lxmert-vqa-uncased / README.md
Jinchen's picture
Update README.md
91a58cf
|
raw
history blame
2.56 kB
metadata
tags:
  - generated_from_trainer
datasets:
  - Graphcore/vqa-lxmert
metrics:
  - accuracy
model-index:
  - name: vqa
    results:
      - task:
          name: Question Answering
          type: question-answering
        dataset:
          name: Graphcore/vqa-lxmert
          type: Graphcore/vqa-lxmert
          args: vqa
        metrics:
          - name: Accuracy
            type: accuracy
            value: 0.7242196202278137

vqa

This model is a fine-tuned version of unc-nlp/lxmert-base-uncased on the Graphcore/vqa-lxmert dataset. It achieves the following results on the evaluation set:

  • Loss: 0.0009
  • Accuracy: 0.7242

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

Graphcore/vqa-lxmert dataset

Training procedure

Trained on 16 Graphcore Mk2 IPUs using optimum-graphcore.

Command line:

python examples/language-modeling/run_clm.py \
  --model_name_or_path gpt2 \
  --ipu_config_name Graphcore/gpt2-small-ipu \
  --dataset_name wikitext \
  --dataset_config_name wikitext-103-raw-v1 \
  --do_train \
  --do_eval \
  --num_train_epochs 10 \
  --dataloader_num_workers 64 \
  --per_device_train_batch_size 1 \
  --per_device_eval_batch_size 1 \
  --gradient_accumulation_steps 128 \
  --output_dir /tmp/clm_output \
  --logging_steps 5 \
  --learning_rate 1e-5 \
  --lr_scheduler_type linear \
  --loss_scaling 16384 \
  --weight_decay 0.01 \
  --warmup_ratio 0.1 \
  --ipu_config_overrides="embedding_serialization_factor=4,optimizer_state_offchip=true,inference_device_iterations=5" \
  --dataloader_drop_last \
  --pod_type pod16

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 5e-05
  • train_batch_size: 1
  • eval_batch_size: 8
  • seed: 42
  • distributed_type: IPU
  • total_train_batch_size: 64
  • total_eval_batch_size: 128
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_ratio: 0.1
  • num_epochs: 4.0
  • training precision: Mixed Precision

Training results

Framework versions

  • Transformers 4.18.0.dev0
  • Pytorch 1.10.0+cpu
  • Datasets 2.0.0
  • Tokenizers 0.11.6