File size: 3,384 Bytes
6ff677e ff7e5aa 6ff677e 91a58cf 6ff677e 3fcba1d 6ff677e 91a58cf 6ff677e 91a58cf 79fcabb 91a58cf 79fcabb 91a58cf 79fcabb 91a58cf 79fcabb 91a58cf 79fcabb 91a58cf 79fcabb 91a58cf 6ff677e fb0ccea 79fcabb fb0ccea 6ff677e |
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 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 |
---
license: apache-2.0
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
---
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# vqa
This model is a fine-tuned version of [unc-nlp/lxmert-base-uncased](https://huggingface.co/unc-nlp/lxmert-base-uncased) on the [Graphcore/vqa-lxmert](https://huggingface.co/datasets/Graphcore/vqa-lxmert) dataset.
It achieves the following results on the evaluation set:
- Loss: 0.0009
- Accuracy: 0.7242
## Model description
LXMERT is a transformer model for learning vision-and-language cross-modality representations. It has a Transformer model that has three encoders: object relationship encoder, a language encoder, and a cross-modality encoder. It is pretrained via a combination of masked language modeling, visual-language text alignment, ROI-feature regression, masked visual-attribute modeling, masked visual-object modeling, and visual-question answering objectives. It acheives the state-of-the-art results on VQA anad GQA.
Paper link : [LXMERT: Learning Cross-Modality Encoder Representations from Transformers](https://arxiv.org/pdf/1908.07490.pdf)
## Intended uses & limitations
More information needed
## Training and evaluation data
[Graphcore/vqa-lxmert](https://huggingface.co/datasets/Graphcore/vqa-lxmert) dataset
## Training procedure
Trained on 16 Graphcore Mk2 IPUs using [optimum-graphcore](https://github.com/huggingface/optimum-graphcore).
Command line:
```
python examples/question-answering/run_vqa.py \
--model_name_or_path unc-nlp/lxmert-base-uncased \
--ipu_config_name Graphcore/lxmert-base-ipu \
--dataset_name Graphcore/vqa-lxmert \
--do_train \
--do_eval \
--max_seq_length 512 \
--per_device_train_batch_size 1 \
--num_train_epochs 4 \
--dataloader_num_workers 64 \
--logging_steps 5 \
--learning_rate 5e-5 \
--lr_scheduler_type linear \
--loss_scaling 16384 \
--weight_decay 0.01 \
--warmup_ratio 0.1 \
--output_dir /tmp/vqa/ \
--dataloader_drop_last \
--replace_qa_head \
--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
```
***** train metrics *****
"epoch": 4.0,
"train_loss": 0.0060005393999575125,
"train_runtime": 13854.802,
"train_samples": 443757,
"train_samples_per_second": 128.116,
"train_steps_per_second": 2.002
***** eval metrics *****
"eval_accuracy": 0.7242196202278137,
"eval_loss": 0.0008745193481445312,
"eval_samples": 214354,
```
### Framework versions
- Transformers 4.18.0.dev0
- Pytorch 1.10.0+cpu
- Datasets 2.0.0
- Tokenizers 0.11.6
|