distily_bench_gpt2_activation_loss

distily_bench_gpt2_activation_loss

This student model is distilled from the teacher model gpt2 using the dataset (unspecified).

The Distily library was used for this distillation.

It achieves the following results on the evaluation set:

• eval_enwikippl: 215.7906
• eval_frwikippl: 1306.3361
• eval_zhwikippl: 583.5945
• eval_loss: 1.2753
• eval_runtime: 34.5544
• eval_samples_per_second: 57.88
• eval_steps_per_second: 7.235

<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. ## 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:

• distillation_objective: MultiObjective(logits_weight=1, logits_loss_fn=(fn:kl_divergence_loss()), hs_weight=0.2, hs_loss_fn=(fn:soft_cross_entropy_loss()), attn_weight=0, attn_loss_fn=(fn:soft_mse_loss()))
• train_embeddings: True
• learning_rate: 4e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant
• num_epochs: 1.0

Resource Usage

Peak GPU Memory: 8.0904 GB

Eval-Phase Metrics

Framework versions

• Distily 0.2.0
• Transformers 4.44.0
• Pytorch 2.3.0
• Datasets 2.20.0