nshepperd/gpt-2 issues and pull requests

#33 - Encode of a new dataset, confused about <|endoftext|> encoding

Issue - State: open - Opened by AliceZhang2016 about 5 years ago - 4 comments

#32 - Apologies, but HELP

Issue - State: open - Opened by OdincoGaming about 5 years ago - 4 comments

#32 - Apologies, but HELP

Issue - State: open - Opened by OdincoGaming about 5 years ago - 4 comments

#31 - Consued about vocab and encoder

Issue - State: open - Opened by weiguowilliam about 5 years ago

#31 - Consued about vocab and encoder

Issue - State: open - Opened by weiguowilliam about 5 years ago

#30 - fix for bigger models (774M)

Pull Request - State: open - Opened by mgrankin about 5 years ago

#30 - fix for bigger models (774M)

Pull Request - State: open - Opened by mgrankin about 5 years ago

#29 - Sampling structure looks weird. Maybe becuase I'm structuring my data wrongly?

Issue - State: open - Opened by dimeldo about 5 years ago

#29 - Sampling structure looks weird. Maybe becuase I'm structuring my data wrongly?

Issue - State: open - Opened by dimeldo about 5 years ago

#28 - Training on distributed machine is slow. Using 8 Nvidia V100.

Issue - State: open - Opened by dimeldo about 5 years ago - 8 comments

#27 - 使用 Colaboratory 创建

Pull Request - State: closed - Opened by LudoArt about 5 years ago

#27 - 使用 Colaboratory 创建

Pull Request - State: closed - Opened by LudoArt about 5 years ago

#26 - Train loss

Issue - State: open - Opened by alecalma about 5 years ago

#26 - Train loss

Issue - State: open - Opened by alecalma about 5 years ago

#25 - Question about training with small dataset entries

Issue - State: closed - Opened by ProtoxiDe22 about 5 years ago - 2 comments

#24 - 774M Model running out of memory

Issue - State: open - Opened by sdan about 5 years ago - 13 comments

#23 - Model directory as a parameter to encode.py

Pull Request - State: open - Opened by erickrf about 5 years ago

#23 - Model directory as a parameter to encode.py

Pull Request - State: open - Opened by erickrf about 5 years ago

#22 - Updated README.md

Pull Request - State: closed - Opened by biranchi2018 about 5 years ago

#22 - Updated README.md

Pull Request - State: closed - Opened by biranchi2018 about 5 years ago

#21 - how to train on multi gpu

Issue - State: open - Opened by brianjcj about 5 years ago - 1 comment

#21 - how to train on multi gpu

Issue - State: open - Opened by brianjcj about 5 years ago - 1 comment

#20 - module 'tensorflow' has no attribute 'sort'

Issue - State: closed - Opened by shoegazerstella over 5 years ago

#19 - Failed to interpret file %s as a pickle

Issue - State: closed - Opened by Ceebox over 5 years ago

#19 - Failed to interpret file %s as a pickle

Issue - State: closed - Opened by Ceebox over 5 years ago

#18 - Zero Division Error

Issue - State: open - Opened by Chris-Rigas over 5 years ago - 2 comments

#18 - Zero Division Error

Issue - State: open - Opened by Chris-Rigas over 5 years ago - 2 comments

#17 - Splitting the model across multiple graphics cards

Issue - State: open - Opened by thoughtsofacrow over 5 years ago

#17 - Splitting the model across multiple graphics cards

Issue - State: open - Opened by thoughtsofacrow over 5 years ago

#16 - "past" is not used in training

Issue - State: open - Opened by cookielee77 over 5 years ago - 2 comments

#16 - "past" is not used in training

Issue - State: open - Opened by cookielee77 over 5 years ago - 2 comments

#15 - Freezing layers while finetuning

Issue - State: open - Opened by bakszero over 5 years ago

#14 - generate_samples() code from gpt2 train.py -- InvalidArgumentError

Issue - State: closed - Opened by radiodee1 over 5 years ago - 1 comment

#13 - Intermediate Layer Output

Issue - State: open - Opened by bakszero over 5 years ago - 4 comments

#12 - Restriction on only training transformer layers?

Issue - State: closed - Opened by bakszero over 5 years ago - 6 comments

#11 - Training from scratch?

Issue - State: open - Opened by bkj over 5 years ago - 2 comments

#10 - [SOLUTION] UnicodeEncodeError: 'charmap' codec can't encode character '\ufffd' in position 29: character maps to <undefined>

Issue - State: open - Opened by ghost over 5 years ago - 1 comment

#9 - Windows doesn't automatically use UTF-8 encoding

Issue - State: open - Opened by MrKrzYch00 over 5 years ago - 2 comments

#8 - ResourceExhaustedError (see above for traceback): OOM when allocating tensor with shape[1,12,1024,1024] and type float on /job:localhost/replica:0/task:0/device:GPU

Issue - State: open - Opened by josai over 5 years ago - 12 comments

#7 - Edit encoding

Pull Request - State: closed - Opened by gyunggyung over 5 years ago - 1 comment

#6 - Edit encoding

Pull Request - State: closed - Opened by gyunggyung over 5 years ago - 1 comment

#5 - OOM With Gradient Checkpointing on 1080 Ti

Issue - State: open - Opened by 9of9 over 5 years ago - 2 comments

#4 - GPT-2: Q/A Training Question

Issue - State: open - Opened by josiahls over 5 years ago - 2 comments

#3 - Add toposort to requirements

Pull Request - State: closed - Opened by Tenoke over 5 years ago - 1 comment

#2 - Added instructions and script for distributed training with Horovod

Pull Request - State: closed - Opened by tlkh over 5 years ago - 3 comments

#1 - fixed too strict dataset sampling restriction

Pull Request - State: closed - Opened by lostmsu over 5 years ago - 6 comments

GitHub / nshepperd/gpt-2 issues and pull requests