yiran-wang3/ds_chat_sppo_hard_new_iter0_constant_1e-6

ds_chat_sppo_hard_new_iter0_2024-09-17-14.35

This model is a fine-tuned version of deepseek-ai/deepseek-llm-7b-chat on the self-generate/ds_chat_original_cn_mining_oj_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: 0.4365
• Rewards/chosen: 0.0466
• Rewards/rejected: -0.0388
• Rewards/accuracies: 0.5769
• Rewards/margins: 0.0854
• Logps/rejected: -287.4509
• Logps/chosen: -250.6894
• Logits/rejected: 1.4344
• Logits/chosen: 1.2252
• Debug/policy Chosen Logits: 1.2252
• Debug/policy Rejected Logits: 1.4344
• Debug/policy Chosen Logps: -250.6894
• Debug/policy Rejected Logps: -287.4509
• Debug/reference Chosen Logps: -255.3497
• Debug/reference Rejected Logps: -283.5705

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-06
• 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: constant
• lr_scheduler_warmup_ratio: 0.1
• lr_scheduler_warmup_steps: 100
• num_epochs: 1.0

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
0.4429	0.6329	100	0.4365	0.0466	-0.0388	0.5769	0.0854	-287.4509	-250.6894	1.4344	1.2252	1.2252	1.4344	-250.6894	-287.4509	-255.3497	-283.5705

Framework versions

• Transformers 4.42.0
• Pytorch 2.3.0+cu121
• Datasets 2.14.6
• Tokenizers 0.19.1