chessgpt2-small-m

This model is a fine-tuned version of gpt2 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.9060

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: 0.0004
• train_batch_size: 32
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.04
• num_epochs: 3
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
2.3011	0.1280	1000	1.7077
1.5883	0.2560	2000	1.4142
1.3992	0.3839	3000	1.2912
1.2978	0.5119	4000	1.2150
1.2322	0.6399	5000	1.1646
1.1846	0.7679	6000	1.1219
1.1477	0.8958	7000	1.0882
1.1142	1.0238	8000	1.0618
1.0801	1.1518	9000	1.0461
1.0616	1.2798	10000	1.0251
1.0409	1.4077	11000	1.0020
1.0253	1.5357	12000	0.9859
1.0098	1.6637	13000	0.9726
0.9947	1.7917	14000	0.9585
0.9817	1.9196	15000	0.9472
0.9591	2.0476	16000	0.9364
0.9338	2.1756	17000	0.9291
0.9273	2.3036	18000	0.9212
0.9219	2.4315	19000	0.9153
0.9167	2.5595	20000	0.9107
0.9123	2.6875	21000	0.9081
0.9103	2.8155	22000	0.9065
0.9092	2.9434	23000	0.9060

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.19.1