--- license: apache-2.0 base_model: TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T tags: - xcomet_xl_xxl - generated_from_trainer model-index: - name: cpo-xcomet-xl_xxl-inc7b-10p-shuff-1e-7-full-tiny results: [] --- # cpo-xcomet-xl_xxl-inc7b-10p-shuff-1e-7-full-tiny This model is a fine-tuned version of [TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T](https://huggingface.co/TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T) on the Unbabel/TowerAligned-v0.1 dataset. It achieves the following results on the evaluation set: - Loss: 2.7960 - Nll Loss: 1.0602 - Logps/best: -102.8461 - Rewards/chosen: -10.2846 - Rewards/rejected: -9.6988 - Rewards/accuracies: 0.4600 - Rewards/margins: -0.5858 - Logps/rejected: -96.9882 - Logps/chosen: -102.8461 - Logits/rejected: -1.8264 - Logits/chosen: -1.9635 ## 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: 1e-07 - train_batch_size: 1 - eval_batch_size: 4 - seed: 42 - gradient_accumulation_steps: 16 - total_train_batch_size: 16 - optimizer: Adam with betas=(0.9,0.95) and epsilon=1e-08 - lr_scheduler_type: linear - lr_scheduler_warmup_ratio: 0.1 - num_epochs: 3 ### Training results | Training Loss | Epoch | Step | Validation Loss | Nll Loss | Logps/best | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen | |:-------------:|:------:|:----:|:---------------:|:--------:|:----------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:| | 2.9687 | 0.1063 | 100 | 2.8066 | 1.0659 | -103.3585 | -10.3358 | -9.7431 | 0.4540 | -0.5928 | -97.4308 | -103.3585 | -1.8274 | -1.9646 | | 3.0173 | 0.2127 | 200 | 2.8063 | 1.0656 | -103.3396 | -10.3340 | -9.7402 | 0.4560 | -0.5937 | -97.4022 | -103.3396 | -1.8275 | -1.9648 | | 2.8267 | 0.3190 | 300 | 2.8046 | 1.0650 | -103.2849 | -10.3285 | -9.7373 | 0.4540 | -0.5912 | -97.3725 | -103.2849 | -1.8273 | -1.9644 | | 2.9404 | 0.4254 | 400 | 2.8046 | 1.0644 | -103.2318 | -10.3232 | -9.7301 | 0.4600 | -0.5931 | -97.3013 | -103.2318 | -1.8271 | -1.9643 | | 3.3065 | 0.5317 | 500 | 2.8002 | 1.0637 | -103.1556 | -10.3156 | -9.7280 | 0.4600 | -0.5875 | -97.2803 | -103.1556 | -1.8268 | -1.9640 | | 2.9333 | 0.6381 | 600 | 2.8021 | 1.0633 | -103.1282 | -10.3128 | -9.7212 | 0.4560 | -0.5916 | -97.2122 | -103.1282 | -1.8271 | -1.9642 | | 3.2698 | 0.7444 | 700 | 2.8006 | 1.0627 | -103.0742 | -10.3074 | -9.7178 | 0.4580 | -0.5897 | -97.1777 | -103.0742 | -1.8268 | -1.9640 | | 2.7002 | 0.8508 | 800 | 2.8003 | 1.0624 | -103.0458 | -10.3046 | -9.7147 | 0.4580 | -0.5899 | -97.1470 | -103.0458 | -1.8269 | -1.9641 | | 3.0848 | 0.9571 | 900 | 2.7984 | 1.0620 | -103.0023 | -10.3002 | -9.7132 | 0.4580 | -0.5870 | -97.1324 | -103.0023 | -1.8267 | -1.9638 | | 2.9243 | 1.0635 | 1000 | 2.7987 | 1.0617 | -102.9805 | -10.2980 | -9.7086 | 0.4580 | -0.5895 | -97.0859 | -102.9805 | -1.8268 | -1.9639 | | 2.7945 | 1.1698 | 1100 | 2.7974 | 1.0615 | -102.9564 | -10.2956 | -9.7084 | 0.4580 | -0.5872 | -97.0842 | -102.9564 | -1.8267 | -1.9638 | | 2.7893 | 1.2762 | 1200 | 2.7979 | 1.0613 | -102.9413 | -10.2941 | -9.7061 | 0.4620 | -0.5880 | -97.0609 | -102.9413 | -1.8266 | -1.9638 | | 3.2162 | 1.3825 | 1300 | 2.7978 | 1.0611 | -102.9208 | -10.2921 | -9.7039 | 0.4540 | -0.5882 | -97.0387 | -102.9208 | -1.8266 | -1.9637 | | 2.8123 | 1.4889 | 1400 | 2.7980 | 1.0611 | -102.9247 | -10.2925 | -9.7032 | 0.4580 | -0.5893 | -97.0320 | -102.9247 | -1.8266 | -1.9637 | | 2.785 | 1.5952 | 1500 | 2.7973 | 1.0606 | -102.8798 | -10.2880 | -9.6993 | 0.4560 | -0.5887 | -96.9928 | -102.8798 | -1.8265 | -1.9636 | | 2.7997 | 1.7016 | 1600 | 2.7952 | 1.0606 | -102.8751 | -10.2875 | -9.7026 | 0.4600 | -0.5849 | -97.0257 | -102.8751 | -1.8267 | -1.9638 | | 2.6655 | 1.8079 | 1700 | 2.7956 | 1.0605 | -102.8628 | -10.2863 | -9.7005 | 0.4620 | -0.5858 | -97.0050 | -102.8628 | -1.8264 | -1.9635 | | 2.7597 | 1.9143 | 1800 | 2.7966 | 1.0605 | -102.8715 | -10.2871 | -9.6999 | 0.4540 | -0.5872 | -96.9991 | -102.8715 | -1.8267 | -1.9637 | | 2.9736 | 2.0206 | 1900 | 2.7955 | 1.0603 | -102.8511 | -10.2851 | -9.6990 | 0.4600 | -0.5861 | -96.9900 | -102.8511 | -1.8266 | -1.9637 | | 2.8977 | 2.1270 | 2000 | 2.7954 | 1.0603 | -102.8550 | -10.2855 | -9.6990 | 0.4560 | -0.5865 | -96.9901 | -102.8550 | -1.8270 | -1.9641 | | 2.7043 | 2.2333 | 2100 | 2.7961 | 1.0604 | -102.8632 | -10.2863 | -9.6997 | 0.4560 | -0.5867 | -96.9967 | -102.8632 | -1.8264 | -1.9635 | | 2.7693 | 2.3396 | 2200 | 2.7951 | 1.0604 | -102.8550 | -10.2855 | -9.6998 | 0.4600 | -0.5857 | -96.9983 | -102.8550 | -1.8263 | -1.9634 | | 2.6632 | 2.4460 | 2300 | 2.7943 | 1.0602 | -102.8407 | -10.2841 | -9.6989 | 0.4600 | -0.5851 | -96.9893 | -102.8407 | -1.8264 | -1.9635 | | 3.2451 | 2.5523 | 2400 | 2.7953 | 1.0602 | -102.8434 | -10.2843 | -9.6989 | 0.4580 | -0.5855 | -96.9885 | -102.8434 | -1.8264 | -1.9635 | | 2.7117 | 2.6587 | 2500 | 2.7955 | 1.0601 | -102.8357 | -10.2836 | -9.6962 | 0.4580 | -0.5873 | -96.9625 | -102.8357 | -1.8263 | -1.9634 | | 3.148 | 2.7650 | 2600 | 2.7967 | 1.0604 | -102.8636 | -10.2864 | -9.6985 | 0.4560 | -0.5878 | -96.9853 | -102.8636 | -1.8265 | -1.9636 | | 3.2951 | 2.8714 | 2700 | 2.7959 | 1.0602 | -102.8490 | -10.2849 | -9.6981 | 0.4620 | -0.5868 | -96.9812 | -102.8490 | -1.8263 | -1.9634 | | 2.8486 | 2.9777 | 2800 | 2.7960 | 1.0602 | -102.8461 | -10.2846 | -9.6988 | 0.4600 | -0.5858 | -96.9882 | -102.8461 | -1.8264 | -1.9635 | ### Framework versions - Transformers 4.41.2 - Pytorch 2.1.2 - Datasets 2.20.0 - Tokenizers 0.19.1