kallewoof 58ec8b1113 feature: loss watchdog for terminating training runs that are failing (#899) ...

Co-authored-by: Karl-Johan Alm <kalle@gmail.com>

2023-12-04 07:54:34 -05:00

..

cerebras

don't compile deepspeed or bitsandbytes from source (#837)

2023-11-08 19:49:55 -05:00

code-llama

don't compile deepspeed or bitsandbytes from source (#837)

2023-11-08 19:49:55 -05:00

falcon

don't compile deepspeed or bitsandbytes from source (#837)

2023-11-08 19:49:55 -05:00

gptj

don't compile deepspeed or bitsandbytes from source (#837)

2023-11-08 19:49:55 -05:00

jeopardy-bot

fix eval_steps to be a sane default (#797)

2023-10-27 22:36:30 -04:00

llama-2

various bugfixes (#856)

2023-11-15 12:23:18 -05:00

mistral

feature: loss watchdog for terminating training runs that are failing (#899)

2023-12-04 07:54:34 -05:00

mpt-7b

fix eval_steps to be a sane default (#797)

2023-10-27 22:36:30 -04:00

openllama-3b

don't compile deepspeed or bitsandbytes from source (#837)

2023-11-08 19:49:55 -05:00

phi

Phi update 202311 (#876)

2023-11-17 12:47:17 -05:00

pythia

fix eval_steps to be a sane default (#797)

2023-10-27 22:36:30 -04:00

pythia-12b

simplify by removing duplicate base_model_config (#772)

2023-10-23 01:42:38 -04:00

qwen

fix: remove FA for qwen examples (#900)

2023-11-27 21:23:54 +09:00

redpajama

fix eval_steps to be a sane default (#797)

2023-10-27 22:36:30 -04:00

replit-3b

fix eval_steps to be a sane default (#797)

2023-10-27 22:36:30 -04:00

xgen-7b

don't compile deepspeed or bitsandbytes from source (#837)

2023-11-08 19:49:55 -05:00