Mirrors/modelscope
1
0
Fork 0
mirror of https://github.com/modelscope/modelscope.git synced 2025-12-21 10:39:24 +01:00
Code Issues Packages Projects Releases Wiki Activity
1,781 Commits 263 Branches 106 Tags
merge_from_internal_0701
Commit Graph

34 Commits

Author SHA1 Message Date
hemu.zp
bb194f1d96 fix llama bug 
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/13133641
* fix bug
 2023-06-30 20:49:30 +08:00
zsl01670416
2a2cc594eb Merge branch 'maas_flex_train_dev' modify finetune_text_classification for arguments collision between eval_strategy and by_e... 2023-06-30 19:04:14 +08:00
hemu.zp
30599dc34f fix llama lora bug 
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/13129620
 2023-06-30 19:03:56 +08:00
hemu.zp
f4c90f2adf Support lora for llama 
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/13080086

* support lora for llama

* update baichuan

* remove work_dir

* fixbug: 1. change ConfigDict to list when hooks key not in config 2. ignore all bin files when preparing output folder

* 1. support device_map 2. remove the operation of to float when using lora

* add inference file

* add comment

* support device_map
 2023-06-29 22:05:34 +08:00
yuze.zyz
d4b9396084 Revert "support device_map" 
This reverts commit 4d0367718d.
 2023-06-29 16:08:42 +08:00
yuze.zyz
4d0367718d support device_map 2023-06-29 08:07:43 +00:00
wenmeng.zwm
0271b9c256 Merge branch 'master-github' into merge_master_github_0628 2023-06-28 20:27:34 +08:00
Wang Qiang
d8381bf9fd Stable diffusion examples of lora and dreambooth. (#341) 
* stable diffusion examples of lora and dreambooth

* pre-commit
 2023-06-28 20:13:28 +08:00
yuze.zyz
a58be34384 Add Lora/Adapter/Prompt and support for chatglm6B and chatglm2-6B 
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/12770413

* add prompt and lora

* add adapter

* add prefix

* add tests

* adapter smoke test passed

* prompt test passed

* support model id in petl

* migrate chatglm6b

* add train script for chatglm6b

* move gen_kwargs to finetune.py

* add chatglm2

* add model definination
 2023-06-27 14:38:18 +08:00
suluyan.sly
7e28ca226b [to #50334474] llama tuned model -> pipeline 
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/13071602
* prepare for inference
 2023-06-26 22:49:44 +08:00
xingjun.wxj
0db0ec5586 Merge code from github 
1. Merge(add) daily regression from github PR (daily_regression.yaml)
2. Add lora stable diffusion from github PR
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/13010802
* fix: device arg not work, rename device to ngpu (#272)

* Correcting the lora stable diffusion example script (#300)

* add vad model and punc model in README.md 

add vad model and punc model

* Merge pull request #302 from modelscope/langgz-patch-1

add vad model and punc model in README.md

* add 1.6

* modify ignore

* Merge pull request #307 from modelscope/dev_rs_16

Merge release 1.6

* undo datetime to 2099

* Merge pull request #311 from modelscope/fix_master_version

undo datetime to 2099

* add daily regression workflow

* modify workflow name

* fix cron format issue

* lora trainer

* Merge pull request #315 from liuyhwangyh/add_regression_workflow

add daily regression workflow
 2023-06-21 10:22:06 +08:00
suluyan.sly
de491cfe6b [to #48855014] llama finetune + deepspeed 
1. llama base finetune：使用trainer从llama finetune至alpaca，效果确认并提供example
2. deepspeed通用性完善：mpu解耦；deepspeed的训练信息支持从ms log透出（目前ms打印的log有误）；支持从modelscope configuration.json 进行 deepspeed config 配置；deepspeed optimizer和lr_scheduler 初始化支持；解决deepspeed和ddp同时使用报错；解决保存ckpt时报错
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/12651323
 2023-06-14 10:00:23 +08:00
zsl01670416
75c90d8a98 add model revision in training_args and modify dataset loading in finetune text classification 
1.add parameter model_revision in training_args.py.
2.add parameter model_revision in kwargs for finetune_text_classification.py and finetune_text_generation.py.
3.modify dataset loading in finetune_text_classification.py for flex training.
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/12869552
* add model revision in training_args and modify dataset loading in finetune text classification
 2023-06-08 19:36:05 +08:00
pangda
73f8d39897 Add ner example for flextrain 
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/12835095
* update named_entity_recognition example
 2023-06-05 15:17:45 +08:00
xingjun.wxj
5b99bb24fe update version: 1.6.1rc0 
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/12754942
 2023-05-25 20:28:36 +08:00
yuze.zyz
d38fa563a8 Re commit lossing files for flextrain 
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/12674746
 2023-05-19 10:48:33 +08:00
yuze.zyz
38fa4cceed fix transformer example and fix some bugs 
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/12626375
 2023-05-16 15:27:17 +08:00
zsl01670416
71a80173e5 modify examples text_classification, text_generation, token_classification for improved trainer 
1.text_classification/run_train.sh
2.text_generation/run_train_mt5.sh, run_train_palm.sh
3.token_classification/finetune_token_classification.py, run_train_mgeo.sh, run_train_structbert.sh
above files were modified to adapt to improved trainer
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/12629010
 2023-05-16 14:32:43 +08:00
hemu.zp
5804ad2dc1 update multi_modal_embedding example 
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/12626062
 2023-05-16 14:31:26 +08:00
xingjun.wang
98e899b8c2 merge master 2023-05-14 02:09:02 +08:00
yuze.zyz
febc0365de Support FlexTrain and update the structure of trainer 
1. Refactor training_args
2. Refactor hooks
3. Add train_id for push_to_hub
4. Support both output_dir/output_sub_dir for checkpoint_hooks
5. Support copy when hardlink fails when checkpointing
6. Support mixed dataset config file as a CLI argument
7. Add eval txt in output folder
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/12384253
* support the ignorance of file pattern
 2023-05-13 12:12:04 +08:00
Wang Qiang
9bfc4a9d83 Add finetuning stable diffusion example (#285) 2023-04-28 10:18:42 +08:00
hemu
9f2f1c066a change file names 2023-04-13 10:41:18 +08:00
hemu
672f25266f Merge branch 'master-github' into master-merge-github-0413 2023-04-13 10:32:01 +08:00
slin000111
92d7eae5b9 add token classification example and gpt3 one layer test (#268) 2023-04-13 10:21:00 +08:00
hemu.zp
9940994d72 Add multi-modal embedding example for CLIP 
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/12251820
 2023-04-11 10:12:11 +08:00
yuze.zyz
4040320346 push to hub 
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/12235855
 2023-04-10 18:17:52 +08:00
hemu.zp
dd16c11f2a Add token classification example for MGeo 
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/12259147
 2023-04-10 17:08:35 +08:00
hemu.zp
f0d69c2aa4 Add palm example 
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/12259265
 2023-04-10 14:55:44 +08:00
hemu.zp
eecdd90412 add finetune_text_generation 
1. Add TrainingArgs and cli call form for text generation task
2. Fix dp+tp finetune bug
Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/11929345
 2023-03-10 15:08:56 +08:00
Firmament-cyou
8092a82577 add directory for image_classification finetuneing scripts (#115) 2023-02-21 11:33:41 +08:00
hemu.zp
82482b3e96 update training args 
Based on feat/0131/nlp_args branch, the original code review: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/11408570

Support for running finetuning from the command line with training args, Compatible with the configuration optimization.
 2023-02-10 05:32:21 +00:00
zhangzhicheng.zzc
42898badf7 [to #42322933] update ast_index logic 2023-01-11 10:43:56 +08:00
wenmeng.zwm
b8ec677739 add training args support and image classification fintune example 
design doc: https://yuque.antfin.com/pai/rwqgvl/khy4uw5dgi39s6ke

usage:
```python

    from modelscope.trainers.training_args import (ArgAttr, MSArgumentParser,
                                               training_args)


    training_args.topk = ArgAttr(cfg_node_name=['train.evaluation.metric_options.topk',
                                                'evaluation.metric_options.topk'],
                                 default=(1,), help='evaluation using topk, tuple format, eg (1,), (1,5)')
    training_args.train_data = ArgAttr(type=str, default='tany0699/cats_and_dogs', help='train dataset')
    training_args.validation_data = ArgAttr(type=str, default='tany0699/cats_and_dogs', help='validation dataset')
    training_args.model_id = ArgAttr(type=str, default='damo/cv_vit-base_image-classification_ImageNet-labels', help='model name')

    parser = MSArgumentParser(training_args)
    cfg_dict = parser.get_cfg_dict()
    args = parser.args
    
    train_dataset = create_dataset(args.train_data, split='train')
    val_dataset = create_dataset(args.validation_data, split='validation')

    def cfg_modify_fn(cfg):
        cfg.merge_from_dict(cfg_dict)
        return cfg

    kwargs = dict(
        model=args.model_id,          # model id
        train_dataset=train_dataset,  # training dataset
        eval_dataset=val_dataset,     # validation dataset
        cfg_modify_fn=cfg_modify_fn     # callback to modify configuration
    )

    trainer = build_trainer(name=Trainers.image_classification, default_args=kwargs)
    # start to train
    trainer.train()
```
        Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/11225071
 2022-12-30 07:35:15 +08:00