Spaces:
Runtime error
Runtime error
File size: 1,687 Bytes
c5ca37a |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 |
description: Train Causal on Yahoo Dataset
auth:
# which virtual cluster you belong to (msrlabs, etc.). Everyone has access to "pnrsy".
vc: resrchprojvc6
# physical cluster to use (cam, gcr, rr1) or Azure clusters (eu1, eu2, etc.)
# cluster: rr2, eu2, eu1 et1
cluster: rr1 # eu2
# docker environment (vm) in which your job will run. we provide "generic" dockers
# with the main deep learning toolkit installed (PyTorch, TF, Chainer, etc.)
docker:
# image: philly/jobs/custom/generic-docker:py27
# registry: phillyregistry.azurecr.io
image: chunyl/pytorch-transformers:v0
registry: index.docker.io
storage:
_default:
#use_phillyfs: True
storage_account_name: textae
container_name: bigtextae
mount_path: /mnt/_default
code:
# local directory of the code. this will be uploaded to the server.
# $CONFIG_DIR is expanded to the directory of this config file
code_upload: False
remote_dir: code/
local_dir: $CONFIG_DIR/code
#data:
# data upload is not required for this example
#data_upload: False
search:
job_template:
name: gpt2_{experiment_name:s}_{bs_option:.0f}
sku: G8 # G4 # G1
command:
- pip install --user --editable .
- python examples/big_ae/run_lm_finetuning_baseline.py --output_dir ../output/philly_clm_yahoo_gpt2 --dataset Yahoo --model_type gpt2 --model_name_or_path gpt2 --do_train --train_data_file ../data/datasets/yahoo_data/train.txt --do_eval --eval_data_file ../data/datasets/yahoo_data/valid.txt --overwrite_output_dir --per_gpu_train_batch_size {bs_option}
max_trials: 20
type: grid
params:
- name: bs_option
spec: discrete
values: [4] # [top,bottom]
|