llama-13b_oasst1_l0.0002_64

This model is a fine-tuned version of huggyllama/llama-13b on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.3242

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: 0.0002
• train_batch_size: 1
• eval_batch_size: 1
• seed: 0
• gradient_accumulation_steps: 16
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant
• lr_scheduler_warmup_ratio: 0.03
• training_steps: 0

Training results

Training Loss	Epoch	Step	Validation Loss
1.4264	0.0018	1	1.6321
1.4346	0.3392	187	1.2420
1.081	0.6783	374	1.2324
1.3295	1.0175	561	1.2309
1.1747	1.3566	748	1.2357
1.1651	1.6958	935	1.2310
0.8447	2.0349	1122	1.2523
1.009	2.3741	1309	1.3121
0.8776	2.7132	1496	1.2949

Framework versions

• PEFT 0.12.1.dev0
• Transformers 4.45.0.dev0
• Pytorch 2.3.0+cu121
• Datasets 2.19.0
• Tokenizers 0.19.1