ft-smollm-135M-instruct-on-hf-ultrafeedback_rob

This model is a fine-tuned version of HuggingFaceTB/SmolLM-135M-Instruct on the HuggingFaceH4/ultrafeedback_binarized dataset. It achieves the following results on the evaluation set:

• Loss: 1.1429
• Rewards/chosen: -0.1303
• Rewards/rejected: -0.1304
• Rewards/accuracies: 0.4670
• Rewards/margins: 0.0000
• Logps/rejected: -1.3036
• Logps/chosen: -1.3032
• Logits/rejected: 27.7664
• Logits/chosen: 27.4331
• Nll Loss: 1.0675
• Log Odds Ratio: -0.7542
• Log Odds Chosen: 0.0132

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: 0.0003
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen	Nll Loss	Log Odds Ratio	Log Odds Chosen
1.3569	0.8	100	1.1429	-0.1303	-0.1304	0.4670	0.0000	-1.3036	-1.3032	27.7664	27.4331	1.0675	-0.7542	0.0132

Framework versions

• Transformers 4.39.3
• Pytorch 2.1.2
• Datasets 2.18.0
• Tokenizers 0.15.2