This model is a fine-tuned version of mistral-7b-base-sft-hh-helpful-4xh200-batch-64-20260418-015332 on the Anthropic/hh-rlhf dataset. It achieves the following results on the evaluation set:

Model description

More information needed

Intended uses & limitations

More information needed

More information needed

The following hyperparameters were used during training:

learning_rate: 5e-07
train_batch_size: 8
eval_batch_size: 8
seed: 42
distributed_type: multi-GPU
num_devices: 4
gradient_accumulation_steps: 2
total_train_batch_size: 64
total_eval_batch_size: 32
optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.1
num_epochs: 1

Training Loss	Epoch	Step	Validation Loss	Margin Dpo/beta	Margin Dpo/loss Margin Mean	Margin Dpo/beta Margin Mean	Margin Dpo/beta Margin Std	Margin Dpo/beta Margin Grad Mean	Margin Dpo/beta Margin Grad Std	Margin Dpo/margin Mean	Margin Dpo/margin Std	Logps/chosen	Logps/rejected	Logps/ref Chosen	Logps/ref Rejected	Logits/chosen	Logits/rejected
0.5701	0.1468	100	0.4649	0.1000	21.4763	2.1476	2.7958	-0.2712	0.2723	21.4763	27.9583	-85.6564	-130.9220	-68.2911	-92.0804	-3.0278	-3.0125
0.4771	0.2937	200	0.4174	0.1000	34.4413	3.4441	3.9020	-0.2232	0.2800	34.4413	39.0204	-102.7860	-161.0166	-68.2911	-92.0804	-2.5747	-2.5539
0.2928	0.4405	300	0.3567	0.1000	36.4753	3.6475	3.8633	-0.2092	0.2657	36.4753	38.6331	-108.3402	-168.6048	-68.2911	-92.0804	-2.6047	-2.5816
0.2693	0.5874	400	0.3643	0.1000	40.9015	4.0902	4.2092	-0.1982	0.2723	40.9015	42.0918	-112.8932	-177.5841	-68.2911	-92.0804	-2.4960	-2.4742
0.3872	0.7342	500	0.3395	0.1000	40.9407	4.0941	4.1148	-0.1926	0.2646	40.9407	41.1480	-112.8636	-177.5936	-68.2911	-92.0804	-2.4965	-2.4753
0.1695	0.8811	600	0.3349	0.1000	41.5652	4.1565	4.1510	-0.1899	0.2630	41.5652	41.5097	-113.7768	-179.1313	-68.2911	-92.0804	-2.5186	-2.4962