--- license: other base_model: deepseek-ai/deepseek-llm-7b-chat tags: - alignment-handbook - trl - dpo - generated_from_trainer - trl - dpo - generated_from_trainer datasets: - self-generate/ds_chat_original_cn_mining_oj_iter0-binarized - self-generate/ds_chat_original_cn_mining_sandbox_iter0-binarized - self-generate/ds_chat_original_cn_rl_oj_iter0-binarized model-index: - name: ds_chat_sppo_hard_cosine_iter0_2024-09-17-09.48 results: [] --- [Visualize in Weights & Biases](https://ml.byteintl.net/experiment/tracking/detail?Id=project_20240915_20321b8f&selectedTrial=run_20240918_9c9733ce) # ds_chat_sppo_hard_cosine_iter0_2024-09-17-09.48 This model is a fine-tuned version of [deepseek-ai/deepseek-llm-7b-chat](https://huggingface.co/deepseek-ai/deepseek-llm-7b-chat) on the self-generate/ds_chat_original_cn_mining_oj_iter0-binarized, the self-generate/ds_chat_original_cn_mining_sandbox_iter0-binarized and the self-generate/ds_chat_original_cn_rl_oj_iter0-binarized datasets. It achieves the following results on the evaluation set: - Loss: 4613.5840 - Rewards/chosen: 0.0069 - Rewards/rejected: -0.0357 - Rewards/accuracies: 0.6053 - Rewards/margins: 0.0425 - Logps/rejected: -263.2242 - Logps/chosen: -252.2310 - Logits/rejected: 1.4371 - Logits/chosen: 1.3941 - Debug/policy Chosen Logits: 1.3941 - Debug/policy Rejected Logits: 1.4371 - Debug/policy Chosen Logps: -252.2310 - Debug/policy Rejected Logps: -263.2242 - Debug/reference Chosen Logps: -252.9185 - Debug/reference Rejected Logps: -259.6586 - Debug/sppo Chosen Reward In Loss: 0.6874 - Debug/sppo Rej Reward In Loss: -3.5656 - Debug/sppo Chosen Loss: 2507.3259 - Debug/sppo Reject Loss: 2312.9116 ## 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: 8 - eval_batch_size: 4 - seed: 42 - distributed_type: multi-GPU - num_devices: 8 - total_train_batch_size: 64 - 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 - lr_scheduler_warmup_steps: 100 - 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 | Debug/policy Chosen Logits | Debug/policy Rejected Logits | Debug/policy Chosen Logps | Debug/policy Rejected Logps | Debug/reference Chosen Logps | Debug/reference Rejected Logps | Debug/sppo Chosen Reward In Loss | Debug/sppo Rej Reward In Loss | Debug/sppo Chosen Loss | Debug/sppo Reject Loss | |:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|:--------------------------:|:----------------------------:|:-------------------------:|:---------------------------:|:----------------------------:|:------------------------------:|:--------------------------------:|:-----------------------------:|:----------------------:|:----------------------:| | 4970.1539 | 0.3623 | 100 | 4979.0801 | -0.0031 | -0.0046 | 0.5658 | 0.0014 | -260.1172 | -253.2325 | 1.6973 | 1.6355 | 1.6355 | 1.6973 | -253.2325 | -260.1172 | -252.9185 | -259.6586 | -0.3140 | -0.4586 | 2532.3372 | 2455.3159 | | 4913.6875 | 0.7246 | 200 | 4922.2964 | -0.0067 | -0.0090 | 0.5395 | 0.0023 | -260.5605 | -253.5932 | 1.6658 | 1.6047 | 1.6047 | 1.6658 | -253.5932 | -260.5605 | -252.9185 | -259.6586 | -0.6748 | -0.9019 | 2570.3391 | 2415.1426 | | 4852.6547 | 1.0870 | 300 | 4861.8960 | -0.0090 | -0.0170 | 0.4605 | 0.0079 | -261.3568 | -253.8218 | 1.6477 | 1.5895 | 1.5895 | 1.6477 | -253.8218 | -261.3568 | -252.9185 | -259.6586 | -0.9033 | -1.6982 | 2599.3752 | 2346.0071 | | 4810.0602 | 1.4493 | 400 | 4799.1152 | -0.0065 | -0.0219 | 0.5395 | 0.0154 | -261.8465 | -253.5692 | 1.6033 | 1.5489 | 1.5489 | 1.6033 | -253.5692 | -261.8465 | -252.9185 | -259.6586 | -0.6507 | -2.1879 | 2584.1985 | 2322.5535 | | 4686.3855 | 1.8116 | 500 | 4767.9019 | -0.0146 | -0.0351 | 0.5132 | 0.0205 | -263.1680 | -254.3759 | 1.5899 | 1.5348 | 1.5348 | 1.5899 | -254.3759 | -263.1680 | -252.9185 | -259.6586 | -1.4575 | -3.5093 | 2678.0864 | 2224.3416 | | 4647.1707 | 2.1739 | 600 | 4725.6548 | -0.0031 | -0.0264 | 0.5395 | 0.0233 | -262.3003 | -253.2256 | 1.5586 | 1.5054 | 1.5054 | 1.5586 | -253.2256 | -262.3003 | -252.9185 | -259.6586 | -0.3071 | -2.6417 | 2562.3191 | 2304.5745 | | 4590.507 | 2.5362 | 700 | 4709.8721 | -0.0028 | -0.0317 | 0.5658 | 0.0289 | -262.8335 | -253.2023 | 1.5311 | 1.4802 | 1.4802 | 1.5311 | -253.2023 | -262.8335 | -252.9185 | -259.6586 | -0.2839 | -3.1748 | 2563.1602 | 2266.7019 | | 4624.6344 | 2.8986 | 800 | 4685.7876 | -0.0021 | -0.0328 | 0.6316 | 0.0307 | -262.9392 | -253.1265 | 1.5168 | 1.4660 | 1.4660 | 1.5168 | -253.1265 | -262.9392 | -252.9185 | -259.6586 | -0.2080 | -3.2806 | 2564.3735 | 2277.4634 | | 4526.798 | 3.2609 | 900 | 4673.5791 | -0.0010 | -0.0339 | 0.5921 | 0.0329 | -263.0450 | -253.0172 | 1.5044 | 1.4543 | 1.4543 | 1.5044 | -253.0172 | -263.0450 | -252.9185 | -259.6586 | -0.0987 | -3.3863 | 2560.7192 | 2277.5515 | | 4599.7109 | 3.6232 | 1000 | 4664.8169 | 0.0018 | -0.0326 | 0.5658 | 0.0344 | -262.9172 | -252.7381 | 1.4973 | 1.4480 | 1.4480 | 1.4973 | -252.7381 | -262.9172 | -252.9185 | -259.6586 | 0.1804 | -3.2586 | 2535.9368 | 2302.0969 | | 4598.4699 | 3.9855 | 1100 | 4659.8091 | 0.0225 | -0.0149 | 0.6579 | 0.0374 | -261.1521 | -250.6732 | 1.4704 | 1.4246 | 1.4246 | 1.4704 | -250.6732 | -261.1521 | -252.9185 | -259.6586 | 2.2452 | -1.4935 | 2330.4351 | 2454.2285 | | 4434.3441 | 4.3478 | 1200 | 4652.3701 | -0.0064 | -0.0448 | 0.5789 | 0.0383 | -264.1339 | -253.5595 | 1.4648 | 1.4176 | 1.4176 | 1.4648 | -253.5595 | -264.1339 | -252.9185 | -259.6586 | -0.6410 | -4.4752 | 2633.1008 | 2222.5164 | | 4673.5336 | 4.7101 | 1300 | 4629.2358 | 0.0059 | -0.0337 | 0.6053 | 0.0396 | -263.0263 | -252.3293 | 1.4597 | 1.4137 | 1.4137 | 1.4597 | -252.3293 | -263.0263 | -252.9185 | -259.6586 | 0.5892 | -3.3676 | 2506.5920 | 2317.5457 | | 4551.7766 | 5.0725 | 1400 | 4636.1592 | 0.0046 | -0.0350 | 0.6053 | 0.0396 | -263.1627 | -252.4586 | 1.4595 | 1.4144 | 1.4144 | 1.4595 | -252.4586 | -263.1627 | -252.9185 | -259.6586 | 0.4598 | -3.5041 | 2524.4553 | 2311.0466 | | 4481.4781 | 5.4348 | 1500 | 4616.7266 | 0.0125 | -0.0289 | 0.5921 | 0.0413 | -262.5467 | -251.6734 | 1.4468 | 1.4029 | 1.4029 | 1.4468 | -251.6734 | -262.5467 | -252.9185 | -259.6586 | 1.2451 | -2.8881 | 2446.6792 | 2368.6218 | | 4557.7566 | 5.7971 | 1600 | 4618.0537 | 0.0014 | -0.0416 | 0.5921 | 0.0430 | -263.8221 | -252.7794 | 1.4428 | 1.3976 | 1.3976 | 1.4428 | -252.7794 | -263.8221 | -252.9185 | -259.6586 | 0.1390 | -4.1635 | 2564.9141 | 2269.4070 | | 4507.4234 | 6.1594 | 1700 | 4618.0 | 0.0009 | -0.0413 | 0.5921 | 0.0422 | -263.7893 | -252.8316 | 1.4382 | 1.3934 | 1.3934 | 1.4382 | -252.8316 | -263.7893 | -252.9185 | -259.6586 | 0.0869 | -4.1307 | 2573.3213 | 2274.9512 | | 4566.6648 | 6.5217 | 1800 | 4619.3325 | 0.0061 | -0.0369 | 0.5921 | 0.0430 | -263.3517 | -252.3105 | 1.4413 | 1.3975 | 1.3975 | 1.4413 | -252.3105 | -263.3517 | -252.9185 | -259.6586 | 0.6080 | -3.6930 | 2512.9187 | 2304.7549 | | 4682.7492 | 6.8841 | 1900 | 4616.8687 | 0.0066 | -0.0366 | 0.5921 | 0.0432 | -263.3144 | -252.2579 | 1.4407 | 1.3967 | 1.3967 | 1.4407 | -252.2579 | -263.3144 | -252.9185 | -259.6586 | 0.6606 | -3.6557 | 2507.0054 | 2307.5239 | | 4486.1707 | 7.2464 | 2000 | 4616.3892 | 0.0062 | -0.0377 | 0.5789 | 0.0439 | -263.4255 | -252.2975 | 1.4378 | 1.3932 | 1.3932 | 1.4378 | -252.2975 | -263.4255 | -252.9185 | -259.6586 | 0.6210 | -3.7668 | 2509.9634 | 2298.5259 | | 4477.8289 | 7.6087 | 2100 | 4617.2290 | 0.0069 | -0.0354 | 0.5789 | 0.0423 | -263.1952 | -252.2293 | 1.4363 | 1.3925 | 1.3925 | 1.4363 | -252.2293 | -263.1952 | -252.9185 | -259.6586 | 0.6892 | -3.5365 | 2506.2578 | 2318.2375 | | 4520.1934 | 7.9710 | 2200 | 4613.5840 | 0.0069 | -0.0357 | 0.6053 | 0.0425 | -263.2242 | -252.2310 | 1.4371 | 1.3941 | 1.3941 | 1.4371 | -252.2310 | -263.2242 | -252.9185 | -259.6586 | 0.6874 | -3.5656 | 2507.3259 | 2312.9116 | ### Framework versions - Transformers 4.42.0 - Pytorch 2.3.0+cu121 - Datasets 2.14.6 - Tokenizers 0.19.1