Ecosyste.ms: Issues
An open API service for providing issue and pull request metadata for open source projects.
GitHub / airaria/TextBrewer issues and pull requests
#121 - 学生模型权重初始化问题
Issue -
State: open - Opened by cgh-code777 6 months ago
- 2 comments
#120 - 请问支持BERT-of-Theseus的蒸馏方式吗
Issue -
State: closed - Opened by zhanghanweii 11 months ago
- 3 comments
Labels: stale
#119 - 麻烦问下,目前支持llama模型吗
Issue -
State: closed - Opened by StevensPrime about 1 year ago
- 2 comments
Labels: stale
#118 - 可以使用chatgpt蒸馏到bert或者T5吗?
Issue -
State: closed - Opened by Hoogck about 1 year ago
- 2 comments
Labels: stale
#117 - 蒸馏后的模型进行evaluate,报错AxisError: axis 2 is out of bounds for array of dimension 1
Issue -
State: closed - Opened by MrRace about 1 year ago
- 5 comments
Labels: stale
#116 - 老师您好,我想问一下,比如roberta蒸馏到tinybert,中间的hidden是通过线性层拉到同样的维度去算mse,那在推理的时候岂不是这些经过梯度更新的线性层毫无作用?那请问这些线性层仅仅就是为了调整维度?
Issue -
State: closed - Opened by lean-wang about 1 year ago
- 2 comments
#115 - 老师,您好,请问有多任务多教师的蒸馏的demo吗?
Issue -
State: closed - Opened by lean-wang about 1 year ago
- 4 comments
Labels: stale
#114 - msra_ner.ipynb最后的trainer.evaluate()显示CUDA out of memory,请问训练的显存要求是多大?十分感谢!
Issue -
State: closed - Opened by jinxiaolinlin over 1 year ago
- 2 comments
Labels: stale
#113 - 不同维度蒸馏有对应的例子吗,从768降到256
Issue -
State: closed - Opened by weidalan over 1 year ago
- 4 comments
Labels: stale
#112 - notebook_examples/msra_ner.ipynb 运行报错
Issue -
State: closed - Opened by MrRace over 1 year ago
- 12 comments
Labels: stale
#111 - 关于ner数据的处理
Issue -
State: closed - Opened by Soulscb over 1 year ago
- 2 comments
Labels: stale
#110 - 在VisionTransformer
Issue -
State: closed - Opened by zym1599 over 1 year ago
- 5 comments
Labels: stale
#109 - CVE-2007-4559 Patch
Pull Request -
State: closed - Opened by TrellixVulnTeam almost 2 years ago
- 2 comments
Labels: stale
#108 - Does it support translation model?
Issue -
State: closed - Opened by AIikai almost 2 years ago
- 2 comments
Labels: stale
#107 - How about the distillation effect of gpt2 ?
Issue -
State: closed - Opened by xk503775229 almost 2 years ago
- 2 comments
Labels: stale
#106 - Picking right layers
Issue -
State: closed - Opened by patryk-at-pieces almost 2 years ago
- 3 comments
Labels: stale
#105 - Fix typo errors in README.md
Pull Request -
State: closed - Opened by mehdie79 almost 2 years ago
#104 - Show the progress bar when training.
Issue -
State: closed - Opened by Gridnn about 2 years ago
- 3 comments
Labels: stale
#103 - interpreting intermediate matches
Issue -
State: closed - Opened by kaliaanup about 2 years ago
- 5 comments
Labels: stale
#102 - Where to find gs4210.pkl file or how to generate it ? thanks
Issue -
State: closed - Opened by Shaukat-Hussain about 2 years ago
- 2 comments
Labels: stale
#101 - pre-trained student weights
Issue -
State: closed - Opened by roymiles about 2 years ago
- 3 comments
Labels: stale
#100 - Modify batch type checking in distiller_utils.py
Pull Request -
State: closed - Opened by yourplatanus about 2 years ago
- 2 comments
#99 - Add conditional statement in distiller_utils.py
Pull Request -
State: closed - Opened by yourplatanus about 2 years ago
- 1 comment
#98 - TextBrewer/src/textbrewer/distiller_utils.py get_outputs_from_batch fails tocheck dicts properly for maskedLM
Issue -
State: closed - Opened by AddedK over 2 years ago
- 4 comments
#97 - TextBrewer/examples/notebook_examples/msra_ner.ipynb have bug?
Issue -
State: closed - Opened by 645187919 over 2 years ago
- 1 comment
#96 - 请问可以直接用于unilm中的NLU和NLG吗?
Issue -
State: closed - Opened by cingtiye over 2 years ago
- 1 comment
Labels: stale
#95 - 如何实现early stopping
Issue -
State: closed - Opened by yuange555 over 2 years ago
#94 - 请问有添加早停机制的打算吗?
Issue -
State: closed - Opened by catqaq over 2 years ago
- 3 comments
#93 - 如何蒸馏不分层的新特征?
Issue -
State: closed - Opened by catqaq over 2 years ago
- 2 comments
#92 - 请问有针对BertForMaskedLM的蒸馏示例吗
Issue -
State: closed - Opened by dongteng over 2 years ago
- 3 comments
Labels: stale
#91 - 关于损失函数的问题
Issue -
State: closed - Opened by LLLLLLoki over 2 years ago
- 1 comment
#90 - 关于损失函数的问题
Issue -
State: closed - Opened by LLLLLLoki over 2 years ago
- 5 comments
Labels: stale
#89 - 关于任务无关的蒸馏
Issue -
State: closed - Opened by savannahfan over 2 years ago
- 1 comment
#88 - mnli main_train
Issue -
State: closed - Opened by Soulscb over 2 years ago
- 7 comments
Labels: stale
#87 - 是否能做预训练的蒸馏
Issue -
State: closed - Opened by YoungErm over 2 years ago
- 3 comments
Labels: stale
#86 - 中间层Loss,会去更新后面网络的参数吗
Issue -
State: closed - Opened by DvHuang over 2 years ago
- 6 comments
Labels: stale
#85 - msra_ner.ipynb复现代码bug
Issue -
State: closed - Opened by HXYstudy over 2 years ago
- 7 comments
Labels: stale
#84 - can't open examples/notebook_examples/sst2.ipynb in colab
Issue -
State: closed - Opened by josephcui over 2 years ago
- 2 comments
Labels: stale
#83 - Is there any methods can simplify the output during training?
Issue -
State: closed - Opened by HarryHe11 over 2 years ago
- 4 comments
Labels: stale
#82 - Notebook JSON is invalid
Issue -
State: closed - Opened by tanyaroosta almost 3 years ago
- 1 comment
#81 - random_token_example error
Issue -
State: closed - Opened by tanyaroosta almost 3 years ago
- 5 comments
Labels: stale
#80 - PyTorch Lightning
Issue -
State: closed - Opened by tchaton almost 3 years ago
- 2 comments
Labels: stale
#79 - examples/random_token_example, when I run python distill.py then exception Killed
Issue -
State: closed - Opened by dulante00 almost 3 years ago
- 2 comments
Labels: stale
#78 - CUDA Error with your Notebook Example
Issue -
State: closed - Opened by cabisarri almost 3 years ago
- 2 comments
Labels: stale
#77 - Cuda Error in scripts
Issue -
State: closed - Opened by cabisarri almost 3 years ago
#76 - 使用自定义的网络结构
Issue -
State: closed - Opened by zhangatao almost 3 years ago
- 2 comments
#75 - Add note examples
Pull Request -
State: closed - Opened by lokwq almost 3 years ago
#74 - Add note examples test2
Pull Request -
State: closed - Opened by lokwq almost 3 years ago
#73 - Add note examples
Pull Request -
State: closed - Opened by lokwq almost 3 years ago
#72 - 关于MNLI任务复现问题
Issue -
State: closed - Opened by sunnan-nn about 3 years ago
- 2 comments
Labels: stale
#71 - 请问此框架的loss函数是否存在问题?
Issue -
State: closed - Opened by Jay2Coomzz about 3 years ago
- 2 comments
Labels: stale
#70 - 模型没有被训练,每个epoch保存的模型weight一模一样。
Issue -
State: closed - Opened by Jay2Coomzz about 3 years ago
- 4 comments
Labels: stale
#69 - GeneralDistiller的train函数报错
Issue -
State: closed - Opened by Jay2Coomzz about 3 years ago
- 6 comments
Labels: stale
#68 - 关于中文阅读理解数据集t4学生模型蒸馏配置的问题
Issue -
State: closed - Opened by SouthBays about 3 years ago
- 5 comments
Labels: stale
#67 - examples/mnli_example: run_mnli_train.sh 模型没被训练
Issue -
State: closed - Opened by Yin169 about 3 years ago
- 15 comments
Labels: stale
#66 - Data preparation
Issue -
State: closed - Opened by liuhl-source about 3 years ago
- 4 comments
Labels: stale
#65 - Type Error: __init__() got an unexpected keyword argument 'unk_token'
Issue -
State: closed - Opened by SnailDM about 3 years ago
- 4 comments
Labels: stale
#64 - earth mover distance的优化求解是如何实现可导的
Issue -
State: closed - Opened by Youarerare about 3 years ago
- 2 comments
Labels: stale
#63 - 维度不匹配问题
Issue -
State: closed - Opened by iamxinxin about 3 years ago
- 1 comment
#62 - 训练mnli任务教师模型准确率很低?
Issue -
State: closed - Opened by binglinchengxiash about 3 years ago
- 1 comment
#61 - 请问下在mnli任务训练中,第一步训练的是教师模型,配置为啥加载的是学生模型的配置?
Issue -
State: closed - Opened by binglinchengxiash about 3 years ago
- 2 comments
#60 - mnli问题
Issue -
State: closed - Opened by binglinchengxiash about 3 years ago
- 2 comments
#59 - 运行时错误,询问。
Issue -
State: closed - Opened by Noeverer about 3 years ago
- 4 comments
Labels: stale
#58 - 运行时出现错误
Issue -
State: closed - Opened by Noeverer about 3 years ago
#57 - 无法复习结果
Issue -
State: closed - Opened by houpanpan over 3 years ago
- 6 comments
Labels: stale
#56 - kd_loss不下降,准确率也基本不动,但是隐层好像拟合的很好,是什么原因呢?
Issue -
State: closed - Opened by Youarerare over 3 years ago
- 1 comment
#55 - RuntimeError: Incoming model is an instance of torch.nn.parallel.DataParallel. Parallel wrappers should only be applied to the model(s) AFTER the model(s) have been returned from amp.initialize.
Issue -
State: closed - Opened by Youarerare over 3 years ago
- 4 comments
#54 - Rename DistillMultiBert.json to DistillMultiBertToTiny.json
Pull Request -
State: closed - Opened by DA-southampton over 3 years ago
#53 - 更换模型无法复现finetune结果
Issue -
State: closed - Opened by houpanpan over 3 years ago
- 3 comments
Labels: stale
#52 - 直接运行ner_ElectraTrain_dist.sh,在半个epoch F1可以达到90以上。你再检查下数据或模型是不是有问题?
Issue -
State: closed - Opened by houpanpan over 3 years ago
- 5 comments
Labels: stale
#51 - NER 任务中teacher model 和student model的matches如匹配
Issue -
State: closed - Opened by houpanpan over 3 years ago
- 3 comments
Labels: stale
#50 - 不支持多标签分类任务蒸馏的原因
Issue -
State: closed - Opened by codemayq over 3 years ago
- 3 comments
#49 - msra命名实体识别任务无法复现结果
Issue -
State: closed - Opened by houpanpan over 3 years ago
- 13 comments
Labels: stale
#48 - msra命名实体识别train 过程无法复现结果
Issue -
State: closed - Opened by houpanpan over 3 years ago
- 2 comments
Labels: stale
#47 - update mnli_exmaple and bert-emd
Pull Request -
State: closed - Opened by airaria over 3 years ago
#46 - update mnli_exmaple and bert-emd
Pull Request -
State: closed - Opened by airaria over 3 years ago
#45 - 请问一下msra实体识别任务中,matches没有给参数,默认是用那种蒸馏方式啊?
Issue -
State: closed - Opened by SunyanGu over 3 years ago
- 3 comments
Labels: stale
#44 - How to customize teacher and student models, such as VGG in CV?
Issue -
State: closed - Opened by LiXuanming over 3 years ago
- 3 comments
Labels: stale
#43 - [Question] 论文中数据增强方法的一些疑惑?
Issue -
State: closed - Opened by wangyuxinwhy over 3 years ago
- 2 comments
Labels: stale
#42 - 关于初始化训练好的student模型做推理预测问题
Issue -
State: closed - Opened by chenjun0210 over 3 years ago
- 1 comment
#41 - mnli_example实验相关
Issue -
State: closed - Opened by qq260612718 over 3 years ago
- 6 comments
Labels: stale
#40 - 请问作者有做中文数据增强吗?
Issue -
State: closed - Opened by hahlw over 3 years ago
- 5 comments
Labels: stale
#39 - Hard loss 有用吗?
Issue -
State: closed - Opened by sz128 over 3 years ago
- 12 comments
Labels: stale
#38 - TypeError: model should be either torch.nn.Module or a dict
Issue -
State: closed - Opened by dakota987 over 3 years ago
- 1 comment
#37 - How to perform data augmentation?
Issue -
State: closed - Opened by LorrinWWW over 3 years ago
- 7 comments
#36 - 关于MMD loss
Issue -
State: closed - Opened by sz128 over 3 years ago
- 5 comments
Labels: stale
#35 - 蒸馏的时候映射层的对应关系是什么样的
Issue -
State: closed - Opened by 652994331 over 3 years ago
- 3 comments
#34 - 对roberta-wwm-ext蒸馏遇到维度不匹配
Issue -
State: closed - Opened by junrong1 over 3 years ago
- 6 comments
#33 - 蒸馏效果不好,请问一下怎么解决?
Issue -
State: closed - Opened by cgq0816 over 3 years ago
- 8 comments
Labels: stale
#32 - shall we give some examples about text classification like cmrc、ner tasks?
Issue -
State: closed - Opened by AlexYoung757 over 3 years ago
- 3 comments
Labels: stale
#31 - the link is fail
Issue -
State: closed - Opened by Soulscb over 3 years ago
- 3 comments
Labels: stale
#30 - 对BERT-wwm-ext进行蒸馏时遇到以下问题,代码已贴出
Issue -
State: closed - Opened by cgq0816 over 3 years ago
- 10 comments
#29 - V0.2.1dev
Pull Request -
State: closed - Opened by airaria over 3 years ago
- 2 comments
#28 - 蒸馏后学生模型乱码
Issue -
State: closed - Opened by peixin-lin over 3 years ago
- 7 comments
Labels: stale
#27 - self.d_config.is_caching_logits=True about results_T
Issue -
State: closed - Opened by chenhaoenen almost 4 years ago
- 1 comment
#26 - 请问有没有BERT蒸馏到简单模型的sample,比如说BiGRU、CNN之类的?
Issue -
State: closed - Opened by yongzhuo almost 4 years ago
- 1 comment
#25 - What is CustomMatch?
Issue -
State: closed - Opened by zhujiangang almost 4 years ago
- 1 comment
#24 - 命名`temperature_scheduler`的合理性
Issue -
State: closed - Opened by shenfe almost 4 years ago
- 1 comment
#23 - 在对中间层求mse loss的时候会出现这层loss 不收敛,其他不同种类的loss 都会有不同程度的收敛?
Issue -
State: closed - Opened by LLLLLLoki almost 4 years ago
- 6 comments
#22 - Question on different input format for teacher and student models
Issue -
State: closed - Opened by meisyarahd almost 4 years ago
- 4 comments