mistral_extended_darulm_20_05_24_part1-2_32000_bpe_full_lr1e4_bs256

This model is a fine-tuned version of RefalMachine/mistral_extended_darulm_20_05_24_part1-2_32000_bpe_mean_init_03_07_24 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 2.0635
• Accuracy: 0.5602

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: 0.0001
• 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
2.3955	0.05	2000	2.1922	0.5420
2.3675	0.09	4000	2.1648	0.5452
2.3365	0.14	6000	2.1490	0.5473
2.3455	0.18	8000	2.1394	0.5484
2.3244	0.23	10000	2.1293	0.5498
2.3287	0.27	12000	2.1223	0.5511
2.3158	0.32	14000	2.1145	0.5521
2.3154	0.36	16000	2.1067	0.5533
2.2754	0.41	18000	2.1002	0.5542
2.3085	0.45	20000	2.0934	0.5552
2.2918	0.5	22000	2.0869	0.5565
2.2725	0.54	24000	2.0813	0.5573
2.286	0.59	26000	2.0756	0.5582
2.2879	0.63	28000	2.0717	0.5588
2.2864	0.68	30000	2.0684	0.5594
2.258	0.72	32000	2.0662	0.5598
2.2726	0.77	34000	2.0648	0.5601
2.2586	0.81	36000	2.0640	0.5602
2.2568	0.86	38000	2.0636	0.5602
2.2492	0.9	40000	2.0635	0.5602
2.2676	0.95	42000	2.0635	0.5603
2.2567	0.99	44000	2.0635	0.5602

Framework versions

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