Tensoic/Tiny-Llama-openhermes-1.1B-step-715k-1.5T

This model is a fine-tuned version of PY007/TinyLlama-1.1B-intermediate-step-715k-1.5T on the openhermes dataset. It achieves the following results on the evaluation set:

• Loss: 1.2355

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0002
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• total_train_batch_size: 8
• total_eval_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 100
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss
1.4654	0.0	1	3.5326
1.2162	0.05	1503	1.9335
1.1918	0.1	3006	1.7391
1.4188	0.15	4509	1.7574
1.8281	0.2	6012	1.6704
0.8639	0.25	7515	1.7459
1.3764	0.3	9018	1.6832
2.1172	0.35	10521	1.6398
1.1855	0.4	12024	1.6007
1.5604	0.45	13527	1.5256
1.0224	0.5	15030	1.4891
1.5582	0.55	16533	1.4903
0.9489	0.6	18036	1.4179
1.67	0.65	19539	1.4585
0.8542	0.7	21042	1.3810
1.5301	0.75	22545	1.3645
0.951	0.8	24048	1.3087
1.1791	0.85	25551	1.3018
1.3342	0.9	27054	1.2595
1.1221	0.95	28557	1.2355

Framework versions

• Transformers 4.34.0
• Pytorch 2.0.1+cu117
• Datasets 2.14.6
• Tokenizers 0.14.1