one-per-step | gpt Overview – Weights & Biases

Skip to main content

Clashluke's group workspace

Group: one-per-step

1-2

of 2

Tags

Notes

Author

State

Crashed

Start time

May 15th, 2023 4:52:46 PM

Runtime

2h 3m 31s

Tracked hours

-

Run path

homebrewnlp/gpt/10uey6jh151q3zkj7re51vju10otixfl

OS

Linux-5.4.0-1043-gcp-x86_64-with-glibc2.29

Python version

3.8.5

Git repository

git clone https://github.com/HomebrewNLP/HomebrewNLP-Jax

Git state

git checkout -b "one-per-step-0" 8603d150f9ac4b932ddf5c1a186683efdb61eca7

Command

main.py

System Hardware

CPU count	48
Logical CPU count	96

W&B CLI Version

0.15.2

Group

Config parameters are your model's inputs. Learn more

▶
Config parameters:{} 11 keys
- ▶
  data:{} 8 keys
- ▶
  dims:{} 10 keys
- fail_on_missing_parameter:
  true
- global_prefix:
  ""
- ▶
  model:{} 4 keys
  - autoregressive:
    true
  - computation_dtype:
    "bfloat16"
  - ▶
    norm:{} 1 key
    - eps:
      0.0000000000000001
  - storage_dtype:
    "float32"
- name_cache_offsets:{} 0 keys
- ▶
  optimizer:{} 12 keys
- parameter_usages:{} 0 keys
- seed:
  0
- ▶
  training:{} 10 keys
- ▶
  wandb:{} 6 keys

Summary metrics are your model's outputs. Learn more

▶
Summary metrics:{} 13 keys
- Accuracy/Current:
  0.02392578125
- Accuracy/Median64:
  0.021728515625
- Accuracy/Median256:
  0.021484375
- Accuracy/Median1024:
  0.02197265625
- Loss/Current:
  0.3970825800079254
- Loss/Median64:
  0.4030077844284268
- Loss/Median256:
  0.40370949709895
- Loss/Median1024:
  0.40306877827752974
- Optimizer/Learning Rate:
  0.00000795023333921563
- Speed/Batches per Second:
  353.86011980089427
- Speed/Parameters * Tokens per Day:
  209,852,723,797,527,812,374,528
- Speed/Tokens per Day:
  128,234,613,535,606.38
- Speed/Tokens Seen:
  12,582,656

This run produced these artifacts as outputs. Total: 1. Learn more

run-10uey6jh151q3zkj7re51vju10otixfl-history:v0

Loading...