatter(
slots=[{"token": "<|user|>"}, "\n", "{{content}}", {"token": "<|assistant|>"}]
),
format_assistant=StringFormatter(slots=["\n", "{{content}}"]),
format_system=StringFormatter(slots=[{"token": "[gMASK]"}, {"token": "sop"}, "{{content}}"]),
format_system=StringFormatter(
slots=[{"token": "[gMASK]"}, {"token": "sop"}, "{{content}}"]
),
format_function=FunctionFormatter(slots=["{{name}}\n{{arguments}}"]),
format_observation=StringFormatter(
slots=[{"token": "<|observation|>"}, "\n", "{{content}}", {"token": "<|assistant|>"}]
slots=[
{"token": "<|observation|>"},
"\n",
"{{content}}",
{"token": "<|assistant|>"},
]
),
stop_words=["<|user|>", "<|observation|>"],
efficient_eos=True,
@@ -481,14 +539,27 @@ _register_template(
_register_template(
name="chatglm3_system",
format_user=StringFormatter(slots=[{"token": "<|user|>"}, "\n", "{{content}}", {"token": "<|assistant|>"}]),
format_user=StringFormatter(
slots=[{"token": "<|user|>"}, "\n", "{{content}}", {"token": "<|assistant|>"}]
),
format_assistant=StringFormatter(slots=["\n", "{{content}}"]),
format_system=StringFormatter(
slots=[{"token": "[gMASK]"}, {"token": "sop"}, {"token": "<|system|>"}, "\n", "{{content}}"]
slots=[
{"token": "[gMASK]"},
{"token": "sop"},
{"token": "<|system|>"},
"\n",
"{{content}}",
]
),
format_function=FunctionFormatter(slots=["{{name}}\n{{arguments}}"]),
format_observation=StringFormatter(
slots=[{"token": "<|observation|>"}, "\n", "{{content}}", {"token": "<|assistant|>"}]
slots=[
{"token": "<|observation|>"},
"\n",
"{{content}}",
{"token": "<|assistant|>"},
]
),
default_system=(
"You are ChatGLM3, a large language model trained by Zhipu.AI. "
@@ -501,9 +572,15 @@ _register_template(
_register_template(
name="chatml",
format_user=StringFormatter(slots=["<|im_start|>user\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]),
format_system=StringFormatter(slots=["<|im_start|>system\n{{content}}<|im_end|>\n"]),
format_observation=StringFormatter(slots=["<|im_start|>tool\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]),
format_user=StringFormatter(
slots=["<|im_start|>user\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]
),
format_system=StringFormatter(
slots=["<|im_start|>system\n{{content}}<|im_end|>\n"]
),
format_observation=StringFormatter(
slots=["<|im_start|>tool\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]
),
format_separator=EmptyFormatter(slots=["\n"]),
stop_words=["<|im_end|>", "<|im_start|>"],
replace_eos=True,
@@ -512,9 +589,15 @@ _register_template(
_register_template(
name="chatml_de",
format_user=StringFormatter(slots=["<|im_start|>user\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]),
format_system=StringFormatter(slots=["<|im_start|>system\n{{content}}<|im_end|>\n"]),
format_observation=StringFormatter(slots=["<|im_start|>tool\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]),
format_user=StringFormatter(
slots=["<|im_start|>user\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]
),
format_system=StringFormatter(
slots=["<|im_start|>system\n{{content}}<|im_end|>\n"]
),
format_observation=StringFormatter(
slots=["<|im_start|>tool\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]
),
format_separator=EmptyFormatter(slots=["\n"]),
default_system="Du bist ein freundlicher und hilfsbereiter KI-Assistent.",
stop_words=["<|im_end|>", "<|im_start|>"],
@@ -524,7 +607,9 @@ _register_template(
_register_template(
name="codegeex2",
format_system=StringFormatter(slots=[{"token": "[gMASK]"}, {"token": "sop"}, "{{content}}"]),
format_system=StringFormatter(
slots=[{"token": "[gMASK]"}, {"token": "sop"}, "{{content}}"]
),
force_system=True,
)
@@ -554,9 +639,15 @@ _register_template(
_register_template(
name="dbrx",
format_user=StringFormatter(slots=["<|im_start|>user\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]),
format_system=StringFormatter(slots=["<|im_start|>system\n{{content}}<|im_end|>\n"]),
format_observation=StringFormatter(slots=["<|im_start|>tool\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]),
format_user=StringFormatter(
slots=["<|im_start|>user\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]
),
format_system=StringFormatter(
slots=["<|im_start|>system\n{{content}}<|im_end|>\n"]
),
format_observation=StringFormatter(
slots=["<|im_start|>tool\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]
),
format_separator=EmptyFormatter(slots=["\n"]),
default_system=(
"You are DBRX, created by Databricks. You were last updated in December 2023. "
@@ -634,7 +725,9 @@ _register_template(
_register_template(
name="gemma",
format_user=StringFormatter(slots=["<start_of_turn>user\n{{content}}<end_of_turn>\n<start_of_turn>model\n"]),
format_user=StringFormatter(
slots=["<start_of_turn>user\n{{content}}<end_of_turn>\n<start_of_turn>model\n"]
),
format_system=StringFormatter(slots=[{"bos_token"}, "{{content}}"]),
format_observation=StringFormatter(
slots=["<start_of_turn>tool\n{{content}}<end_of_turn>\n<start_of_turn>model\n"]
@@ -647,7 +740,9 @@ _register_template(
_register_template(
name="intern",
format_user=StringFormatter(slots=["<|User|>:{{content}}", {"token": "<eoh>"}, "\n<|Bot|>:"]),
format_user=StringFormatter(
slots=["<|User|>:{{content}}", {"token": "<eoh>"}, "\n<|Bot|>:"]
),
format_separator=EmptyFormatter(slots=[{"token": "<eoa>"}, "\n"]),
stop_words=["<eoa>"],
efficient_eos=True,
@@ -656,8 +751,12 @@ _register_template(
_register_template(
name="intern2",
format_user=StringFormatter(slots=["<|im_start|>user\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]),
format_system=StringFormatter(slots=[{"bos_token"}, "<|im_start|>system\n{{content}}<|im_end|>\n"]),
format_user=StringFormatter(
slots=["<|im_start|>user\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]
),
format_system=StringFormatter(
slots=[{"bos_token"}, "<|im_start|>system\n{{content}}<|im_end|>\n"]
),
format_separator=EmptyFormatter(slots=["\n"]),
default_system=(
"You are an AI assistant whose name is InternLM (书生·浦语).\n"
@@ -707,7 +806,10 @@ _register_template(
]
),
format_system=StringFormatter(
slots=[{"bos_token"}, "<|start_header_id|>system<|end_header_id|>\n\n{{content}}<|eot_id|>"]
slots=[
{"bos_token"},
"<|start_header_id|>system<|end_header_id|>\n\n{{content}}<|eot_id|>",
]
),
format_observation=StringFormatter(
slots=[
@@ -742,7 +844,13 @@ _register_template(
_register_template(
name="openchat",
format_user=StringFormatter(slots=["GPT4 Correct User: {{content}}", {"eos_token"}, "GPT4 Correct Assistant:"]),
format_user=StringFormatter(
slots=[
"GPT4 Correct User: {{content}}",
{"eos_token"},
"GPT4 Correct Assistant:",
]
),
format_assistant=StringFormatter(slots=["{{content}}", {"eos_token"}]),
format_system=StringFormatter(slots=[{"bos_token"}, "{{content}}"]),
force_system=True,
@@ -751,7 +859,9 @@ _register_template(
_register_template(
name="orion",
format_user=StringFormatter(slots=["Human: {{content}}\n\nAssistant: ", {"eos_token"}]),
format_user=StringFormatter(
slots=["Human: {{content}}\n\nAssistant: ", {"eos_token"}]
),
format_system=StringFormatter(slots=[{"bos_token"}, "{{content}}"]),
force_system=True,
)
@@ -759,9 +869,15 @@ _register_template(
_register_template(
name="phi",
format_user=StringFormatter(slots=["<|user|>\n{{content}}<|end|>\n<|assistant|>\n"]),
format_system=StringFormatter(slots=[{"bos_token"}, "<|system|>\n{{content}}<|end|>\n"]),
format_observation=StringFormatter(slots=["<|function_output|>\n{{content}}<|end|>\n<|assistant|>\n"]),
format_user=StringFormatter(
slots=["<|user|>\n{{content}}<|end|>\n<|assistant|>\n"]
),
format_system=StringFormatter(
slots=[{"bos_token"}, "<|system|>\n{{content}}<|end|>\n"]
),
format_observation=StringFormatter(
slots=["<|function_output|>\n{{content}}<|end|>\n<|assistant|>\n"]
),
format_separator=EmptyFormatter(slots=["\n"]),
default_system="You are a helpful AI assistant.",
stop_words=["<|end|>"],
@@ -771,9 +887,15 @@ _register_template(
_register_template(
name="qwen",
format_user=StringFormatter(slots=["<|im_start|>user\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]),
format_system=StringFormatter(slots=["<|im_start|>system\n{{content}}<|im_end|>\n"]),
format_observation=StringFormatter(slots=["<|im_start|>tool\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]),
format_user=StringFormatter(
slots=["<|im_start|>user\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]
),
format_system=StringFormatter(
slots=["<|im_start|>system\n{{content}}<|im_end|>\n"]
),
format_observation=StringFormatter(
slots=["<|im_start|>tool\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]
),
format_separator=EmptyFormatter(slots=["\n"]),
default_system="You are a helpful assistant.",
stop_words=["<|im_end|>"],
@@ -829,8 +951,12 @@ _register_template(
_register_template(
name="yayi",
format_user=StringFormatter(slots=[{"token": "<|Human|>"}, ":\n{{content}}\n\n", {"token": "<|YaYi|>"}, ":"]),
format_system=StringFormatter(slots=[{"token": "<|System|>"}, ":\n{{content}}\n\n"]),
format_user=StringFormatter(
slots=[{"token": "<|Human|>"}, ":\n{{content}}\n\n", {"token": "<|YaYi|>"}, ":"]
),
format_system=StringFormatter(
slots=[{"token": "<|System|>"}, ":\n{{content}}\n\n"]
),
format_separator=EmptyFormatter(slots=["\n\n"]),
default_system=(
"You are a helpful, respectful and honest assistant named YaYi "
@@ -849,7 +975,9 @@ _register_template(
_register_template(
name="yi",
format_user=StringFormatter(slots=["<|im_start|>user\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]),
format_user=StringFormatter(
slots=["<|im_start|>user\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]
),
format_separator=EmptyFormatter(slots=["\n"]),
stop_words=["<|im_end|>"],
replace_eos=True,
@@ -867,7 +995,9 @@ _register_template(
_register_template(
name="zephyr",
format_user=StringFormatter(slots=["<|user|>\n{{content}}", {"eos_token"}, "<|assistant|>"]),
format_user=StringFormatter(
slots=["<|user|>\n{{content}}", {"eos_token"}, "<|assistant|>"]
),
format_assistant=StringFormatter(slots=["\n{{content}}", {"eos_token"}]),
format_system=StringFormatter(slots=["<|system|>\n{{content}}", {"eos_token"}]),
default_system="You are a friendly chatbot who always responds in the style of a pirate",
@@ -879,3 +1009,13 @@ _register_template(
format_user=StringFormatter(slots=["<human>:{{content}}\n<bot>:"]),
format_separator=EmptyFormatter(slots=["\n"]),
)
_register_template(
name="llava",
format_user=StringFormatter(slots=["USER: {{content}} "]),
format_assistant=StringFormatter(slots=["ASSISTANT: {{content}}"]),
default_system=(
"A chat between a curious user and an artificial intelligence assistant. "
"The assistant gives helpful, detailed, and polite answers to the user's questions."
),
)