small_scale_cql_chaotic_lstm_sweep-v0 | tlab Workspace

Howuhh's group workspace

Group: small_scale_cql_chaotic_lstm_sweep-v0

1-11

of 11

Tags

Notes

Author

howuhh

State

Finished

Start time

May 26th, 2023 8:37:53 AM

Sweep

k5ut0gpe

Runtime

6h 16m 5s

Tracked hours

6h 15m 56s

Run path

tlab/NetHack/fobltx1p

Linux-5.4.0-146-generic-x86_64-with-glibc2.29

Python version

3.8.10

Git repository

git clone git@github.com:tinkoff-ai/katakomba.git

Git state

git checkout -b "cql-mon-hum-neu-a1e30f02" 2bac99f941f8ffcbb7d2b4d878d22cacff02aca5

Command

algorithms/small_scale/cql_chaotic_lstm.py --alpha=5000 --character=mon-hum-neu --data_mode=memmap --group=small_scale_cql_chaotic_lstm_sweep --train_seed=2 --update_steps=250000 --version=0

System Hardware

CPU count	128
GPU count	1
GPU type	NVIDIA A100-PCIE-40GB

W&B CLI Version

0.12.21

Group

small_scale_cql_chaotic_lstm_sweep-v0

Config parameters are your model's inputs. Learn more

▶
Config parameters:{} 26 keys
- alpha:
  5,000
- batch_size:
  64
- character:
  "mon-hum-neu"
- checkpoints_path:
  null
- clip_grad_norm:
  null
- clip_range:
  10
- data_mode:
  "memmap"
- eval_episodes:
  50
- eval_every:
  10,000
- eval_processes:
  14
- eval_seed:
  50
- gamma:
  0.999
- group:
  "small_scale_cql_chaotic_lstm_sweep"
- learning_rate:
  0.0003
- name:
  "cql-mon-hum-neu-a1e30f02"
- project:
  "NetHack"
- rnn_dropout:
  0
- rnn_hidden_dim:
  2,048
- rnn_layers:
  2
- seq_len:
  16
- tau:
  0.005
- train_seed:
  2
- update_steps:
  250,000
- use_prev_action:
  true
- version:
  0
- weight_decay:
  0

Summary metrics are your model's outputs. Learn more

▶
Summary metrics:{} 21 keys
- cql_loss:
  1,723.2021484375
- depth_max:
  1
- depth_mean:
  0.12
- depth_median:
  0
- depth_min:
  0
- depth_std:
  0.32496153618543844
- loss:
  1,726.5411376953125
- q_target:
  9.887938499450684
- reward_max:
  3,845
- reward_mean:
  608.78
- reward_median:
  455
- reward_min:
  0
- reward_std:
  632.653057844503
- td_loss:
  3.3389830589294434
- times/backward_pass:
  0.013950400352478028
- times/batch_loading_cpu:
  0.047113656997680664
- times/batch_loading_gpu:
  0.04707369613647461
- times/evaluation_cpu:
  126.77609848976137
- times/evaluation_gpu:
  126.77534375
- times/forward_pass:
  0.00894976043701172
- transitions:
  256,000,000

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

wandb-history

run-fobltx1p-history:v0