--- base_model: roneneldan/TinyStories-33M library_name: Distily tags: - generated_from_trainer model-index: - name: distily_bench_obj_cross results: [] --- # distily_bench_obj_cross This student model is distilled from the teacher model [roneneldan/TinyStories-33M](https://huggingface.co/roneneldan/TinyStories-33M) using the dataset (unspecified). The [Distily](https://github.com/lapp0/distily) library was used for this distillation. It achieves the following results on the evaluation set: - eval_enwikippl: 3944.9531 - eval_frwikippl: 30197.7344 - eval_zhwikippl: 52496.3438 - eval_tinystoriesppl: 1385.5492 - eval_loss: 16.6107 - eval_runtime: 66.9937 - eval_samples_per_second: 74.634 - eval_steps_per_second: 9.329 ## Training procedure ### Training hyperparameters The following hyperparameters were used during training: - distillation_objective: DistillationObjective(logits_loss_component=LossComponent(label=logits, weight=1, loss_fn=kl, layer_mapper=None, projector=None), hs_loss_component=LossComponent(label=hs, weight=10.0, loss_fn=raw_mse, layer_mapper=None, projector=None), attn_loss_component=LossComponent(label=attn, weight=10.0, loss_fn=mse, layer_mapper=None, projector=None)) - train_embeddings: True - learning_rate: 0.0004 - 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 - lr_scheduler_warmup_ratio: 0.1 - num_epochs: 1.0 ### Resource Usage Peak GPU Memory: 8.2677 GB ### Eval-Phase Metrics | step | epoch | enwikippl | frwikippl | loss | runtime | samples_per_second | steps_per_second | tinystoriesppl | zhwikippl | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | **teacher eval** | | 169.9865 | 47377.9414 | | | | | 3.9789 | 4998.1294 | | 0 | 0 | 21397.4785 | 57946.0117 | 18.3162 | 67.1093 | 74.505 | 9.313 | 12321.8145 | 60955.8008 | | 3000 | 0.0485 | 3940.0654 | 30197.7344 | 16.6107 | 67.006 | 74.62 | 9.328 | 1383.7178 | 52496.3438 | | 6000 | 0.0970 | 3937.6238 | 30180.7188 | 16.6119 | 67.1095 | 74.505 | 9.313 | 1383.9467 | 52496.3438 | | 9000 | 0.1455 | 3944.9531 | 30197.7344 | 16.6107 | 66.9937 | 74.634 | 9.329 | 1385.5492 | 52496.3438 | | 12000 | 0.1939 | 3944.9531 | 30197.7344 | 16.6115 | 67.0666 | 74.553 | 9.319 | 1384.8617 | 52496.3438 | | 15000 | 0.2424 | 3937.6238 | 30180.7188 | 16.6121 | 66.6143 | 75.059 | 9.382 | 1385.3201 | 52524.3359 | | 18000 | 0.2909 | 3937.6238 | 30180.7188 | 16.6117 | 66.7575 | 74.898 | 9.362 | 1384.1750 | 52552.3945 | | 21000 | 0.3394 | 3944.9531 | 30197.7344 | 16.6115 | 66.679 | 74.986 | 9.373 | 1384.4041 | 52496.3438 | | 24000 | 0.3879 | 3944.9531 | 30197.7344 | 16.6121 | 66.8908 | 74.749 | 9.344 | 1384.4041 | 52468.3164 | | 27000 | 0.4364 | 3942.5085 | 30197.7344 | 16.6117 | 66.4311 | 75.266 | 9.408 | 1383.0317 | 52496.3438 | | 30000 | 0.4848 | 3940.0654 | 30180.7188 | 16.6107 | 66.4762 | 75.215 | 9.402 | 1383.2599 | 52496.3438 | | 33000 | 0.5333 | 3937.6238 | 30197.7344 | 16.6107 | 66.4814 | 75.209 | 9.401 | 1382.8029 | 52496.3438 | | 36000 | 0.5818 | 3942.5085 | 30180.7188 | 16.6111 | 67.3001 | 74.294 | 9.287 | 1385.3201 | 52496.3438 | | 39000 | 0.6303 | 3937.6238 | 30180.7188 | 16.6115 | 67.0065 | 74.62 | 9.327 | 1383.4888 | 52496.3438 | | 42000 | 0.6788 | 3942.5085 | 30197.7344 | 16.6109 | 66.7444 | 74.913 | 9.364 | 1384.1750 | 52496.3438 | | 45000 | 0.7273 | 3941.2869 | 30197.7344 | 16.6115 | 67.1516 | 74.458 | 9.307 | 1382.8029 | 52496.3438 | | 48000 | 0.7758 | 3944.9531 | 30180.7188 | 16.6107 | 66.7762 | 74.877 | 9.36 | 1386.6947 | 52524.3359 | | 51000 | 0.8242 | 3942.5085 | 30197.7344 | 16.6111 | 67.2623 | 74.336 | 9.292 | 1384.8617 | 52496.3438 | | 54000 | 0.8727 | 3944.9531 | 30180.7188 | 16.6107 | 66.724 | 74.936 | 9.367 | 1385.3201 | 52496.3438 | | 57000 | 0.9212 | 3941.2869 | 30197.7344 | 16.6115 | 67.0602 | 74.56 | 9.32 | 1382.8029 | 52468.3164 | | 60000 | 0.9697 | 3942.5085 | 30197.7344 | 16.6119 | 67.4137 | 74.169 | 9.271 | 1382.8029 | 52468.3164 | | 61875 | 1.0 | 3937.6238 | 30180.7188 | 16.6119 | 67.1794 | 74.428 | 9.303 | 1383.7178 | 52496.3438 | ### Framework versions - Distily 0.2.0 - Transformers 4.44.0 - Pytorch 2.3.0 - Datasets 2.21.0