llama3_1_8B_all_zhtw_turn_lr1e-5_ep1_16_32_128

This model is a fine-tuned version of meta-llama/Llama-3.1-8B-Instruct on the multi_turn_miss_func_zh_tw_function_mix_sharegpt500_turn, the multi_turn_miss_param_zh_tw_function_mix_sharegpt500_turn, the multi_turn_zh_tw_function_mix_sharegpt500_turn, the irrelevance_zh_tw_sharegpt3000 and the apigen_zhtwV3_remove_sys_sharegpt datasets. It achieves the following results on the evaluation set:

• Loss: 0.7852

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-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 128
• total_train_batch_size: 128
• 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_steps: 15
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss
1.5538	0.0467	10	1.7379
1.4577	0.0933	20	1.6165
1.4283	0.1400	30	1.4638
0.9057	0.1866	40	1.3257
1.0109	0.2333	50	1.2057
0.9861	0.2799	60	1.0982
0.9197	0.3266	70	1.0025
0.7071	0.3732	80	0.9370
0.8973	0.4199	90	0.8938
0.8647	0.4665	100	0.8634
0.9744	0.5132	110	0.8408
0.7256	0.5598	120	0.8232
0.7378	0.6065	130	0.8116
0.7114	0.6531	140	0.8037
0.7931	0.6998	150	0.7972
0.754	0.7464	160	0.7923
0.9558	0.7931	170	0.7890
0.8876	0.8397	180	0.7870
0.7552	0.8864	190	0.7857
0.7593	0.9330	200	0.7851
0.8785	0.9797	210	0.7850

Framework versions

• PEFT 0.17.0
• Transformers 4.57.1
• Pytorch 2.4.1+cu121
• Datasets 4.0.0
• Tokenizers 0.22.2