Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Add replication instructions for training #53

Merged
merged 4 commits into from
Sep 27, 2023
Merged

Add replication instructions for training #53

Changes from 1 commit
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
6baf3c4
add draft replicability instructions
Sep 26, 2023
6df7364
add note about global BS
Sep 27, 2023
1bc7bf0
unify commits used
Sep 27, 2023
e8c9cb3
update neox submodule pointer
Sep 27, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
update neox submodule pointer
  • Loading branch information
haileyschoelkopf committed Sep 27, 2023
commit e8c9cb32f09b3ce7dd189ac4f4895be56f70462e
2 changes: 1 addition & 1 deletion gpt-neox
Open in desktop
Submodule gpt-neox updated 90 files
+1 −8 .gitignore
+3 −3 README.md
+3 −11 configs/neox_arguments.md
+0 −101 configs/v0.4/1.4b-1000-early.yml
+0 −101 configs/v0.4/1.4b-12000-early.yml
+0 −101 configs/v0.4/1.4b-128-early.yml
+0 −101 configs/v0.4/1.4b-2000-early.yml
+0 −101 configs/v0.4/1.4b-24000-early.yml
+0 −101 configs/v0.4/1.4b-256-early.yml
+0 −101 configs/v0.4/1.4b-48000-early.yml
+0 −101 configs/v0.4/1.4b-6000-early.yml
+0 −101 configs/v0.4/1.4b-64-early.yml
+0 −101 configs/v0.4/1.4b-71000-early.yml
+0 −101 configs/v0.4/1.4b-8-early.yml
+0 −99 configs/v0.5/410M_flash.yml
+0 −94 configs/v0.5/6-9B.yml
+0 −97 configs/v0.5/6-9B_flash.yml
+0 −99 configs/v0.5/6-9B_universal_flash.yml
+0 −99 configs/v0.5/70M_flash.yml
+0 −106 configs/v0.5/llama_7B.yml
+0 −106 configs/v0.5/llama_7B_correct-scheduler.yml
+0 −106 configs/v0.5/llama_7B_correct-scheduler_3e-5.yml
+0 −106 configs/v0.5/llama_7B_correct-scheduler_low-lr.yml
+0 −107 configs/v0.5/llama_7B_correct-scheduler_resume.yml
+0 −107 configs/v0.5/llama_7B_correct-scheduler_validation.yml
+0 −104 configs/v0.5/llama_7B_resume.yml
+0 −102 configs/v0.5/llama_7B_scratch.yml
+0 −106 configs/v0.6/llama_7b.yml
+0 −106 configs/v0.6/llama_7b_1e-5.yml
+127 −12 megatron/checkpointing.py
+14 −16 megatron/learning_rates.py
+32 −19 megatron/model/positional_embeddings.py
+113 −67 megatron/model/transformer.py
+0 −3 megatron/neox_arguments/deepspeed_args.py
+34 −4 megatron/neox_arguments/neox_args.py
+1 −1 megatron/tokenizer/train_tokenizer.py
+23 −5 megatron/training.py
+1 −1 requirements/requirements-flashattention.txt
+2 −0 requirements/requirements.txt
+0 −60 slurm/v0.4/1.4b-1000-early.sh
+0 −60 slurm/v0.4/1.4b-12000-early.sh
+0 −60 slurm/v0.4/1.4b-128-early.sh
+0 −60 slurm/v0.4/1.4b-2000-early.sh
+0 −60 slurm/v0.4/1.4b-24000-early.sh
+0 −60 slurm/v0.4/1.4b-256-early.sh
+0 −60 slurm/v0.4/1.4b-48000-early.sh
+0 −60 slurm/v0.4/1.4b-6000-early.sh
+0 −60 slurm/v0.4/1.4b-64-early.sh
+0 −60 slurm/v0.4/1.4b-71000-early.sh
+0 −60 slurm/v0.4/1.4b-8-early.sh
+ − slurm/v0.5/01_all.png
+ − slurm/v0.5/02_fixed-checkpoint.png
+0 −39 slurm/v0.5/410m_flash.sh
+0 −35 slurm/v0.5/6-9b.sh
+0 −38 slurm/v0.5/6-9b_flash.sh
+0 −39 slurm/v0.5/70m_flash.sh
+0 −21 slurm/v0.5/README.md
+0 −42 slurm/v0.5/convert_llama_to_hf.sh
+0 −39 slurm/v0.5/llama_7b.sh
+0 −39 slurm/v0.5/llama_7b_correct-scheduler.sh
+0 −39 slurm/v0.5/llama_7b_correct-scheduler_3e-5.yml
+0 −39 slurm/v0.5/llama_7b_correct-scheduler_low-lr.sh
+0 −39 slurm/v0.5/llama_7b_correct-scheduler_resume.sh
+0 −39 slurm/v0.5/llama_7b_correct-scheduler_validation.sh
+0 −39 slurm/v0.5/llama_7b_resume.sh
+0 −39 slurm/v0.5/llama_7b_scratch.sh
+0 −42 slurm/v0.6/convert_llama_to_hf.sh
+0 −39 slurm/v0.6/llama_7b.yml
+0 −39 slurm/v0.6/llama_7b_1e-5.yml
+0 −15 tools/README.md
+0 −50 tools/checkpoints/README.md
+0 −434 tools/checkpoints/convert_v1.0_to_hf.py
+0 −225 tools/checkpoints/deepspeed_to_deepspeed.py
+0 −360 tools/checkpoints/ds_to_universal.py
+0 −128 tools/checkpoints/inspect_ds_checkpoint.py
+0 −441 tools/convert_llama_sequential_to_hf.py
+2 −2 tools/convert_raw_llama_weights_to_neox.py
+0 −0 tools/convert_sequential_to_hf.py
+18 −12 tools/corpora.py
+0 −0 tools/inspect_checkpoints.py
+0 −0 tools/kill.sh
+0 −0 tools/killall.sh
+0 −0 tools/merge20b.py
+0 −0 tools/merge_datasets.py
+0 −0 tools/merge_mp_partitions.py
+5 −9 tools/preprocess_data.py
+0 −0 tools/sync.sh
+0 −0 tools/sync_cmd.sh
+0 −0 tools/syncdir.sh
+0 −0 tools/upload.py