PaddlePaddle/PaddleFormers v1.2.0

PaddleFormers v1.2

Repository: PaddlePaddle/PaddleFormers

Tag: v1.2.0

Published: 2026-06-19T14:10:57Z

Prerelease: no

Release notes:

What's Changed

• [CI]add build scripts by @Liujie0926 in https://github.com/PaddlePaddle/PaddleFormers/pull/2433
• [CI]update bucket for daily build by @Liujie0926 in https://github.com/PaddlePaddle/PaddleFormers/pull/2441
• Add tensor parallelism on QLoRA by @tugang-baidu in https://github.com/PaddlePaddle/PaddleFormers/pull/2424
• Cherry-pick hybrid expert parallel sharding_metas by @pkuzyc in https://github.com/PaddlePaddle/PaddleFormers/pull/2447
• Solve DPO pin-memory problem by hacking HybridParallelOptimizer by @WYB27 in https://github.com/PaddlePaddle/PaddleFormers/pull/2428
• lazy more_elegant by @miao200years in https://github.com/PaddlePaddle/PaddleFormers/pull/2451
• support multi download source by @fjjF77 in https://github.com/PaddlePaddle/PaddleFormers/pull/2427
• Fix paddle.distributed.checkpoint path by @xingmingyyj in https://github.com/PaddlePaddle/PaddleFormers/pull/2452
• fix sentencepiece.bpe.model download by @fjjF77 in https://github.com/PaddlePaddle/PaddleFormers/pull/2454
• hf tokenizer adaptation by @SdeeRK in https://github.com/PaddlePaddle/PaddleFormers/pull/2445
• Support general design for modeling by @cheng221 in https://github.com/PaddlePaddle/PaddleFormers/pull/2446
• [CI]add ce yml by @Liujie0926 in https://github.com/PaddlePaddle/PaddleFormers/pull/2468
• Legacy tokenizer by @SdeeRK in https://github.com/PaddlePaddle/PaddleFormers/pull/2465
• Tokenizer update by @SdeeRK in https://github.com/PaddlePaddle/PaddleFormers/pull/2457
• fix vl lora uc. by @wtmlon in https://github.com/PaddlePaddle/PaddleFormers/pull/2463
• FIX: E_cpu by @miao200years in https://github.com/PaddlePaddle/PaddleFormers/pull/2475
• fix import download_utils & support ci set network proxy by @fjjF77 in https://github.com/PaddlePaddle/PaddleFormers/pull/2477
• [BugFix] fix lazy_import error of importlib.machinery by @EmmonsCurse in https://github.com/PaddlePaddle/PaddleFormers/pull/2482
• [feature] lazyimport-and-tokenizer by @miao200years in https://github.com/PaddlePaddle/PaddleFormers/pull/2481
• model download source add ernie by @fjjF77 in https://github.com/PaddlePaddle/PaddleFormers/pull/2484
• fix test_configuration download model ci bug by @fjjF77 in https://github.com/PaddlePaddle/PaddleFormers/pull/2488
• Support HF torch load & save by @llbdyiu66 in https://github.com/PaddlePaddle/PaddleFormers/pull/2437
• delete import distutils in pdc_utils by @cheng221 in https://github.com/PaddlePaddle/PaddleFormers/pull/2493
• [feature] testcase-without-paddle by @miao200years in https://github.com/PaddlePaddle/PaddleFormers/pull/2494
• Try import ckpt convert by @xingmingyyj in https://github.com/PaddlePaddle/PaddleFormers/pull/2476
• add sink_attention by @xxyux in https://github.com/PaddlePaddle/PaddleFormers/pull/2461
• fix uc tp name mapping. by @wtmlon in https://github.com/PaddlePaddle/PaddleFormers/pull/2502
• [CI] update trigger conditions by @Liujie0926 in https://github.com/PaddlePaddle/PaddleFormers/pull/2499
• [FIx_v0.2] PreTrainedTokenizer by @miao200years in https://github.com/PaddlePaddle/PaddleFormers/pull/2498
• support ci which download models from hugging face by @fjjF77 in https://github.com/PaddlePaddle/PaddleFormers/pull/2490
• enhance apply_chat_template by @SdeeRK in https://github.com/PaddlePaddle/PaddleFormers/pull/2513
• add train sft examples by @llbdyiu66 in https://github.com/PaddlePaddle/PaddleFormers/pull/2491
• remove bos download by @fjjF77 in https://github.com/PaddlePaddle/PaddleFormers/pull/2517
• [CI]Add codecov by @Liujie0926 in https://github.com/PaddlePaddle/PaddleFormers/pull/2528
• add decode_token function by @SdeeRK in https://github.com/PaddlePaddle/PaddleFormers/pull/2519
• fix ernie4_5vl tokenizer unitest for network connection error by @fjjF77 in https://github.com/PaddlePaddle/PaddleFormers/pull/2529
• change apply_chat_template add_generation_prompt param by @SdeeRK in https://github.com/PaddlePaddle/PaddleFormers/pull/2533
• [CE]fix ce scripts by @Liujie0926 in https://github.com/PaddlePaddle/PaddleFormers/pull/2534
• update v0.2 by @lugimzzz in https://github.com/PaddlePaddle/PaddleFormers/pull/2540
• [CI] Update image by @Liujie0926 in https://github.com/PaddlePaddle/PaddleFormers/pull/2542
• [BugFix] fix decode_token by @yuanlehome in https://github.com/PaddlePaddle/PaddleFormers/pull/2544
• move text data streams from Erniekit to PaddleFormers. by @Jonathans575 in https://github.com/PaddlePaddle/PaddleFormers/pull/2453
• [fix] ADDTOKEN by @miao200years in https://github.com/PaddlePaddle/PaddleFormers/pull/2545
• Add ernie4 5 moe by @cheng221 in https://github.com/PaddlePaddle/PaddleFormers/pull/2520
• fix mistakes change in tp by @cheng221 in https://github.com/PaddlePaddle/PaddleFormers/pull/2550
• [BufFix] fix decode_token by @yuanlehome in https://github.com/PaddlePaddle/PaddleFormers/pull/2553
• fix_decode_token by @yuanlehome in https://github.com/PaddlePaddle/PaddleFormers/pull/2559
• fix general pipeline model by @cheng221 in https://github.com/PaddlePaddle/PaddleFormers/pull/2560
• add test_tokenizer_decode_token by @SdeeRK in https://github.com/PaddlePaddle/PaddleFormers/pull/2562
• 【model】add Gpt oss model support sft/lora and infer by @xiaoguoguo626807 in https://github.com/PaddlePaddle/PaddleFormers/pull/2555
• remove old dataset. by @Jonathans575 in https://github.com/PaddlePaddle/PaddleFormers/pull/2561
• fix chat_template bug. by @Jonathans575 in https://github.com/PaddlePaddle/PaddleFormers/pull/2552
• update examples, add dpo & lora training by @llbdyiu66 in https://github.com/PaddlePaddle/PaddleFormers/pull/2563
• Refactoring Qwen2/3 with general design by @Ace-To-HYB in https://github.com/PaddlePaddle/PaddleFormers/pull/2480
• 【Fix Bug】fix startend_row_indices bug by @cheng221 in https://github.com/PaddlePaddle/PaddleFormers/pull/2565
• update v0.3 by @lugimzzz in https://github.com/PaddlePaddle/PaddleFormers/pull/2567
• add estimate max_steps by @Jonathans575 in https://github.com/PaddlePaddle/PaddleFormers/pull/2566
• fix pp_seg_method and unfiy training attention with attn_impl by @cheng221 in https://github.com/PaddlePaddle/PaddleFormers/pull/2572
• add estimate training in dpo training by @Jonathans575 in https://github.com/PaddlePaddle/PaddleFormers/pull/2573
• rename loss_subbatch_seqlen to loss_subbatch_sequence_length by @cheng221 in...