中文

output

Model description

This model is a fine-tuned version of EleutherAI/gpt-neo-2.7B on the Lila-IID-train/dev set from the Lila dataset .

Usage

Bhaskara was trained with the following format:

Question: ...

Answer: ...

Program:
```python
...
```

It will perform best if queried in this way.

Intended uses & limitations

If you use this model, please cite our work.

@INPROCEEDINGS{Mishra2022Lila,
  author = {
    Swaroop Mishra 
      and Matthew Finlayson 
      and Pan Lu 
      and Leonard Tang 
      and Sean Welleck 
      and Chitta Baral 
      and Tanmay Rajpurohit 
      and Oyvind Tafjord 
      and Ashish Sabharwal 
      and Peter Clark 
      and Ashwin Kalyan},
  title = {Lila: A Unified Benchmark for Mathematical Reasoning},
  booktitle = {Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing (EMNLP)},
  year = {2022}
}

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 5e-05
  • train_batch_size: 4
  • eval_batch_size: 4
  • seed: 42
  • distributed_type: multi-GPU
  • num_devices: 2
  • total_train_batch_size: 8
  • total_eval_batch_size: 8
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 10.0

Training results

Training Loss Epoch Step Validation Loss Accuracy
No log 0.06 100 0.7930 0.8214
No log 0.11 200 0.7544 0.8290
No log 0.17 300 0.7358 0.8328
No log 0.23 400 0.7192 0.8357
0.8156 0.28 500 0.7012 0.8397
0.8156 0.34 600 0.6904 0.8419
0.8156 0.4 700 0.6802 0.8440
0.8156 0.45 800 0.6670 0.8465
0.8156 0.51 900 0.6572 0.8486
0.7219 0.57 1000 0.6499 0.8500
0.7219 0.62 1100 0.6411 0.8522
0.7219 0.68 1200 0.6343 0.8537
0.7219 0.74 1300 0.6299 0.8546
0.7219 0.79 1400 0.6221 0.8561
0.662 0.85 1500 0.6157 0.8574
0.662 0.91 1600 0.6138 0.8579
0.662 0.96 1700 0.6055 0.8595
0.662 1.02 1800 0.6143 0.8598
0.662 1.08 1900 0.6191 0.8599
0.5707 1.14 2000 0.6118 0.8607
0.5707 1.19 2100 0.6123 0.8611
0.5707 1.25 2200 0.6089 0.8617
0.5707 1.31 2300 0.6064 0.8619
0.5707 1.36 2400 0.6079 0.8625
0.4923 1.42 2500 0.6040 0.8625
0.4923 1.48 2600 0.6030 0.8630
0.4923 1.53 2700 0.6021 0.8636
0.4923 1.59 2800 0.6001 0.8643
0.4923 1.65 2900 0.5981 0.8644
0.4909 1.7 3000 0.5942 0.8648
0.4909 1.76 3100 0.5918 0.8650
0.4909 1.82 3200 0.5923 0.8659
0.4909 1.87 3300 0.5884 0.8664
0.4909 1.93 3400 0.5884 0.8663
0.4964 1.99 3500 0.5903 0.8669
0.4964 2.04 3600 0.6421 0.8655
0.4964 2.1 3700 0.6401 0.8651
0.4964 2.16 3800 0.6411 0.8649
0.4964 2.21 3900 0.6387 0.8645
0.345 2.27 4000 0.6362 0.8654
0.345 2.33 4100 0.6362 0.8654
0.345 2.38 4200 0.6362 0.8654
0.345 2.44 4300 0.6357 0.8655
0.345 2.5 4400 0.6362 0.8656
0.3463 2.55 4500 0.6377 0.8658
0.3463 2.61 4600 0.6357 0.8660
0.3463 2.67 4700 0.6294 0.8665
0.3463 2.72 4800 0.6333 0.8665
0.3463 2.78 4900 0.6362 0.8662
0.3508 2.84 5000 0.6357 0.8666
0.3508 2.89 5100 0.6299 0.8673
0.3508 2.95 5200 0.6313 0.8668
0.3508 3.01 5300 0.7188 0.8646
0.3508 3.06 5400 0.7017 0.8656
0.295 3.12 5500 0.6982 0.8653
0.295 3.18 5600 0.7031 0.8655
0.295 3.23 5700 0.6992 0.8651
0.295 3.29 5800 0.6997 0.8653
0.295 3.35 5900 0.7041 0.8651
0.2348 3.41 6000 0.7075 0.8649
0.2348 3.46 6100 0.6992 0.8650
0.2348 3.52 6200 0.7065 0.8647
0.2348 3.58 6300 0.6997 0.8652
0.2348 3.63 6400 0.7026 0.8651
0.2411 3.69 6500 0.7046 0.8656
0.2411 3.75 6600 0.7007 0.8655
0.2411 3.8 6700 0.7026 0.8651
0.2411 3.86 6800 0.7031 0.8655
0.2411 3.92 6900 0.7012 0.8658
0.251 3.97 7000 0.7051 0.8656
0.251 4.03 7100 0.7607 0.8650
0.251 4.09 7200 0.7632 0.8656
0.251 4.14 7300 0.7588 0.8655
0.251 4.2 7400 0.7578 0.8651
0.1797 4.26 7500 0.7710 0.8645
0.1797 4.31 7600 0.7627 0.8648
0.1797 4.37 7700 0.7583 0.8650
0.1797 4.43 7800 0.7646 0.8649
0.1797 4.48 7900 0.7598 0.8646
0.1784 4.54 8000 0.7656 0.8650
0.1784 4.6 8100 0.7617 0.8648
0.1784 4.65 8200 0.7573 0.8651
0.1784 4.71 8300 0.7671 0.8648
0.1784 4.77 8400 0.7563 0.8651
0.1827 4.82 8500 0.7651 0.8649
0.1827 4.88 8600 0.7637 0.8650
0.1827 4.94 8700 0.7607 0.8654
0.1827 4.99 8800 0.7607 0.8650
0.1827 5.05 8900 0.8149 0.8646
0.167 5.11 9000 0.8081 0.8648
0.167 5.16 9100 0.8184 0.8644
0.167 5.22 9200 0.8140 0.8647
0.167 5.28 9300 0.8169 0.8644
0.167 5.33 9400 0.8120 0.8645
0.1371 5.39 9500 0.8154 0.8643
0.1371 5.45 9600 0.8179 0.8642
0.1371 5.51 9700 0.8154 0.8643
0.1371 5.56 9800 0.8120 0.8645
0.1371 5.62 9900 0.8110 0.8650
0.1425 5.68 10000 0.8159 0.8645
0.1425 5.73 10100 0.8174 0.8646
0.1425 5.79 10200 0.8159 0.8649
0.1425 5.85 10300 0.8110 0.8639
0.1425 5.9 10400 0.8135 0.8645
0.1505 5.96 10500 0.8140 0.8642
0.1505 6.02 10600 0.8628 0.8640
0.1505 6.07 10700 0.8540 0.8644
0.1505 6.13 10800 0.8530 0.8642
0.1505 6.19 10900 0.8560 0.8647
0.1086 6.24 11000 0.8555 0.8649
0.1086 6.3 11100 0.8604 0.8644
0.1086 6.36 11200 0.8569 0.8642
0.1086 6.41 11300 0.8530 0.8639
0.1086 6.47 11400 0.8589 0.8643
0.1076 6.53 11500 0.8525 0.8639
0.1076 6.58 11600 0.8579 0.8640
0.1076 6.64 11700 0.8594 0.8640
0.1076 6.7 11800 0.8599 0.8643
0.1076 6.75 11900 0.8564 0.8640
0.1109 6.81 12000 0.8633 0.8640
0.1109 6.87 12100 0.8584 0.8638
0.1109 6.92 12200 0.8647 0.8636
0.1109 6.98 12300 0.8599 0.8635
0.1109 7.04 12400 0.8979 0.8632
0.1028 7.09 12500 0.8936 0.8635
0.1028 7.15 12600 0.9043 0.8637
0.1028 7.21 12700 0.8989 0.8642
0.1028 7.26 12800 0.8936 0.8642
0.1028 7.32 12900 0.8921 0.8641
0.0774 7.38 13000 0.8955 0.8634
0.0774 7.43 13100 0.8950 0.8636
0.0774 7.49 13200 0.8994 0.8635
0.0774 7.55 13300 0.8999 0.8635
0.0774 7.6 13400 0.8936 0.8631
0.0852 7.66 13500 0.9048 0.8634
0.0852 7.72 13600 0.8960 0.8632
0.0852 7.78 13700 0.9023 0.8635
0.0852 7.83 13800 0.8984 0.8638
0.0852 7.89 13900 0.9019 0.8635
0.0879 7.95 14000 0.9014 0.8634
0.0879 8.0 14100 0.9136 0.8630
0.0879 8.06 14200 0.9312 0.8639
0.0879 8.12 14300 0.9346 0.8635
0.0879 8.17 14400 0.9307 0.8635
0.0611 8.23 14500 0.9419 0.8641
0.0611 8.29 14600 0.9331 0.8631
0.0611 8.34 14700 0.9375 0.8636
0.0611 8.4 14800 0.9292 0.8626
0.0611 8.46 14900 0.9458 0.8637
0.061 8.51 15000 0.9336 0.8634
0.061 8.57 15100 0.9409 0.8630
0.061 8.63 15200 0.9390 0.8632
0.061 8.68 15300 0.9375 0.8628
0.061 8.74 15400 0.9365 0.8630
0.0646 8.8 15500 0.9370 0.8628
0.0646 8.85 15600 0.9355 0.8629
0.0646 8.91 15700 0.9375 0.8632
0.0646 8.97 15800 0.9390 0.8630
0.0646 9.02 15900 0.9717 0.8630
0.0593 9.08 16000 0.9673 0.8626
0.0593 9.14 16100 0.9644 0.8630
0.0593 9.19 16200 0.9624 0.8631
0.0593 9.25 16300 0.9648 0.8633
0.0593 9.31 16400 0.9673 0.8632
0.0415 9.36 16500 0.9658 0.8633
0.0415 9.42 16600 0.9688 0.8628
0.0415 9.48 16700 0.9653 0.8632
0.0415 9.53 16800 0.9658 0.8628
0.0415 9.59 16900 0.9668 0.8629
0.0471 9.65 17000 0.9604 0.8625
0.0471 9.7 17100 0.9658 0.8621
0.0471 9.76 17200 0.9731 0.8630
0.0471 9.82 17300 0.9692 0.8626
0.0471 9.88 17400 0.9673 0.8623
0.0528 9.93 17500 0.9614 0.8620
0.0528 9.99 17600 0.9697 0.8621

Framework versions

  • Transformers 4.21.0.dev0
  • Pytorch 1.12.1+cu113
  • Datasets 2.4.0
  • Tokenizers 0.12.1