ruadapt_qwen2.5_3B_darulm_bpe_64000_full_lr2e4_bs256

This model is a fine-tuned version of RefalMachine/ruadapt_qwen2.5_3B_darulm_bpe_64000_mean_init on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 2.5958
• Accuracy: 0.4892

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: 4
• eval_batch_size: 4
• seed: 42
• distributed_type: multi-GPU
• num_devices: 64
• total_train_batch_size: 256
• total_eval_batch_size: 256
• optimizer: Adam with betas=(0.9,0.95) and epsilon=1e-05
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 100
• num_epochs: 1.0

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
2.8396	0.09	1000	2.6993	0.4783
2.7801	0.18	2000	2.6373	0.4849
2.7488	0.26	3000	2.6154	0.4868
2.7519	0.35	4000	2.6053	0.4880
2.7159	0.44	5000	2.6004	0.4888
2.7303	0.53	6000	2.5980	0.4888
2.7572	0.61	7000	2.5966	0.4891
2.7377	0.7	8000	2.5961	0.4892
2.7412	0.79	9000	2.5959	0.4893
2.7246	0.88	10000	2.5958	0.4892
2.7197	0.97	11000	2.5958	0.4892

Framework versions

• Transformers 4.37.2
• Pytorch 2.3.0a0+6ddf5cf85e.nv24.04
• Datasets 2.18.0
• Tokenizers 0.15.2