support rank0 logger

Former-commit-id: c38aa29336
This commit is contained in:
hiyouga
2024-11-02 18:31:04 +08:00
parent 4b2c47fcae
commit e83cb17f97
42 changed files with 316 additions and 252 deletions

View File

@@ -19,6 +19,7 @@ from typing import Any, Dict, Optional, Tuple
from yaml import safe_dump, safe_load
from ..extras import logging
from ..extras.constants import (
CHECKPOINT_NAMES,
DATA_CONFIG,
@@ -30,7 +31,6 @@ from ..extras.constants import (
VISION_MODELS,
DownloadSource,
)
from ..extras.logging import get_logger
from ..extras.misc import use_modelscope, use_openmind
from ..extras.packages import is_gradio_available
@@ -39,7 +39,7 @@ if is_gradio_available():
import gradio as gr
logger = get_logger(__name__)
logger = logging.get_logger(__name__)
DEFAULT_CACHE_DIR = "cache"
@@ -56,7 +56,7 @@ def get_save_dir(*paths: str) -> os.PathLike:
Gets the path to saved model checkpoints.
"""
if os.path.sep in paths[-1]:
logger.warning("Found complex path, some features may be not available.")
logger.warning_rank0("Found complex path, some features may be not available.")
return paths[-1]
paths = (path.replace(" ", "").strip() for path in paths)
@@ -172,14 +172,14 @@ def load_dataset_info(dataset_dir: str) -> Dict[str, Dict[str, Any]]:
Loads dataset_info.json.
"""
if dataset_dir == "ONLINE" or dataset_dir.startswith("REMOTE:"):
logger.info(f"dataset_dir is {dataset_dir}, using online dataset.")
logger.info_rank0(f"dataset_dir is {dataset_dir}, using online dataset.")
return {}
try:
with open(os.path.join(dataset_dir, DATA_CONFIG), encoding="utf-8") as f:
return json.load(f)
except Exception as err:
logger.warning(f"Cannot open {os.path.join(dataset_dir, DATA_CONFIG)} due to {str(err)}.")
logger.warning_rank0(f"Cannot open {os.path.join(dataset_dir, DATA_CONFIG)} due to {str(err)}.")
return {}

View File

@@ -22,7 +22,7 @@ from transformers.trainer import TRAINING_ARGS_NAME
from ..extras.constants import LLAMABOARD_CONFIG, PEFT_METHODS, TRAINING_STAGES
from ..extras.misc import is_gpu_or_npu_available, torch_gc
from ..extras.packages import is_gradio_available
from ..extras.packages import is_gradio_available, is_transformers_version_equal_to_4_46
from .common import DEFAULT_CACHE_DIR, DEFAULT_CONFIG_DIR, QUANTIZATION_BITS, get_save_dir, load_config
from .locales import ALERTS, LOCALES
from .utils import abort_process, gen_cmd, get_eval_results, get_trainer_info, load_args, save_args, save_cmd
@@ -152,7 +152,7 @@ class Runner:
pure_bf16=(get("train.compute_type") == "pure_bf16"),
plot_loss=True,
ddp_timeout=180000000,
include_num_input_tokens_seen=True,
include_num_input_tokens_seen=False if is_transformers_version_equal_to_4_46() else True, # FIXME
**json.loads(get("train.extra_args")),
)