metadata

tags:
  - generated_from_trainer
datasets:
  - squad_v2
model-index:
  - name: roberta-finetuned-squad_v2
    results:
      - task:
          type: question-answering
          name: Question Answering
        dataset:
          name: SQuAD v2
          type: squad_v2
          split: validation
        metrics:
          - type: exact
            value: 100
            name: Exact
          - type: f1
            value: 100
            name: F1
          - type: total
            value: 2
            name: Total
          - type: HasAns_exact
            value: 100
            name: Hasans_exact
          - type: HasAns_f1
            value: 100
            name: Hasans_f1
          - type: HasAns_total
            value: 2
            name: Hasans_total
          - type: best_exact
            value: 100
            name: Best_exact
          - type: best_exact_thresh
            value: 0.9603068232536316
            name: Best_exact_thresh
          - type: best_f1
            value: 100
            name: Best_f1
          - type: best_f1_thresh
            value: 0.9603068232536316
            name: Best_f1_thresh
          - type: total_time_in_seconds
            value: 0.036892927000735654
            name: Total_time_in_seconds
          - type: samples_per_second
            value: 54.21093316776193
            name: Samples_per_second
          - type: latency_in_seconds
            value: 0.018446463500367827
            name: Latency_in_seconds

roberta-finetuned-squad_v2

This model was trained from scratch on the squad_v2 dataset. It achieves the following results on the evaluation set:

Loss: 0.8582

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 2e-05
train_batch_size: 128
eval_batch_size: 128
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 512
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 4

Training results

Training Loss	Epoch	Step	Validation Loss
2.9129	0.2	100	1.4700
1.4395	0.39	200	1.2407
1.2356	0.59	300	1.0325
1.1284	0.78	400	0.9750
1.0821	0.98	500	0.9345
0.9978	1.18	600	0.9893
0.9697	1.37	700	0.9300
0.9455	1.57	800	0.9351
0.9322	1.76	900	0.9451
0.9269	1.96	1000	0.9064
0.9105	2.16	1100	0.8837
0.8805	2.35	1200	0.8876
0.8703	2.55	1300	0.9853
0.8699	2.75	1400	0.9235
0.8633	2.94	1500	0.8930
0.828	3.14	1600	0.8582
0.8284	3.33	1700	0.9203
0.8076	3.53	1800	0.8866
0.7805	3.73	1900	0.9099
0.7974	3.92	2000	0.8746

Framework versions

Transformers 4.34.1
Pytorch 2.1.0+cu118
Datasets 2.14.5
Tokenizers 0.14.1