Mitsuki-Sakamoto's picture
Model save
4272544 verified
metadata
license: apache-2.0
base_model: EleutherAI/pythia-160m
tags:
  - generated_from_trainer
model-index:
  - name: pythia_160m_alpaca_farm_instructions_sft_constant_pa_seed_1
    results: []

pythia_160m_alpaca_farm_instructions_sft_constant_pa_seed_1

This model is a fine-tuned version of EleutherAI/pythia-160m on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 2.1686

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: 8e-06
  • train_batch_size: 4
  • eval_batch_size: 8
  • seed: 1
  • distributed_type: multi-GPU
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: constant
  • num_epochs: 3.0

Training results

Training Loss Epoch Step Validation Loss
2.3096 0.02 50 2.2544
2.2692 0.04 100 2.2374
2.2021 0.06 150 2.2228
2.2268 0.08 200 2.2338
2.1433 0.1 250 2.2146
2.0708 0.12 300 2.2004
2.163 0.14 350 2.1996
2.2518 0.16 400 2.1898
2.0717 0.18 450 2.1899
2.2137 0.2 500 2.1847
2.2232 0.22 550 2.1760
2.2455 0.24 600 2.1757
2.1936 0.26 650 2.1732
2.1352 0.28 700 2.1619
2.1215 0.3 750 2.1608
2.1568 0.32 800 2.1506
2.1319 0.34 850 2.1514
2.0831 0.36 900 2.1494
2.0788 0.38 950 2.1430
2.0901 0.4 1000 2.1376
2.1374 0.42 1050 2.1343
1.9484 0.44 1100 2.1298
2.204 0.46 1150 2.1284
2.108 0.48 1200 2.1249
1.9353 0.5 1250 2.1210
2.1352 0.52 1300 2.1178
1.9498 0.54 1350 2.1162
2.1571 0.56 1400 2.1153
2.1804 0.58 1450 2.1114
1.988 0.6 1500 2.1107
2.0485 0.62 1550 2.1055
2.0596 0.64 1600 2.1020
1.98 0.66 1650 2.1027
2.0626 0.68 1700 2.0980
2.097 0.7 1750 2.0949
2.2013 0.72 1800 2.0893
2.1234 0.74 1850 2.0913
1.9662 0.76 1900 2.0971
2.138 0.78 1950 2.0929
2.0816 0.8 2000 2.0898
2.1506 0.82 2050 2.0848
2.0585 0.84 2100 2.0860
2.099 0.86 2150 2.0862
2.084 0.88 2200 2.0816
2.1046 0.9 2250 2.0790
2.02 0.92 2300 2.0865
2.0548 0.94 2350 2.0776
2.0819 0.96 2400 2.0766
1.9181 0.98 2450 2.0755
2.0345 1.0 2500 2.0793
1.7741 1.02 2550 2.0922
1.6556 1.04 2600 2.0921
1.6168 1.06 2650 2.0921
1.8017 1.08 2700 2.0927
1.8055 1.1 2750 2.0893
1.7298 1.12 2800 2.0910
1.6924 1.14 2850 2.0969
1.853 1.16 2900 2.0951
1.7641 1.18 2950 2.1020
1.7529 1.2 3000 2.0991
1.7556 1.22 3050 2.1005
1.7273 1.24 3100 2.0984
1.8478 1.26 3150 2.1000
1.8965 1.28 3200 2.0932
1.761 1.3 3250 2.0917
1.7579 1.32 3300 2.0943
1.7347 1.34 3350 2.0914
1.7725 1.36 3400 2.0928
1.8931 1.38 3450 2.0913
1.7301 1.4 3500 2.1030
1.741 1.42 3550 2.0953
1.8009 1.44 3600 2.0971
1.8397 1.46 3650 2.0932
1.7941 1.48 3700 2.0932
1.7136 1.5 3750 2.0936
1.723 1.52 3800 2.0913
1.7837 1.54 3850 2.0878
1.7988 1.56 3900 2.0859
1.7759 1.58 3950 2.0883
1.8608 1.6 4000 2.0926
1.5859 1.62 4050 2.0918
1.8474 1.64 4100 2.0888
1.7921 1.66 4150 2.0932
1.755 1.68 4200 2.0950
1.8437 1.7 4250 2.0880
1.826 1.72 4300 2.0861
1.8548 1.74 4350 2.0886
1.7668 1.76 4400 2.0832
1.7818 1.78 4450 2.0877
1.8981 1.8 4500 2.0900
1.9266 1.82 4550 2.0855
1.8589 1.84 4600 2.0795
1.7587 1.86 4650 2.0833
1.6735 1.88 4700 2.0886
1.7961 1.9 4750 2.0874
1.8099 1.92 4800 2.0801
1.8481 1.94 4850 2.0802
1.8418 1.96 4900 2.0774
1.8471 1.98 4950 2.0876
1.829 2.0 5000 2.0820
1.4073 2.02 5050 2.1485
1.4951 2.04 5100 2.1651
1.4291 2.06 5150 2.1522
1.3912 2.08 5200 2.1545
1.5581 2.1 5250 2.1462
1.5533 2.12 5300 2.1613
1.5436 2.14 5350 2.1562
1.4632 2.16 5400 2.1437
1.5859 2.18 5450 2.1563
1.4974 2.2 5500 2.1749
1.464 2.22 5550 2.1648
1.4689 2.24 5600 2.1623
1.565 2.26 5650 2.1656
1.5491 2.28 5700 2.1696
1.5382 2.3 5750 2.1659
1.4154 2.32 5800 2.1614
1.4636 2.34 5850 2.1570
1.4858 2.36 5900 2.1634
1.4295 2.38 5950 2.1897
1.6108 2.4 6000 2.1653
1.4283 2.42 6050 2.1633
1.4685 2.44 6100 2.1720
1.4443 2.46 6150 2.1618
1.4918 2.48 6200 2.1577
1.5742 2.5 6250 2.1665
1.49 2.52 6300 2.1697
1.552 2.54 6350 2.1489
1.5577 2.56 6400 2.1660
1.4348 2.58 6450 2.1766
1.5508 2.6 6500 2.1564
1.4666 2.62 6550 2.1644
1.4784 2.64 6600 2.1611
1.6065 2.66 6650 2.1770
1.559 2.68 6700 2.1635
1.5579 2.7 6750 2.1605
1.5103 2.72 6800 2.1735
1.5369 2.74 6850 2.1711
1.6012 2.76 6900 2.1650
1.5058 2.78 6950 2.1683
1.6553 2.8 7000 2.1613
1.5858 2.82 7050 2.1664
1.6428 2.84 7100 2.1566
1.4619 2.86 7150 2.1620
1.5989 2.88 7200 2.1571
1.6181 2.9 7250 2.1598
1.5831 2.92 7300 2.1560
1.555 2.94 7350 2.1529
1.5387 2.96 7400 2.1593
1.5477 2.98 7450 2.1608
1.4989 3.0 7500 2.1686

Framework versions

  • Transformers 4.36.2
  • Pytorch 1.13.1+cu117
  • Datasets 2.17.1
  • Tokenizers 0.15.2