results

This model is a fine-tuned version of defog/llama-3-sqlcoder-8b on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.2568
• Rewards/chosen: -0.0126
• Rewards/rejected: -0.0217
• Rewards/accuracies: 0.8944
• Rewards/margins: 0.0091
• Logps/rejected: -0.2167
• Logps/chosen: -0.1258
• Logits/rejected: 0.1307
• Logits/chosen: 0.1283
• Nll Loss: 0.2132
• Log Odds Ratio: -0.4354
• Log Odds Chosen: 0.6768

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: 8e-06
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 10
• 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
0.9481	0.2	72	0.9541	-0.0776	-0.0797	0.7143	0.0021	-0.7975	-0.7765	-0.3031	-0.3167	0.8875	-0.6703	0.0480
0.7313	0.4	144	0.7089	-0.0551	-0.0596	0.8292	0.0045	-0.5962	-0.5513	-0.1005	-0.1135	0.6459	-0.6312	0.1330
0.547	0.6	216	0.4407	-0.0292	-0.0367	0.8882	0.0075	-0.3670	-0.2924	-0.0064	-0.0109	0.3866	-0.5408	0.3609
0.2547	0.8	288	0.3018	-0.0164	-0.0250	0.8882	0.0085	-0.2498	-0.1644	0.0633	0.0592	0.2551	-0.4664	0.5805
0.3407	1.0	360	0.2568	-0.0126	-0.0217	0.8944	0.0091	-0.2167	-0.1258	0.1307	0.1283	0.2132	-0.4354	0.6768

Framework versions

• PEFT 0.11.1
• Transformers 4.42.1
• Pytorch 2.3.1+cu118
• Datasets 2.20.0
• Tokenizers 0.19.1