Open jhyukjang opened 1 year ago
Thx for your great work! I have some questions about your code.
- Did you set the random seed to 0 for all experiments?
- I reproduced as you did (NUM_GPUS=2, BS_FITS_YOUR_GPU=2) like below
ARROW_ROOT=./datasets/mmimdb NUM_GPUS=2 NUM_NODES=1 BS_FITS_YOUR_GPU=2 PRETRAINED_MODEL_PATH=./pretrained_weight/vilt_200k_mlm_itm.ckpt EXP_NAME=mmimdb
python run.py with data_root=${ARROW_ROOT} num_gpus=${NUM_GPUS} num_nodes=${NUM_NODES} per_gpu_batchsize=${BS_FITS_YOUR_GPU} task_finetune_mmimdb load_path=${PRETRAINED_MODEL_PATH} exp_name=${EXP_NAME}
and I got 40.65 (paper: 42.66) on test set with same setting. Can I reproduce the paper's results without changing parameters like a learning rate or are there some optimized hyperparameters for each dataset?
Hi.How many epochs
do you trained?
same problem.Hope the author can answer and help us
same problem, it would be helpful if the authors could provide more details on getting the results on paper
lots of version issue how did you handle guys?
Is the code complete, and it can't be run directly using the training command given by the author?
Thx for your great work! I have some questions about your code.
ARROW_ROOT=./datasets/mmimdb NUM_GPUS=2 NUM_NODES=1 BS_FITS_YOUR_GPU=2 PRETRAINED_MODEL_PATH=./pretrained_weight/vilt_200k_mlm_itm.ckpt EXP_NAME=mmimdb
python run.py with data_root=${ARROW_ROOT} \ num_gpus=${NUM_GPUS} \ num_nodes=${NUM_NODES} \ per_gpu_batchsize=${BS_FITS_YOUR_GPU} \ task_finetune_mmimdb \ load_path=${PRETRAINED_MODEL_PATH} \ exp_name=${EXP_NAME}
and I got 40.65 (paper: 42.66) on test set with same setting. Can I reproduce the paper's results without changing parameters like a learning rate or are there some optimized hyperparameters for each dataset?