时隔5个月，MindNLP 0.4来啦！！！这一次在开源社区同学们的努力下带来了数个重大特性更新！！！

1. 模型大量更新，支持250+ transformers模型，覆盖90% huggingface模型，使用MindSpore进行模型复现、开发训练再无迁移之忧（社区会同步huggingface和业界新模型，随时更新）。
2. 全平台支持：全面支持昇腾910A、910B、310B(香橙派)、GPU、CPU。（PS：目前香橙派唯一可用的AI开发套件）
3. 分布式并行推理：支持多卡多进程并行推理（超过10B的模型可以玩啦！）
4. 量化算法支持：香橙派支持Smooth Quant，GPU支持bitsandbytes int8量化
5. Sentence transformer支持：可以开心的搞RAG啦
6. 动态图性能优化：昇腾硬件下动态图推理速度达到Pytorch+GPU同水平（Llama实测性能85ms/token）
7. 真正的动静统一：通过mindspore.jit一行切换图模式，完全使用huggingface code style，既保证易用性，又能快速提升性能。昇腾硬件下Llama实测性能为动态图2倍(45ms/token)，与mindspore其他静态图based套件性能完全一致。
8. 海量LLM应用更新：包括文本信息抽取、聊天机器人、语音识别、ChatPDF、音乐生成、代码生成、声音克隆等等，伴随着模型支持度的提高，更多好玩的应用等你来开发！

欢迎大家使用、提Issue和Star一下~

安装方法：

pip install mindnlp

代码仓：
https://github.com/mindspore-lab/mindnlp

PS：获取昇腾算力的几个途径：

• 参加MindSpore社区实习和社区活动，免费领取华为云代金券：https://gitee.com/mindspore/community/issues/IAUJE8
• 使用启智社区免费算力：https://www.openi.org.cn/

What's Changed

• add musicgen gradio demo by @lvyufeng in #1066
• Peft lokr support by @ultranationalism in #1067
• use mkdocs to generate documentation by @lvyufeng in #1070
• Add tutorial quick start by @hubosynapse in #1069
• fix precision error of flash_attn kernel in some case by @hypertseng in #1072
• support prompt_tuning by @lvyufeng in #1073
• add peft integration by @lvyufeng in #1076
• implement cogvlm model and demo by @antigone660 in #1075
• fix pipeline base error by @lvyufeng in #1077
• peft module documentation completion by @ultranationalism in #1078
• Add MindNLP Starcoder vscode demo by @xing-yiren in #1079
• fix starcoder example by @lvyufeng in #1080
• update docs by @lvyufeng in #1084

New Contributors

• @ultranationalism made their first contribution in #1067
• @antigone660 made their first contribution in #1075
• @xing-yiren made their first contribution in #1079

Full Changelog: v0.3.0...v0.3.1

What's Changed

• add new Trainer like hf-transformers
• Add timesformer, MPNet, pegasus, segformer, sam, llama3, MPT,
  olmo, llava_next, llava, vipllava, convnext, cvt, resnet, van, openelm, phi3
• support flash_attn_v2, flash_attn_bwd
• peft module support ia3, adalora
• support resume download
• add gsm8k cot prompt example

New Contributors

• @Brain100destruction made their first contribution in #1038
• @wjy4399 made their first contribution in #1042
• @zhouyifeng888 made their first contribution in #1047

Full Changelog: v0.2.4...v0.3.0

What's Changed

• update flashattention kernel by @hypertseng in #962
• support for internlm by @NANDSS1 in #961
• add qwen2_moe & fix bugs by @lvyufeng in #965
• skip big and slow ut on CPU by @lvyufeng in #966
• Add table_transformer model by @ShaoqLin in #976
• fix find_cuda_home raise error by @lvyufeng in #967
• fix error on macOS by @lvyufeng in #977
• add jamba by @lvyufeng in #978
• add bit by @lvyufeng in #979
• add musicgen_melody by @lvyufeng in #980
• Fix tokenized index problem by @WilliamLiuAtCPC in #982
• add text2vec module by @lvyufeng in #983
• support blenderbot by @lvyufeng in #984
• support blenderbot samll by @lvyufeng in #985
• support blip by @lvyufeng in #987
• Add ConvBert model and pass ut. by @neoming in #991
• add model: xlnet by @2802427218 in #988
• support blip_2 by @lvyufeng in #993
• support bridgetower & bros model by @lvyufeng in #994
• Added hypercomplex Tensor Decomposition and example of using it by @maxmgrdv in #995
• loar fintune convbert by @neoming in #997
• support jetmoe & fix python id() caused bugs by @lvyufeng in #998
• update supported list by @lvyufeng in #999

New Contributors

• @NANDSS1 made their first contribution in #961
• @ShaoqLin made their first contribution in #976
• @WilliamLiuAtCPC made their first contribution in #982

Full Changelog: v0.2.3...v0.2.4

New models

• ALIGN
• AltCLIP
• Audio Spectrogram Transformer
• BioGPT

What's Changed

• add tokenization_layoutlm by @FLoutione in #935
• support wav2vec2_with_lm by @lvyufeng in #936
• fix whisper inference on Ascend by @lvyufeng in #937
• fix whisper return_token_timestamps error by @lvyufeng in #938
• add evaluate module by @lvyufeng in #939
• add biogpt by @lvyufeng in #940
• support align model by @lvyufeng in #942
• fix _scaled_dot_product_attention bug by @hypertseng in #941
• add altclip & ast model by @lvyufeng in #944

Full Changelog: v0.2.1...v0.2.2

New features

Pipelines support

• Automatic Speech Recognition
• Question Answering
• Text to Text
• Text Classification
• Text Generation

What's Changed

• update readme by @lvyufeng in #889
• correct bark encodec inference bug by @JinPan-Second in #892
• support starcoder2 by @lvyufeng in #898
• big_bird has a ut failure by @stringency in #905
• fix baichuan error by @lvyufeng in #909
• fix big_bird error by @lvyufeng in #910
• support mixtral by @lvyufeng in #912
• windows use synchronize since fatal errors by @lvyufeng in #914
• support mamba by @lvyufeng in #916
• mamba support graph mode by @lvyufeng in #919
• Text generation by @2802427218 in #918
• Text2Text generation pipeline feat by @fanxing-6 in #917
• fix codegen error on GPU by @lvyufeng in #920
• Added dual-valued Bert model and example of using it by @maxmgrdv in #921
• fix hypercomplex casued errors by @lvyufeng in #922
• add:question answering by @2802427218 in #923
• fix mixtral ut on windows by @lvyufeng in #924
• layoutlm module by @FLoutione in #925
• update peft examples by @lvyufeng in #926
• udate ci pipeline by @lvyufeng in #927
• support deberta model by @lvyufeng in #928
• add bce example by @lvyufeng in #930
• support reformer by @lvyufeng in #931
• fix roll not support on CPU by @lvyufeng in #932
• support automatic_speech_recognition pipeline by @lvyufeng in #934

New Contributors

• @stringency made their first contribution in #905
• @2802427218 made their first contribution in #918
• @fanxing-6 made their first contribution in #917
• @maxmgrdv made their first contribution in #921

Full Changelog: v0.2.0...v0.2.1