cyllama3

This model is a fine-tuned version of meta-llama/Meta-Llama-3-8B on the khangmacon/llmtrain dataset. It achieves the following results on the evaluation set:

  • Loss: 1.9930
  • Accuracy: 0.5590

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: 3e-05
  • train_batch_size: 16
  • eval_batch_size: 16
  • seed: 42
  • distributed_type: multi-GPU
  • num_devices: 2
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 128
  • total_eval_batch_size: 32
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_ratio: 0.05
  • num_epochs: 1.0

Training results

Training Loss Epoch Step Validation Loss Accuracy
2.2432 0.01 500 2.1239 0.5358
2.209 0.02 1000 2.0922 0.5404
2.1988 0.03 1500 2.0742 0.5436
2.1877 0.04 2000 2.0615 0.5463
2.1743 0.05 2500 2.0514 0.5479
2.1885 0.06 3000 2.0427 0.5495
2.1883 0.07 3500 2.0355 0.5509
2.1954 0.08 4000 2.0298 0.5519
2.1597 0.09 4500 2.0254 0.5526
2.1763 0.1 5000 2.0222 0.5532
2.1413 0.11 5500 2.0195 0.5541
2.1812 0.12 6000 2.0169 0.5545
2.1526 0.14 6500 2.0148 0.5547
2.155 0.15 7000 2.0131 0.5554
2.1594 0.16 7500 2.0110 0.5558
2.1681 0.17 8000 2.0097 0.5559
2.1572 0.18 8500 2.0083 0.5562
2.0943 0.19 9000 2.0074 0.5566
2.1421 0.2 9500 2.0063 0.5566
2.1196 0.21 10000 2.0049 0.5568
2.1634 0.22 10500 2.0042 0.5568
2.1361 0.23 11000 2.0035 0.5573
2.1614 0.24 11500 2.0027 0.5572
2.1205 0.25 12000 2.0021 0.5576
2.0984 0.26 12500 2.0011 0.5576
2.1226 0.27 13000 2.0006 0.5575
2.1054 0.28 13500 2.0001 0.5577
2.1297 0.29 14000 1.9997 0.5578
2.1233 0.3 14500 1.9988 0.5581
2.1348 0.31 15000 1.9984 0.5581
2.1494 0.32 15500 1.9980 0.5582
2.0827 0.33 16000 1.9976 0.5584
2.0991 0.34 16500 1.9975 0.5582
2.1108 0.35 17000 1.9972 0.5582
2.1209 0.36 17500 1.9968 0.5583
2.1012 0.37 18000 1.9963 0.5584
2.1155 0.38 18500 1.9959 0.5585
2.1493 0.4 19000 1.9956 0.5585
2.1219 0.41 19500 1.9953 0.5587
2.1584 0.42 20000 1.9952 0.5588
2.1167 0.43 20500 1.9950 0.5587
2.1507 0.44 21000 1.9948 0.5586
2.1043 0.45 21500 1.9946 0.5587
2.0864 0.46 22000 1.9945 0.5587
2.1074 0.47 22500 1.9943 0.5587
2.0858 0.48 23000 1.9942 0.5590
2.1178 0.49 23500 1.9941 0.5588
2.1148 0.5 24000 1.9940 0.5588
2.1165 0.51 24500 1.9939 0.5588
2.1012 0.52 25000 1.9938 0.5590
2.1573 0.53 25500 1.9936 0.5590
2.1674 0.54 26000 1.9936 0.5589
2.1184 0.55 26500 1.9935 0.5590
2.1424 0.56 27000 1.9935 0.5590
2.1437 0.57 27500 1.9935 0.5590
2.1244 0.58 28000 1.9933 0.5591
2.0767 0.59 28500 1.9933 0.5589
2.1182 0.6 29000 1.9934 0.5591
2.1277 0.61 29500 1.9933 0.5591
2.1407 0.62 30000 1.9932 0.5591
2.1222 0.63 30500 1.9932 0.5591
2.1146 0.64 31000 1.9931 0.5591
2.1441 0.65 31500 1.9932 0.5591
2.1224 0.67 32000 1.9931 0.5590
2.0878 0.68 32500 1.9932 0.5591
2.1172 0.69 33000 1.9932 0.5590
2.1166 0.7 33500 1.9931 0.5592
2.1054 0.71 34000 1.9931 0.5591
2.0972 0.72 34500 1.9931 0.5590
2.1228 0.73 35000 1.9931 0.5590
2.1231 0.74 35500 1.9931 0.5592
2.0974 0.75 36000 1.9931 0.5590
2.1025 0.76 36500 1.9931 0.5591
2.1217 0.77 37000 1.9931 0.5590
2.1227 0.78 37500 1.9930 0.5591
2.1272 0.79 38000 1.9931 0.5592
2.117 0.8 38500 1.9931 0.5591
2.1325 0.81 39000 1.9931 0.5591
2.1046 0.82 39500 1.9930 0.5591
2.1096 0.83 40000 1.9930 0.5591
2.1149 0.84 40500 1.9931 0.5591
2.122 0.85 41000 1.9931 0.5591
2.1137 0.86 41500 1.9931 0.5591
2.0983 0.87 42000 1.9930 0.5590
2.1109 0.88 42500 1.9931 0.5591
2.172 0.89 43000 1.9930 0.5590
2.0882 0.9 43500 1.9930 0.5591
2.0646 0.91 44000 1.9930 0.5591
2.1223 0.93 44500 1.9930 0.5591
2.1342 0.94 45000 1.9930 0.5591
2.0991 0.95 45500 1.9930 0.5590
2.1431 0.96 46000 1.9930 0.5592
2.0965 0.97 46500 1.9931 0.5590
2.1377 0.98 47000 1.9931 0.5592
2.1118 0.99 47500 1.9931 0.5592
2.089 1.0 48000 1.9930 0.5590

Framework versions

  • PEFT 0.10.1.dev0
  • Transformers 4.39.3
  • Pytorch 2.2.0
  • Datasets 2.18.0
  • Tokenizers 0.15.2
Downloads last month
0
Inference API
Unable to determine this model’s pipeline type. Check the docs .

Model tree for khangmacon/cyllama3

Adapter
(529)
this model

Evaluation results