genv3pair1NoGT_1.5B_cdpo_ebs32_lr5e-06_beta0.1_epoch8.0_42

This model is a fine-tuned version of Qwen/Qwen2.5-1.5B-Instruct on the YuchenLi01/MATH_Qwen2.5-1.5BInstruct_DPO_MoreUniqueResponseNoGTv3pair1 dataset. It achieves the following results on the evaluation set:

  • Loss: 1.2478
  • Rewards/chosen: -0.2745
  • Rewards/rejected: 0.0
  • Rewards/accuracies: 0.4250
  • Rewards/margins: -0.2745
  • Logps/rejected: -53.9242
  • Logps/chosen: -32.8150
  • Logits/rejected: -3.5353
  • Logits/chosen: -3.4027

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-06
  • train_batch_size: 4
  • eval_batch_size: 4
  • seed: 42
  • distributed_type: multi-GPU
  • num_devices: 8
  • total_train_batch_size: 32
  • 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: 0.1
  • num_epochs: 8.0

Training results

Training Loss Epoch Step Validation Loss Rewards/chosen Rewards/rejected Rewards/accuracies Rewards/margins Logps/rejected Logps/chosen Logits/rejected Logits/chosen
0.6339 0.1117 20 0.6293 0.1239 0.0 1.0 0.1239 -40.1835 -28.8307 -2.3134 -2.4461
0.4231 0.2235 40 0.4187 0.7203 0.0 0.9500 0.7203 -33.8767 -22.8669 -3.0016 -3.0384
0.3949 0.3352 60 0.3773 0.9640 0.0 0.9750 0.9640 -32.2958 -20.4299 -3.3011 -3.2774
0.3546 0.4469 80 0.3637 0.9914 0.0 1.0 0.9914 -31.6176 -20.1563 -3.3013 -3.2770
0.3059 0.5587 100 0.3664 1.0027 0.0 0.9750 1.0027 -31.7914 -20.0426 -3.3690 -3.3327
0.3504 0.6704 120 0.3799 0.9659 0.0 1.0 0.9659 -32.6139 -20.4109 -3.3379 -3.3099
0.4161 0.7821 140 0.4016 0.8722 0.0 0.9750 0.8722 -33.4282 -21.3484 -3.2793 -3.2535
0.4314 0.8939 160 0.4172 0.8576 0.0 0.9750 0.8576 -33.2324 -21.4936 -3.2940 -3.2712
0.2584 1.0056 180 0.4375 0.7763 0.0 0.9250 0.7763 -33.9398 -22.3070 -3.2443 -3.2321
0.2477 1.1173 200 0.5238 0.7154 0.0 0.8000 0.7154 -36.2846 -22.9161 -3.4659 -3.4084
0.3476 1.2291 220 0.4947 0.7315 0.0 0.875 0.7315 -36.0092 -22.7547 -3.4213 -3.3765
0.2478 1.3408 240 0.5380 0.7329 0.0 0.875 0.7329 -37.9193 -22.7412 -3.4929 -3.4322
0.3341 1.4525 260 0.5285 0.7788 0.0 0.9000 0.7788 -36.5301 -22.2823 -3.4798 -3.4204
0.2892 1.5642 280 0.5287 0.7061 0.0 0.9250 0.7061 -36.8482 -23.0091 -3.3743 -3.3223
0.2768 1.6760 300 0.5382 0.6573 0.0 0.8500 0.6573 -37.3328 -23.4967 -3.4178 -3.3609
0.3726 1.7877 320 0.5422 0.6608 0.0 0.8500 0.6608 -35.9154 -23.4616 -3.3283 -3.3046
0.3073 1.8994 340 0.5740 0.6415 0.0 0.8250 0.6415 -36.9027 -23.6551 -3.4064 -3.3511
0.1934 2.0112 360 0.5502 0.7417 0.0 0.9000 0.7417 -36.7407 -22.6530 -3.3965 -3.3432
0.2036 2.1229 380 0.6724 0.6468 0.0 0.8500 0.6468 -39.6545 -23.6024 -3.5622 -3.4664
0.1782 2.2346 400 0.6572 0.5434 0.0 0.8500 0.5434 -38.8220 -24.6361 -3.5365 -3.4469
0.2196 2.3464 420 0.6808 0.4513 0.0 0.75 0.4513 -40.6678 -25.5571 -3.5865 -3.4942
0.1398 2.4581 440 0.7209 0.3850 0.0 0.75 0.3850 -39.6323 -26.2196 -3.5268 -3.4407
0.1952 2.5698 460 0.7184 0.4272 0.0 0.75 0.4272 -39.7520 -25.7979 -3.5159 -3.4299
0.3116 2.6816 480 0.6880 0.5085 0.0 0.7750 0.5085 -39.4111 -24.9850 -3.5145 -3.4305
0.1611 2.7933 500 0.7117 0.4137 0.0 0.75 0.4137 -41.1864 -25.9328 -3.4832 -3.4002
0.1948 2.9050 520 0.6730 0.5021 0.0 0.875 0.5021 -39.4969 -25.0492 -3.4868 -3.4069
0.1743 3.0168 540 0.6875 0.3881 0.0 0.8000 0.3881 -39.2651 -26.1893 -3.5315 -3.4441
0.2073 3.1285 560 0.8458 0.1726 0.0 0.625 0.1726 -42.8931 -28.3435 -3.5456 -3.4402
0.1496 3.2402 580 0.7679 0.2554 0.0 0.6750 0.2554 -41.5129 -27.5165 -3.4945 -3.4113
0.1503 3.3520 600 0.7629 0.2099 0.0 0.6000 0.2099 -42.3083 -27.9710 -3.5254 -3.4360
0.1578 3.4637 620 0.7733 0.1533 0.0 0.6000 0.1533 -41.1396 -28.5373 -3.5135 -3.4252
0.1335 3.5754 640 0.8319 0.1997 0.0 0.6000 0.1997 -41.4449 -28.0729 -3.5372 -3.4387
0.1696 3.6872 660 0.8017 0.2201 0.0 0.6500 0.2201 -42.0740 -27.8688 -3.5692 -3.4636
0.2641 3.7989 680 0.8066 0.2394 0.0 0.7250 0.2394 -42.6891 -27.6759 -3.5624 -3.4601
0.1268 3.9106 700 0.7793 0.3316 0.0 0.75 0.3316 -42.6197 -26.7540 -3.5242 -3.4228
0.1236 4.0223 720 0.7696 0.3849 0.0 0.8000 0.3849 -42.0404 -26.2206 -3.5181 -3.4178
0.1061 4.1341 740 0.9666 0.1498 0.0 0.6750 0.1498 -46.4133 -28.5724 -3.5567 -3.4378
0.1186 4.2458 760 0.9323 0.1752 0.0 0.6000 0.1752 -44.5664 -28.3183 -3.6046 -3.4827
0.1112 4.3575 780 0.9042 0.1862 0.0 0.7000 0.1862 -44.8347 -28.2085 -3.5715 -3.4571
0.1463 4.4693 800 0.8225 0.2410 0.0 0.6750 0.2410 -42.9694 -27.6601 -3.5655 -3.4553
0.1564 4.5810 820 0.8811 0.1677 0.0 0.625 0.1677 -44.3126 -28.3932 -3.5381 -3.4259
0.1985 4.6927 840 0.9132 0.1664 0.0 0.6000 0.1664 -45.6402 -28.4064 -3.5504 -3.4339
0.1374 4.8045 860 0.8452 0.1916 0.0 0.6000 0.1916 -44.4828 -28.1538 -3.5435 -3.4322
0.1626 4.9162 880 0.8745 0.1316 0.0 0.6000 0.1316 -44.5274 -28.7537 -3.5277 -3.4138
0.1003 5.0279 900 0.9217 0.0483 0.0 0.5 0.0483 -46.2505 -29.5872 -3.5361 -3.4139
0.0927 5.1397 920 1.0600 -0.0258 0.0 0.4750 -0.0258 -48.8408 -30.3276 -3.5497 -3.4235
0.1022 5.2514 940 0.9659 0.0427 0.0 0.5500 0.0427 -47.0927 -29.6433 -3.5460 -3.4240
0.109 5.3631 960 1.0517 -0.1151 0.0 0.5 -0.1151 -49.9369 -31.2207 -3.5436 -3.4193
0.1338 5.4749 980 1.0318 -0.0630 0.0 0.5250 -0.0630 -49.2989 -30.6998 -3.5513 -3.4235
0.1032 5.5866 1000 1.0205 -0.0941 0.0 0.5500 -0.0941 -48.3879 -31.0113 -3.5495 -3.4234
0.0994 5.6983 1020 1.0377 -0.0663 0.0 0.5500 -0.0663 -49.2657 -30.7334 -3.5589 -3.4317
0.1406 5.8101 1040 1.0168 -0.0241 0.0 0.5500 -0.0241 -49.0492 -30.3108 -3.5694 -3.4398
0.1197 5.9218 1060 0.9964 -0.0000 0.0 0.5750 -0.0000 -48.1696 -30.0701 -3.5505 -3.4231
0.0783 6.0335 1080 1.0153 -0.0575 0.0 0.5250 -0.0575 -48.8631 -30.6448 -3.5704 -3.4415
0.0717 6.1453 1100 1.1374 -0.1786 0.0 0.4750 -0.1786 -51.5463 -31.8565 -3.5625 -3.4296
0.101 6.2570 1120 1.1705 -0.2142 0.0 0.4500 -0.2142 -52.2915 -32.2120 -3.5480 -3.4154
0.117 6.3687 1140 1.1203 -0.1841 0.0 0.4250 -0.1841 -51.6086 -31.9114 -3.5513 -3.4207
0.095 6.4804 1160 1.1487 -0.2081 0.0 0.4000 -0.2081 -51.3992 -32.1507 -3.5420 -3.4088
0.0921 6.5922 1180 1.1640 -0.2049 0.0 0.4750 -0.2049 -51.9258 -32.1190 -3.5373 -3.4033
0.0818 6.7039 1200 1.1760 -0.1961 0.0 0.5 -0.1961 -52.2654 -32.0309 -3.5403 -3.4070
0.0854 6.8156 1220 1.1823 -0.2121 0.0 0.4500 -0.2121 -52.4322 -32.1909 -3.5436 -3.4127
0.1399 6.9274 1240 1.1804 -0.2021 0.0 0.4500 -0.2021 -52.6683 -32.0915 -3.5352 -3.4029
0.0886 7.0391 1260 1.1777 -0.1845 0.0 0.5 -0.1845 -52.7352 -31.9150 -3.5362 -3.4040
0.1105 7.1508 1280 1.2006 -0.2368 0.0 0.4500 -0.2368 -52.9880 -32.4379 -3.5347 -3.4014
0.0773 7.2626 1300 1.2167 -0.2458 0.0 0.4500 -0.2458 -53.3666 -32.5277 -3.5404 -3.4086
0.0836 7.3743 1320 1.2340 -0.2461 0.0 0.4500 -0.2461 -53.7346 -32.5307 -3.5380 -3.4057
0.1214 7.4860 1340 1.2435 -0.2655 0.0 0.4500 -0.2655 -54.0182 -32.7252 -3.5406 -3.4093
0.115 7.5978 1360 1.2474 -0.2650 0.0 0.4250 -0.2650 -54.0682 -32.7199 -3.5435 -3.4125
0.0801 7.7095 1380 1.2451 -0.2708 0.0 0.4500 -0.2708 -54.1074 -32.7779 -3.5365 -3.4037
0.1084 7.8212 1400 1.2457 -0.2575 0.0 0.4500 -0.2575 -53.9336 -32.6446 -3.5311 -3.3971
0.1042 7.9330 1420 1.2461 -0.2697 0.0 0.4500 -0.2697 -54.1756 -32.7669 -3.5368 -3.4044

Framework versions

  • Transformers 4.45.2
  • Pytorch 2.5.1+cu121
  • Datasets 3.5.0
  • Tokenizers 0.20.3
Downloads last month
6
Safetensors
Model size
2B params
Tensor type
BF16
ยท
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for YuchenLi01/genv3pair1NoGT_1.5B_cdpo_ebs32_lr5e-06_beta0.1_epoch8.0_42

Base model

Qwen/Qwen2.5-1.5B
Finetuned
(1319)
this model