mistral_darulm_20_05_24_part1-2_32000_unigram_part1_lr2e4_bs256

This model is a fine-tuned version of RefalMachine/mistral_darulm_20_05_24_part1-2_32000_unigram_mean_init_03_07_24 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 2.0916
• Accuracy: 0.5524

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: 32
• total_train_batch_size: 128
• total_eval_batch_size: 128
• 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
2.5086	0.09	2000	2.2880	0.5264
2.4581	0.18	4000	2.2482	0.5315
2.4032	0.26	6000	2.2253	0.5347
2.4038	0.35	8000	2.1971	0.5379
2.4058	0.44	10000	2.1745	0.5408
2.3731	0.53	12000	2.1466	0.5449
2.3023	0.61	14000	2.1238	0.5477
2.3136	0.7	16000	2.1059	0.5502
2.2773	0.79	18000	2.0954	0.5517
2.2638	0.88	20000	2.0920	0.5522
2.2632	0.96	22000	2.0917	0.5524

Framework versions

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