adpretko/train-armv8-O2-epoch1and2-AMD

train-armv8-O2_epoch1and2

This model is a fine-tuned version of saves/train-armv8-O2_epoch1and2/checkpoint-3200 on the train-armv8-O2-verbose_part_00, the train-armv8-O2-verbose_part_01, the train-armv8-O2-verbose_part_02, the train-armv8-O2-verbose_part_03, the train-armv8-O2-verbose_part_04, the train-armv8-O2-verbose_part_05, the train-armv8-O2-verbose_part_06, the train-armv8-O2-verbose_part_07, the train-armv8-O2-verbose_part_08, the train-armv8-O2-verbose_part_09, the train-armv8-O2-verbose_part_10, the train-armv8-O2-verbose_part_11, the train-armv8-O2-verbose_part_12, the train-armv8-O2-verbose_part_13, the train-armv8-O2-verbose_part_14, the train-armv8-O2-verbose_part_15, the train-armv8-O2-verbose_part_16, the train-armv8-O2-verbose_part_17, the train-armv8-O2-verbose_part_18, the train-armv8-O2-verbose_part_19, the train-armv8-O2-verbose_part_20, the train-armv8-O2-verbose_part_21, the train-armv8-O2-verbose_part_22, the train-armv8-O2-verbose_part_23, the train-armv8-O2-verbose_part_24, the train-armv8-O2-verbose_part_25, the train-armv8-O2-verbose_part_26, the train-armv8-O2-verbose_part_27, the train-armv8-O2-verbose_part_28, the train-armv8-O2-verbose_part_29, the train-armv8-O2-verbose_part_30, the train-armv8-O2-verbose_part_31, the train-armv8-O2-verbose_part_32, the train-armv8-O2-verbose_part_33, the train-armv8-O2-verbose_part_34, the train-armv8-O2-verbose_part_35, the train-armv8-O2-verbose_part_36, the train-armv8-O2-verbose_part_37, the train-armv8-O2-verbose_part_38, the train-armv8-O2-verbose_part_39, the train-armv8-O2-verbose_part_40 and the train-armv8-O2-verbose_part_41 datasets.

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: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 8
• total_train_batch_size: 512
• total_eval_batch_size: 64
• 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: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 2.0

Training results

Framework versions

• Transformers 4.55.0
• Pytorch 2.8.0+rocm6.3
• Datasets 3.6.0
• Tokenizers 0.21.1