Ecosyste.ms: Issues
An open API service for providing issue and pull request metadata for open source projects.
GitHub / THUDM/GLM issues and pull requests
#91 - glm_10B_chinese在finetune的时候需要多久,目前已经六个小时还未结束,运行的命令是github上给出的bash scripts/generate_block.sh \ config_tasks/model_blocklm_10B_chinese.sh,且一直未有log输出 ,但gpu是有利用率的
Issue -
State: closed - Opened by haiqizhang over 1 year ago
- 3 comments
#90 - 1
Issue -
State: closed - Opened by zhangyipin over 1 year ago
#89 - MPU module
Issue -
State: closed - Opened by Ant0082 over 1 year ago
- 2 comments
#88 - For `GLM-10B-Chinese`, the fine-tuning loss barely decrease within each epoch and it only decreases when starting a new epoch.
Issue -
State: closed - Opened by silverriver over 1 year ago
- 3 comments
#87 - 生成结果“随机性固定”的问题
Issue -
State: closed - Opened by yipintiancheng over 1 year ago
- 2 comments
#86 - ImportError: cannot import name 'torch_required' from 'transformers.utils'
Issue -
State: closed - Opened by sh0416 over 1 year ago
- 2 comments
#85 - GLM-10B chinese and MP_SIZE= 2 for pretrain just stay in the function of get_train_val_test_data ?
Issue -
State: closed - Opened by LovesportsMcDull over 1 year ago
- 3 comments
#84 - generate empty sample
Issue -
State: closed - Opened by mx8435 over 1 year ago
#82 - 运行scripts/generate_block.sh,在生成的过程中中断并报错
Issue -
State: closed - Opened by haiqizhang over 1 year ago
- 2 comments
#81 - Questions about 10B-chinese
Issue -
State: closed - Opened by mx8435 over 1 year ago
- 2 comments
#80 - 同一个句子中多个[MASK]无法同时预测
Issue -
State: closed - Opened by robotsp over 1 year ago
- 2 comments
Labels: enhancement
#79 - How to finetune for text generation?
Issue -
State: closed - Opened by ouyangliqi over 1 year ago
- 10 comments
#78 - AutoModelForMultipleChoice无法加载glm-large-chinese模型
Issue -
State: closed - Opened by Lollipop over 1 year ago
- 2 comments
#77 - Add text classification examples on rotten_tomatoes and emotion datasets
Pull Request -
State: open - Opened by atfortes over 1 year ago
- 1 comment
#76 - KQA Pro example added!
Pull Request -
State: closed - Opened by jiudingsun01 over 1 year ago
#75 - Question about how to finetune 10b-chinese model for summarization task
Issue -
State: open - Opened by siyuanxue over 1 year ago
#74 - classification task using the 'art' dataset
Pull Request -
State: open - Opened by liku-amare over 1 year ago
#73 - Classification task commonsense_qa and Generation task multi_news
Pull Request -
State: open - Opened by REIGN12 over 1 year ago
#72 - How are the escape characters '\n' or '\t' in data processed during pretraining or finetuing?
Issue -
State: closed - Opened by Tebmer over 1 year ago
- 8 comments
#71 - 如何操作:glm-10b-chinese不做finetune直接加载pretrained model做inference
Issue -
State: closed - Opened by haiqizhang over 1 year ago
- 13 comments
#70 - Bug of finetuning code? the attention mask of padding is not 0.
Issue -
State: closed - Opened by Tebmer over 1 year ago
- 2 comments
#69 - glm-10B-chinese是如何finetune的,运行的脚本文件是哪个
Issue -
State: closed - Opened by haiqizhang over 1 year ago
- 1 comment
#68 - Deepspeed zero stage 3
Issue -
State: open - Opened by Porraio over 1 year ago
- 3 comments
#67 - 如果用 AutoModelForSeq2SeqLM 的格式进行下游finetune 后 除了使用save_pretrained 方法进行储存外 还需要进行哪些操作 才能再次用 AutoModelForSeq2SeqLM.from_pretrained本地初始化?
Issue -
State: open - Opened by svjack over 1 year ago
- 1 comment
#66 - Generation task on squad dataset.
Pull Request -
State: open - Opened by yuwenmichael over 1 year ago
#65 - 4bit quantization of the 10b model
Issue -
State: open - Opened by phills11 over 1 year ago
#64 - Model Warmup for ICL
Issue -
State: closed - Opened by Ant0082 over 1 year ago
- 2 comments
#63 - Can not reproduce SQuAD v1.1 result using GLM-Large
Issue -
State: closed - Opened by cklsoft over 1 year ago
- 1 comment
#62 - Align test speed
Pull Request -
State: closed - Opened by ccssu over 1 year ago
#61 - Why not release GLM-base-chinese?
Issue -
State: closed - Opened by mx8435 over 1 year ago
#60 - Train the glm-10B-chinese model using 4 V100 GPUs, with no error logs printed, and then exit
Issue -
State: closed - Opened by Ant0082 over 1 year ago
- 6 comments
#59 - The pretraining corpus of GLM-Large-Chinese
Issue -
State: closed - Opened by cklsoft over 1 year ago
- 1 comment
#58 - Hello, below are some questions I encountered while learning code, I hope you can answer them when you have time, thank you.
Issue -
State: closed - Opened by Ant0082 over 1 year ago
- 1 comment
#57 - Aboutlength
Issue -
State: closed - Opened by llllooong over 1 year ago
#56 - How many cards do you need to fine-tune this model?
Issue -
State: closed - Opened by Ant0082 over 1 year ago
#55 - In `GLM-10B-Chinese`, token id for `[gMASK]` and `[eop]` is the same. Is it a designed behavior?
Issue -
State: closed - Opened by silverriver almost 2 years ago
- 1 comment
#54 - Unrecognized configuration class
Issue -
State: closed - Opened by 980202006 almost 2 years ago
- 1 comment
#53 - Which config is used to pretrain the released `GLM-10B-Chinese` model? is `ds_block_10B_chinese_longer.sh` or `ds_block_10B_chinese.sh`
Issue -
State: closed - Opened by silverriver almost 2 years ago
- 1 comment
#52 - Unable to use `AutoModelForSeq2SeqLM`
Issue -
State: closed - Opened by larrylawl almost 2 years ago
- 3 comments
#51 - convert pretrained pt to huggingface
Issue -
State: closed - Opened by xv44586 almost 2 years ago
- 1 comment
#50 - add examples directory and related requirements
Pull Request -
State: closed - Opened by Xiao9905 almost 2 years ago
- 1 comment
#49 - Create Examples for GLM
Pull Request -
State: closed - Opened by Xiao9905 almost 2 years ago
#48 - CVE-2007-4559 Patch
Pull Request -
State: closed - Opened by TrellixVulnTeam almost 2 years ago
#47 - Accelerate the model inference of GLM-10B
Issue -
State: closed - Opened by Ant0082 almost 2 years ago
- 2 comments
#46 - Hardware requirements for GLM-chinese-10B
Issue -
State: open - Opened by shaomai00 almost 2 years ago
- 9 comments
#45 - fix df_finetune_seq2seq.sh save path
Pull Request -
State: closed - Opened by xv44586 almost 2 years ago
#44 - Information about those new released multi-task model
Issue -
State: closed - Opened by siriusctrl almost 2 years ago
- 1 comment
#43 - 自定义tokenizer
Issue -
State: closed - Opened by maojinyang almost 2 years ago
#42 - Hardware requirements
Issue -
State: open - Opened by eli-halych almost 2 years ago
Labels: documentation
#41 - 模型权重加载问题
Issue -
State: closed - Opened by maojinyang almost 2 years ago
- 2 comments
#40 - run infer failed
Issue -
State: closed - Opened by xv44586 almost 2 years ago
- 4 comments
#39 - how to choose the finetuning script for question-answering task
Issue -
State: closed - Opened by shaomai00 almost 2 years ago
- 2 comments
#38 - 运行(bash scripts/generate_block.sh config_tasks/model_blocklm_10B_chinese.sh)代码时生成的文本与示例中的不一致
Issue -
State: closed - Opened by Ant0082 almost 2 years ago
- 2 comments
#37 - remove unused imports
Pull Request -
State: closed - Opened by WrRan almost 2 years ago
#36 - typo
Pull Request -
State: closed - Opened by WrRan almost 2 years ago
#35 - 配置问题
Issue -
State: open - Opened by zfstr almost 2 years ago
- 3 comments
#34 - Google Colab error
Issue -
State: open - Opened by MultiTrickFox almost 2 years ago
- 1 comment
#33 - generate有没有并行的方法
Issue -
State: closed - Opened by debby1103 about 2 years ago
- 3 comments
#32 - continue pretrain的时候遇到loss scale的问题,怎么解决?
Issue -
State: closed - Opened by dinglei8908 about 2 years ago
- 4 comments
#31 - Training and inference issue
Issue -
State: closed - Opened by qtli about 2 years ago
- 3 comments
#30 - text infilling cases
Issue -
State: closed - Opened by qtli about 2 years ago
- 11 comments
#29 - Optimizer state when changing MP(Model Parallelism) SIZE
Issue -
State: closed - Opened by SeonggwanAhn about 2 years ago
- 1 comment
#28 - Simple questions on GLM pretraining mechanism
Issue -
State: closed - Opened by buttercutter about 2 years ago
- 1 comment
#27 - 有没有模型压缩量化等方法提高模型的推理性能
Issue -
State: closed - Opened by jiangliqin about 2 years ago
- 3 comments
#26 - GLM-2B-Chinese Pretrained Model please
Issue -
State: closed - Opened by lockmatrix about 2 years ago
- 1 comment
#25 - GLM-chinese compare to nezha and roformer-v2 ?
Issue -
State: closed - Opened by XiaoqingNLP about 2 years ago
#24 - 继续预训练如何加载模型?
Issue -
State: closed - Opened by fade-color about 2 years ago
- 4 comments
#23 - The multi-task learning setting is different from the original paper
Issue -
State: closed - Opened by Aurora-slz about 2 years ago
- 1 comment
#22 - If the data of a row exceeds 512, will the excess part be discarded?
Issue -
State: closed - Opened by Aurora-slz about 2 years ago
- 1 comment
#21 - Why vocabulary is divided by GPU number and how to load it?
Issue -
State: closed - Opened by Aurora-slz about 2 years ago
- 3 comments
#20 - attention mask between spans
Issue -
State: closed - Opened by ChuanTianML over 2 years ago
- 1 comment
#19 - Evaluation on SQuAD
Issue -
State: closed - Opened by SeonggwanAhn over 2 years ago
- 1 comment
#18 - 10B 中文模型有下游数据集的效果么?
Issue -
State: closed - Opened by OleNet over 2 years ago
- 1 comment
#17 - model.backward在iter == arg.eval-interval时卡住无法继续
Issue -
State: closed - Opened by xikaluo over 2 years ago
- 3 comments
#16 - Can you provide the test.json file mentioned in TestDataset in data_utils/corpora.py?
Issue -
State: closed - Opened by zhufq00 over 2 years ago
- 2 comments
#15 - 请问有没有什么办法可以快速的使用GLM得到词向量呢?
Issue -
State: open - Opened by wangleiai over 2 years ago
#14 - EOFError: Ran out of input
Issue -
State: open - Opened by helloeng over 2 years ago
#13 - WARNING: could not find the metadata file /root/data/checkpoints/blocklm-large-chinese/latest_checkpointed_iteration.txt
Issue -
State: closed - Opened by Xiang-Pan over 2 years ago
- 1 comment
#12 - [quesiton] block_position_ids的含义
Issue -
State: closed - Opened by starkhu over 2 years ago
- 1 comment
#11 - HuggingFace module
Issue -
State: closed - Opened by peregilk about 3 years ago
- 3 comments
#10 - does the GLM perform well than bert on text similarity task and ner task?
Issue -
State: open - Opened by boy-be-ambitious about 3 years ago
#9 - How to pre-train? has anyone start pre-train successful?
Issue -
State: open - Opened by wdyxwzyh about 3 years ago
- 4 comments
#8 - Are pretraining codes released?
Issue -
State: closed - Opened by swaggy-TN about 3 years ago
#7 - Do we have a GLM model of Chinese version?
Issue -
State: open - Opened by ericg108 over 3 years ago
#6 - About "Text Summarization"
Issue -
State: open - Opened by ShiYaya over 3 years ago
#5 - Does it support multilingual ?
Issue -
State: open - Opened by puraminy over 3 years ago
#4 - Can't run models : RuntimeError: CUDA error: invalid device ordinal
Issue -
State: closed - Opened by puraminy over 3 years ago
- 1 comment
#3 - Hi, i got two points.
Issue -
State: open - Opened by wdyxwzyh over 3 years ago
#2 - Will you release this code in the future?
Issue -
State: closed - Opened by yikedouer over 3 years ago
- 1 comment
#1 - Consider using an different acronym than "GLM"
Issue -
State: open - Opened by ledell over 3 years ago
- 4 comments