OpenELM-1_1B-DPO-full-max-random-reward

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

  • Loss: 194.2460
  • Rewards/chosen: -660.0
  • Rewards/rejected: -568.0
  • Rewards/accuracies: 0.4277
  • Rewards/margins: -89.5
  • Logps/rejected: -57344.0
  • Logps/chosen: -66560.0
  • Logits/rejected: 7.5
  • Logits/chosen: 7.0

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: 5e-05
  • train_batch_size: 8
  • eval_batch_size: 16
  • seed: 42
  • distributed_type: multi-GPU
  • num_devices: 4
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 64
  • total_eval_batch_size: 64
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_ratio: 0.1
  • num_epochs: 3

Training results

Training Loss Epoch Step Validation Loss Rewards/chosen Rewards/rejected Rewards/accuracies Rewards/margins Logps/rejected Logps/chosen Logits/rejected Logits/chosen
0.6914 0.1047 100 0.6983 -0.3262 -0.3223 0.4004 -0.0052 -320.0 -352.0 -9.4375 -9.8125
0.6914 0.2094 200 58.1418 -201.0 -173.0 0.4453 -28.0 -17664.0 -20480.0 0.2969 0.2178
0.6914 0.3141 300 97.2609 -330.0 -284.0 0.4258 -44.75 -28800.0 -33280.0 -0.3262 -0.3027
0.6914 0.4188 400 102.8539 -348.0 -300.0 0.4297 -47.75 -30464.0 -35328.0 -0.7656 -0.7461
0.6914 0.5236 500 108.8187 -368.0 -318.0 0.4277 -50.25 -32128.0 -37120.0 -0.2490 -0.2812
0.6914 0.6283 600 114.7604 -388.0 -336.0 0.4355 -53.0 -33792.0 -39168.0 1.25 1.1016
0.6914 0.7330 700 120.9475 -410.0 -354.0 0.4277 -55.75 -35584.0 -41216.0 2.3906 2.1875
0.6914 0.8377 800 127.3012 -432.0 -372.0 0.4336 -58.75 -37632.0 -43520.0 4.2812 3.9062
0.6914 0.9424 900 133.8314 -454.0 -392.0 0.4297 -62.0 -39424.0 -45824.0 4.4375 4.0938
0.6914 1.0471 1000 140.0195 -476.0 -410.0 0.4355 -64.5 -41472.0 -47872.0 5.875 5.4062
0.6914 1.1518 1100 146.3645 -496.0 -430.0 0.4316 -67.5 -43264.0 -49920.0 5.7812 5.375
0.6914 1.2565 1200 151.9910 -516.0 -446.0 0.4336 -70.0 -44800.0 -51968.0 6.375 5.9375
0.6914 1.3613 1300 157.8106 -536.0 -462.0 0.4297 -73.0 -46592.0 -54016.0 7.0 6.5
0.6914 1.4660 1400 163.0493 -552.0 -478.0 0.4316 -75.5 -48128.0 -55552.0 7.3438 6.8125
0.6914 1.5707 1500 168.1114 -572.0 -494.0 0.4277 -77.5 -49664.0 -57344.0 7.2812 6.75
0.6914 1.6754 1600 172.7765 -588.0 -506.0 0.4316 -80.0 -50944.0 -58880.0 7.0625 6.5938
0.6914 1.7801 1700 176.9677 -600.0 -520.0 0.4395 -81.5 -52224.0 -60416.0 7.4688 6.9375
0.6914 1.8848 1800 180.6313 -612.0 -532.0 0.4355 -83.0 -53248.0 -61696.0 7.7812 7.25
0.6914 1.9895 1900 183.7843 -624.0 -540.0 0.4258 -84.5 -54272.0 -62720.0 7.625 7.125
0.6914 2.0942 2000 186.4619 -632.0 -548.0 0.4277 -86.0 -55040.0 -63744.0 7.6562 7.125
0.6914 2.1990 2100 188.7695 -640.0 -552.0 0.4258 -87.0 -55808.0 -64512.0 7.5938 7.125
0.6914 2.3037 2200 190.4722 -648.0 -560.0 0.4355 -87.5 -56320.0 -65024.0 7.5625 7.0625
0.6914 2.4084 2300 191.8555 -652.0 -564.0 0.4258 -88.5 -56576.0 -65536.0 7.5 7.0312
0.6914 2.5131 2400 192.9321 -656.0 -564.0 0.4258 -89.0 -56832.0 -66048.0 7.4375 6.9688
0.6914 2.6178 2500 193.6570 -656.0 -568.0 0.4258 -89.0 -57088.0 -66048.0 7.4688 7.0
0.6914 2.7225 2600 193.9604 -660.0 -568.0 0.4238 -89.5 -57344.0 -66048.0 7.5312 7.0625
0.6914 2.8272 2700 194.1360 -660.0 -568.0 0.4258 -89.5 -57344.0 -66048.0 7.5 7.0312
0.6914 2.9319 2800 194.2460 -660.0 -568.0 0.4277 -89.5 -57344.0 -66560.0 7.5 7.0

Framework versions

  • Transformers 4.44.2
  • Pytorch 2.3.0
  • Datasets 2.21.0
  • Tokenizers 0.19.1
Downloads last month
9
Safetensors
Model size
1.08B params
Tensor type
BF16
·
Inference Examples
Inference API (serverless) does not yet support model repos that contain custom code.