423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-12-16 03:40:34 +08:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'main' into main

Browse Source
This commit is contained in:
hoshi-hiyouga
2024-07-01 21:01:09 +08:00
committed by GitHub
parent 770f75dc83 33f2ddb8b6
commit e8e6af2651
181 changed files with 4835 additions and 1382 deletions
Download Patch File Download Diff File Expand all files Collapse all files

41
src/llamafactory/webui/runner.py
View File

@@ -1,3 +1,17 @@
# Copyright 2024 the LlamaFactory team.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import os
from copy import deepcopy
from subprocess import Popen, TimeoutExpired
@@ -8,9 +22,9 @@ 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 .common import DEFAULT_CACHE_DIR, DEFAULT_CONFIG_DIR, get_save_dir, load_config
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_leaf_process, gen_cmd, get_eval_results, get_trainer_info, load_args, save_args, save_cmd
from .utils import abort_process, gen_cmd, get_eval_results, get_trainer_info, load_args, save_args, save_cmd
if is_gradio_available():
@@ -38,7 +52,7 @@ class Runner:
def set_abort(self) -> None:
self.aborted = True
if self.trainer is not None:
abort_leaf_process(self.trainer.pid)
abort_process(self.trainer.pid)
def _initialize(self, data: Dict["Component", Any], do_train: bool, from_preview: bool) -> str:
get = lambda elem_id: data[self.manager.get_elem_by_id(elem_id)]
@@ -90,6 +104,11 @@ class Runner:
model_name, finetuning_type = get("top.model_name"), get("top.finetuning_type")
user_config = load_config()
if get("top.quantization_bit") in QUANTIZATION_BITS:
quantization_bit = int(get("top.quantization_bit"))
else:
quantization_bit = None
args = dict(
stage=TRAINING_STAGES[get("train.training_stage")],
do_train=True,
@@ -97,7 +116,8 @@ class Runner:
cache_dir=user_config.get("cache_dir", None),
preprocessing_num_workers=16,
finetuning_type=finetuning_type,
quantization_bit=int(get("top.quantization_bit")) if get("top.quantization_bit") in ["8", "4"] else None,
quantization_bit=quantization_bit,
quantization_method=get("top.quantization_method"),
template=get("top.template"),
rope_scaling=get("top.rope_scaling") if get("top.rope_scaling") in ["linear", "dynamic"] else None,
flash_attn="fa2" if get("top.booster") == "flashattn2" else "auto",
@@ -160,6 +180,8 @@ class Runner:
args["create_new_adapter"] = get("train.create_new_adapter")
args["use_rslora"] = get("train.use_rslora")
args["use_dora"] = get("train.use_dora")
args["pissa_init"] = get("train.use_pissa")
args["pissa_convert"] = get("train.use_pissa")
args["lora_target"] = get("train.lora_target") or "all"
args["additional_target"] = get("train.additional_target") or None
@@ -219,13 +241,19 @@ class Runner:
model_name, finetuning_type = get("top.model_name"), get("top.finetuning_type")
user_config = load_config()
if get("top.quantization_bit") in QUANTIZATION_BITS:
quantization_bit = int(get("top.quantization_bit"))
else:
quantization_bit = None
args = dict(
stage="sft",
model_name_or_path=get("top.model_path"),
cache_dir=user_config.get("cache_dir", None),
preprocessing_num_workers=16,
finetuning_type=finetuning_type,
quantization_bit=int(get("top.quantization_bit")) if get("top.quantization_bit") in ["8", "4"] else None,
quantization_bit=quantization_bit,
quantization_method=get("top.quantization_method"),
template=get("top.template"),
rope_scaling=get("top.rope_scaling") if get("top.rope_scaling") in ["linear", "dynamic"] else None,
flash_attn="fa2" if get("top.booster") == "flashattn2" else "auto",
@@ -283,6 +311,7 @@ class Runner:
env = deepcopy(os.environ)
env["LLAMABOARD_ENABLED"] = "1"
env["LLAMABOARD_WORKDIR"] = args["output_dir"]
if args.get("deepspeed", None) is not None:
env["FORCE_TORCHRUN"] = "1"
@@ -291,7 +320,7 @@ class Runner:
def _form_config_dict(self, data: Dict["Component", Any]) -> Dict[str, Any]:
config_dict = {}
skip_ids = ["top.lang", "top.model_path", "train.output_dir", "train.config_path", "train.device_count"]
skip_ids = ["top.lang", "top.model_path", "train.output_dir", "train.config_path"]
for elem, value in data.items():
elem_id = self.manager.get_id_by_elem(elem)
if elem_id not in skip_ids: