rumodernbert_small_distill

This model is a fine-tuned version of deepvk/RuModernBERT-small on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.4484
• F1: 0.8670
• Precision: 0.8632
• Recall: 0.8708

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: 3e-05
• train_batch_size: 16
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 0.1
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	F1	Precision	Recall
4.8701	0.32	200	2.0232	0.4841	0.4288	0.5558
2.6246	0.64	400	1.0239	0.6799	0.6425	0.7219
2.0708	0.96	600	0.9332	0.7623	0.7271	0.8010
1.5292	1.28	800	0.7393	0.7809	0.7670	0.7953
1.1811	1.6	1000	0.6396	0.8126	0.8101	0.8150
1.2625	1.92	1200	0.6523	0.8085	0.7990	0.8182
0.8249	2.24	1400	0.5636	0.8322	0.8294	0.8351
0.8175	2.56	1600	0.5814	0.8361	0.8258	0.8467
0.7697	2.88	1800	0.5125	0.8446	0.8333	0.8563
0.5251	3.2	2000	0.5297	0.8374	0.8268	0.8483
0.5485	3.52	2200	0.4821	0.8542	0.8513	0.8571
0.5174	3.84	2400	0.4683	0.8614	0.8549	0.8680
0.4142	4.16	2600	0.4663	0.8614	0.8564	0.8664
0.3958	4.48	2800	0.4507	0.8622	0.8560	0.8684
0.4127	4.8	3000	0.4406	0.8576	0.8490	0.8664
0.2867	5.12	3200	0.4370	0.8677	0.8592	0.8764
0.3076	5.44	3400	0.4394	0.8612	0.8502	0.8724
0.2746	5.76	3600	0.4325	0.8664	0.8624	0.8704
0.2701	6.08	3800	0.4484	0.8670	0.8632	0.8708

Framework versions

• Transformers 5.3.0
• Pytorch 2.10.0+cu128
• Datasets 4.7.0
• Tokenizers 0.22.2