sheared_llama_1.3b-reazon_v2-ja_en_trans-T2T

This model is a fine-tuned version of princeton-nlp/Sheared-LLaMA-1.3B-ShareGPT on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.3844

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: 2e-05
• train_batch_size: 32
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 4
• total_train_batch_size: 1024
• total_eval_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.01
• training_steps: 200

Training results

Training Loss	Epoch	Step	Validation Loss
2.0409	0.0114	10	1.7202
1.6901	0.0229	20	1.6101
1.5859	0.0343	30	1.5446
1.5533	0.0458	40	1.5029
1.4937	0.0572	50	1.4722
1.4802	0.0687	60	1.4492
1.4484	0.0801	70	1.4302
1.4292	0.0916	80	1.4183
1.4203	0.1030	90	1.4078
1.4184	0.1145	100	1.3985
1.4045	0.1259	110	1.3923
1.4125	0.1374	120	1.3886
1.4098	0.1488	130	1.3877
1.3921	0.1603	140	1.3859
1.3984	0.1717	150	1.3851
1.3858	0.1832	160	1.3845
1.3995	0.1946	170	1.3842
1.3943	0.2061	180	1.3847
1.3988	0.2175	190	1.3844
1.3969	0.2290	200	1.3844

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1