OpenELM-1_1B-KTO

This model was trained from scratch on the None dataset. It achieves the following results on the evaluation set:

  • Kl: 0.0
  • Logits/chosen: -1428926080.0
  • Logits/rejected: -1154165888.0
  • Logps/chosen: -548.5443
  • Logps/rejected: -840.7200
  • Loss: 0.4491
  • Rewards/chosen: -2.1686
  • Rewards/margins: 3.2859
  • Rewards/rejected: -5.4545

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 Kl Logits/chosen Logits/rejected Logps/chosen Logps/rejected Validation Loss Rewards/chosen Rewards/margins Rewards/rejected
0.4859 0.0523 100 0.0 -3175687424.0 -3128284928.0 -393.0411 -370.7767 0.4855 -0.6136 0.1415 -0.7551
0.4556 0.1047 200 0.0 -3518155520.0 -3502177280.0 -452.2593 -491.0910 0.4658 -1.2023 0.7579 -1.9602
0.4658 0.1570 300 0.0 -3454242304.0 -3428943360.0 -416.2217 -439.7414 0.4630 -0.8430 0.6032 -1.4462
0.4543 0.2094 400 0.0 -3808428032.0 -3728536576.0 -562.5413 -589.5574 0.4728 -2.3096 0.6357 -2.9453
0.4445 0.2617 500 0.0 -3128418048.0 -2977955840.0 -534.3256 -585.3565 0.4657 -2.0271 0.8745 -2.9015
0.4654 0.3141 600 0.0 -3596449280.0 -3444389120.0 -529.1321 -582.2695 0.4658 -1.9723 0.8963 -2.8686
0.4517 0.3664 700 0.0 -3631068928.0 -3583666688.0 -452.7501 -492.3958 0.4690 -1.2099 0.7624 -1.9723
0.4701 0.4187 800 0.0 -3269426688.0 -3207910144.0 -550.3810 -656.9627 0.4595 -2.1877 1.4299 -3.6176
0.4711 0.4711 900 0.0 -3022695168.0 -3008713984.0 -592.4354 -758.0388 0.4626 -2.6066 2.0224 -4.6290
0.4534 0.5234 1000 0.0 -2621240320.0 -2359728640.0 -548.9560 -656.3580 0.4594 -2.1704 1.4414 -3.6118
0.4428 0.5758 1100 0.0 -2962243840.0 -2838944512.0 -587.9386 -759.0254 0.4583 -2.5622 2.0794 -4.6416
0.4619 0.6281 1200 0.0 -2887944704.0 -2875961088.0 -570.1098 -742.3173 0.4549 -2.3843 2.0891 -4.4734
0.4627 0.6805 1300 0.0 -2942404096.0 -2700332800.0 -705.3934 -1037.3665 0.4565 -3.7379 3.6828 -7.4207
0.4622 0.7328 1400 0.0 -3010711296.0 -2812979968.0 -783.2637 -1089.1775 0.4620 -4.5181 3.4185 -7.9366
0.4571 0.7851 1500 0.0 -2759985152.0 -2638283776.0 -567.8931 -773.6330 0.4572 -2.3622 2.4227 -4.7849
0.4714 0.8375 1600 0.0 -2749732352.0 -2697137152.0 -526.0604 -662.9458 0.4618 -1.9416 1.7369 -3.6785
0.4266 0.8898 1700 0.0 -1904080896.0 -1675191680.0 -554.2761 -683.2819 0.4548 -2.2239 1.6573 -3.8812
0.449 0.9422 1800 0.0 -2063198080.0 -1781980032.0 -630.7531 -780.4754 0.4612 -2.9899 1.8639 -4.8538
0.4773 0.9945 1900 0.0 -2405932544.0 -2133635840.0 -658.1771 -792.9826 0.4612 -3.2661 1.7145 -4.9806
0.4654 1.0468 2000 0.0 -2678495744.0 -2597272832.0 -528.9421 -616.7041 0.4568 -1.9690 1.2455 -3.2144
0.4228 1.0992 2100 0.0 -1897423232.0 -1554289280.0 -605.4824 -827.4172 0.4540 -2.7379 2.5883 -5.3262
0.4094 1.1515 2200 0.0 -1938966784.0 -1658281344.0 -495.0579 -585.3565 0.4561 -1.6318 1.2711 -2.9029
0.3779 1.2039 2300 0.0 -2399674624.0 -2125380352.0 -514.6284 -627.5883 0.4564 -1.8295 1.4960 -3.3256
0.3731 1.2562 2400 0.0 -1886105216.0 -1661210752.0 -496.7679 -605.2153 0.4578 -1.6493 1.4530 -3.1023
0.3538 1.3086 2500 0.0 -1523131520.0 -1173539584.0 -520.2969 -641.4639 0.4578 -1.8841 1.5789 -3.4630
0.3699 1.3609 2600 0.0 -2620041984.0 -2228706560.0 -464.2454 -571.5127 0.4552 -1.3250 1.4383 -2.7633
0.3293 1.4132 2700 0.0 -1547698176.0 -1145178112.0 -559.4696 -805.1397 0.4508 -2.2760 2.8229 -5.0990
0.3376 1.4656 2800 0.0 -1543770240.0 -1201035648.0 -495.7229 -643.6280 0.4576 -1.6388 1.8453 -3.4841
0.3545 1.5179 2900 0.0 -1516407296.0 -1264482816.0 -736.2058 -1167.2123 0.4523 -4.0433 4.6733 -8.7166
0.3399 1.5703 3000 0.0 -2308331776.0 -2173847808.0 -634.2682 -983.0731 0.4497 -3.0224 3.8548 -6.8772
0.3429 1.6226 3100 0.0 -2065728000.0 -1775322368.0 -641.7100 -945.9015 0.4497 -3.1008 3.4052 -6.5060
0.3005 1.6750 3200 0.0 -2172250112.0 -2024051328.0 -515.8318 -719.3396 0.4492 -1.8422 2.3998 -4.2420
0.3468 1.7273 3300 0.0 -2299277568.0 -2052279552.0 -650.1019 -1102.9896 0.4503 -3.1821 4.8952 -8.0773
0.3361 1.7796 3400 0.0 -1953080960.0 -1726854912.0 -586.8936 -953.3486 0.4488 -2.5504 4.0314 -6.5818
0.3405 1.8320 3500 0.0 -1600359936.0 -1368940928.0 -595.8555 -917.2273 0.4473 -2.6417 3.5806 -6.2223
0.362 1.8843 3600 0.0 -1622263552.0 -1444837888.0 -594.8738 -933.2352 0.4472 -2.6312 3.7478 -6.3789
0.3153 1.9367 3700 0.0 -1449431680.0 -1236653952.0 -548.1009 -828.2765 0.4482 -2.1611 3.1750 -5.3362
0.332 1.9890 3800 0.0 -1527192704.0 -1300766848.0 -517.2568 -750.3690 0.4465 -1.8534 2.6999 -4.5532
0.3281 2.0414 3900 0.0 -1679585792.0 -1427328256.0 -596.9639 -954.6852 0.4463 -2.6517 3.9427 -6.5944
0.3181 2.0937 4000 0.0 -1560081408.0 -1263351040.0 -525.1103 -727.0094 0.4486 -1.9321 2.3847 -4.3169
0.2603 2.1460 4100 0.0 -1459418112.0 -1138054528.0 -524.6037 -721.5356 0.4512 -1.9273 2.3376 -4.2649
0.2388 2.1984 4200 0.0 -1377462656.0 -1113288064.0 -484.2593 -613.2352 0.4556 -1.5242 1.6601 -3.1844
0.224 2.2507 4300 0.0 -1391976320.0 -1133061248.0 -479.9842 -590.8304 0.4576 -1.4813 1.4770 -2.9584
0.26 2.3031 4400 0.0 -1215682432.0 -961427712.0 -496.7046 -633.5714 0.4562 -1.6511 1.7328 -3.3839
0.2234 2.3554 4500 0.0 -1145577600.0 -884432256.0 -540.8808 -735.6340 0.4557 -2.0918 2.3138 -4.4056
0.235 2.4077 4600 0.0 -1404625792.0 -1132328960.0 -488.9302 -651.3615 0.4559 -1.5702 1.9911 -3.5613
0.2246 2.4601 4700 0.0 -1378527872.0 -1094446976.0 -506.0782 -691.9065 0.4523 -1.7423 2.2245 -3.9668
0.24 2.5124 4800 0.0 -1400231808.0 -1111490560.0 -509.5299 -701.2312 0.4519 -1.7776 2.2838 -4.0614
0.2557 2.5648 4900 0.0 -1397235840.0 -1123008256.0 -528.5304 -749.2869 0.4514 -1.9623 2.5779 -4.5403
0.2413 2.6171 5000 0.0 -1477659904.0 -1194644352.0 -527.0421 -771.0870 0.4506 -1.9521 2.8087 -4.7608
0.2273 2.6695 5100 0.0 -1420737280.0 -1144246016.0 -524.3503 -753.7106 0.4495 -1.9271 2.6609 -4.5880
0.2645 2.7218 5200 0.0 -1426662528.0 -1147708032.0 -529.0688 -767.9682 0.4501 -1.9730 2.7549 -4.7279
0.2637 2.7741 5300 0.0 -1414479104.0 -1139319424.0 -551.0460 -835.7553 0.4497 -2.1932 3.2146 -5.4078
0.2683 2.8265 5400 0.0 -1420271232.0 -1144445824.0 -552.9777 -847.9125 0.4496 -2.2117 3.3187 -5.5303
0.2551 2.8788 5500 0.0 -1425663872.0 -1148307200.0 -551.4577 -846.3531 0.4494 -2.1969 3.3124 -5.5093
0.2695 2.9312 5600 0.0 -1428859520.0 -1154099328.0 -548.8293 -841.0065 0.4490 -2.1721 3.2870 -5.4591
0.2664 2.9835 5700 0.0 -1428926080.0 -1154165888.0 -548.5443 -840.7200 0.4491 -2.1686 3.2859 -5.4545

Framework versions

  • Transformers 4.44.2
  • Pytorch 2.3.0
  • Datasets 3.0.0
  • Tokenizers 0.19.1
Downloads last month
10
Safetensors
Model size
1.08B params
Tensor type
BF16
·
Inference Examples
Inference API (serverless) does not yet support model repos that contain custom code.