childes-segmentation-18M-gpt2_lm-model

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

• Loss: 1.5598
• Model Preparation Time: 0.0013
• Perplexity: 4.7580
• Bpc: 2.2503
• Spike Seg Type Fscore Entropy: 0.5424
• Spike Seg Boundary Fscore Entropy: 0.7652
• Absolute Seg Type Fscore Entropy: 0.4188
• Absolute Seg Boundary Fscore Entropy: 0.6411
• Spike Seg Type Fscore Increase in entropy: 0.5339
• Spike Seg Boundary Fscore Increase in entropy: 0.7796
• Absolute Seg Type Fscore Increase in entropy: 0.5744
• Absolute Seg Boundary Fscore Increase in entropy: 0.7708
• Spike Seg Type Fscore Loss: 0.4461
• Spike Seg Boundary Fscore Loss: 0.6948
• Absolute Seg Type Fscore Loss: 0.3397
• Absolute Seg Boundary Fscore Loss: 0.6138
• Spike Seg Type Fscore Increase in loss: 0.5024
• Spike Seg Boundary Fscore Increase in loss: 0.7430
• Absolute Seg Type Fscore Increase in loss: 0.5046
• Absolute Seg Boundary Fscore Increase in loss: 0.7437
• Spike Seg Type Fscore Rank: 0.4778
• Spike Seg Boundary Fscore Rank: 0.6585
• Absolute Seg Type Fscore Rank: 0.3314
• Absolute Seg Boundary Fscore Rank: 0.5551
• Spike Seg Type Fscore Increase in rank: 0.4977
• Spike Seg Boundary Fscore Increase in rank: 0.6963
• Absolute Seg Type Fscore Increase in rank: 0.4902
• Absolute Seg Boundary Fscore Increase in rank: 0.7065
• Spike Seg Type Fscore Boundary prediction: 0.5365
• Spike Seg Boundary Fscore Boundary prediction: 0.8041
• Absolute Seg Type Fscore Boundary prediction: 0.3187
• Absolute Seg Boundary Fscore Boundary prediction: 0.7456
• Spike Seg Type Fscore Increase in boundary prediction: 0.5171
• Spike Seg Boundary Fscore Increase in boundary prediction: 0.7895
• Absolute Seg Type Fscore Increase in boundary prediction: 0.2577
• Absolute Seg Boundary Fscore Increase in boundary prediction: 0.5526
• Spike Seg Type Fscore Majority vote cutoff: 0.6165
• Spike Seg Type Fscore Majority vote spike: 0.4770
• Absolute Seg Type Fscore Majority vote cutoff: 0.5211
• Absolute Seg Type Fscore Majority vote spike: 0.6022
• Spike Seg Boundary Fscore Majority vote cutoff: 0.8101
• Spike Seg Boundary Fscore Majority vote spike: 0.7717
• Absolute Seg Boundary Fscore Majority vote cutoff: 0.7609
• Absolute Seg Boundary Fscore Majority vote spike: 0.8128

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: 0.001
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 60000
• training_steps: 200000

Training results

Training Loss	Epoch	Step	Validation Loss	Model Preparation Time	Perplexity	Bpc	Spike Seg Type Fscore Entropy	Spike Seg Boundary Fscore Entropy	Absolute Seg Type Fscore Entropy	Absolute Seg Boundary Fscore Entropy	Spike Seg Type Fscore Increase in entropy	Spike Seg Boundary Fscore Increase in entropy	Absolute Seg Type Fscore Increase in entropy	Absolute Seg Boundary Fscore Increase in entropy	Spike Seg Type Fscore Loss	Spike Seg Boundary Fscore Loss	Absolute Seg Type Fscore Loss	Absolute Seg Boundary Fscore Loss	Spike Seg Type Fscore Increase in loss	Spike Seg Boundary Fscore Increase in loss	Absolute Seg Type Fscore Increase in loss	Absolute Seg Boundary Fscore Increase in loss	Spike Seg Type Fscore Rank	Spike Seg Boundary Fscore Rank	Absolute Seg Type Fscore Rank	Absolute Seg Boundary Fscore Rank	Spike Seg Type Fscore Increase in rank	Spike Seg Boundary Fscore Increase in rank	Absolute Seg Type Fscore Increase in rank	Absolute Seg Boundary Fscore Increase in rank	Spike Seg Type Fscore Boundary prediction	Spike Seg Boundary Fscore Boundary prediction	Absolute Seg Type Fscore Boundary prediction	Absolute Seg Boundary Fscore Boundary prediction	Spike Seg Type Fscore Increase in boundary prediction	Spike Seg Boundary Fscore Increase in boundary prediction	Absolute Seg Type Fscore Increase in boundary prediction	Absolute Seg Boundary Fscore Increase in boundary prediction	Spike Seg Type Fscore Majority vote cutoff	Spike Seg Type Fscore Majority vote spike	Absolute Seg Type Fscore Majority vote cutoff	Absolute Seg Type Fscore Majority vote spike	Spike Seg Boundary Fscore Majority vote cutoff	Spike Seg Boundary Fscore Majority vote spike	Absolute Seg Boundary Fscore Majority vote cutoff	Absolute Seg Boundary Fscore Majority vote spike
1.418	4.5290	20000	1.5456	0.0013	4.6908	2.2298	0.5202	0.7537	0.3779	0.6326	0.4886	0.7542	0.5462	0.7705	0.4673	0.7125	0.1852	0.6119	0.5	0.7439	0.5140	0.7503	0.4580	0.6515	0.3252	0.5828	0.4965	0.6950	0.5032	0.6947	0.5137	0.7850	0.3688	0.5036	0.4720	0.7564	0.2699	0.7468	0.6117	0.4695	0.4865	0.5951	0.8190	0.7707	0.7754	0.8128
1.3419	9.0580	40000	1.5062	0.0013	4.5097	2.1730	0.5334	0.7731	0.4017	0.6446	0.4934	0.7641	0.5823	0.7738	0.4661	0.7199	0.3633	0.6170	0.5182	0.7655	0.5230	0.7541	0.4670	0.6554	0.3283	0.5868	0.5086	0.7047	0.5374	0.7079	0.5384	0.8	0.2665	0.7782	0.4865	0.7603	0.2625	0.7599	0.6162	0.4752	0.5467	0.6404	0.8207	0.7733	0.8083	0.8297
1.2911	13.5870	60000	1.4740	0.0013	4.3665	2.1265	0.5431	0.7827	0.4017	0.6226	0.5042	0.7663	0.5776	0.7816	0.4832	0.7214	0.2106	0.6109	0.5060	0.7533	0.5344	0.7594	0.4732	0.6519	0.3198	0.5685	0.4923	0.6900	0.4931	0.6954	0.5379	0.8083	0.3506	0.4930	0.5008	0.7768	0.2621	0.7390	0.6045	0.4492	0.4242	0.6183	0.8186	0.7659	0.7554	0.8234
1.2397	18.1159	80000	1.4710	0.0013	4.3537	2.1222	0.5355	0.7742	0.4044	0.6203	0.5169	0.7687	0.5692	0.7722	0.4724	0.7140	0.3523	0.6225	0.5088	0.7554	0.5271	0.7526	0.4918	0.6667	0.3442	0.5695	0.4949	0.6899	0.5318	0.7059	0.5409	0.8024	0.2643	0.785	0.5060	0.7725	0.2590	0.7676	0.6034	0.4954	0.5495	0.6285	0.8290	0.7749	0.8150	0.8230
1.1906	22.6449	100000	1.4768	0.0013	4.3788	2.1305	0.5342	0.7807	0.4052	0.6284	0.5238	0.7770	0.5770	0.7649	0.4817	0.7269	0.3506	0.6181	0.5196	0.7627	0.5321	0.7583	0.4850	0.6691	0.3317	0.5690	0.5012	0.6983	0.4975	0.7142	0.5420	0.8090	0.2637	0.7085	0.5230	0.7840	0.2821	0.4171	0.6129	0.4882	0.5175	0.6171	0.8043	0.7814	0.7775	0.8289
1.1539	27.1739	120000	1.4986	0.0013	4.4756	2.1621	0.5355	0.7782	0.4135	0.6490	0.5242	0.7819	0.5790	0.7795	0.4570	0.7061	0.3286	0.6123	0.4988	0.7528	0.5187	0.7281	0.4779	0.6674	0.3452	0.5604	0.4854	0.6910	0.5449	0.7106	0.5502	0.8088	0.2884	0.8028	0.5251	0.7881	0.3504	0.7872	0.6119	0.4789	0.5543	0.6131	0.8316	0.7727	0.7959	0.8165
1.1198	31.7029	140000	1.4979	0.0013	4.4723	2.1610	0.5628	0.7849	0.4080	0.5883	0.5267	0.7764	0.5820	0.7557	0.4490	0.6987	0.3389	0.6187	0.4901	0.7447	0.5149	0.7496	0.4686	0.6553	0.3383	0.5647	0.5059	0.6940	0.5319	0.7036	0.5503	0.8056	0.2686	0.7966	0.5293	0.7900	0.2607	0.7840	0.6003	0.4854	0.5448	0.6101	0.8329	0.7729	0.8068	0.8146
1.0878	36.2319	160000	1.5223	0.0013	4.5827	2.1962	0.5553	0.7755	0.4237	0.6483	0.5196	0.7746	0.5848	0.7763	0.4497	0.6927	0.3273	0.6138	0.4858	0.7384	0.5113	0.7470	0.4716	0.6550	0.3289	0.5669	0.5098	0.69	0.5040	0.6965	0.5400	0.8044	0.3216	0.7546	0.5179	0.7898	0.5233	0.7859	0.6214	0.4608	0.5760	0.6141	0.8290	0.7650	0.8015	0.8115
1.0617	40.7609	180000	1.5411	0.0013	4.6699	2.2234	0.5562	0.7730	0.4066	0.6411	0.5280	0.7766	0.5836	0.7781	0.4479	0.6957	0.3336	0.6154	0.4893	0.7420	0.4984	0.7377	0.4808	0.6601	0.3386	0.5917	0.4836	0.6912	0.4857	0.7079	0.5423	0.8068	0.3296	0.7652	0.5232	0.7876	0.5623	0.4156	0.6383	0.4685	0.5665	0.6055	0.8162	0.7709	0.7762	0.8144
1.0394	45.2899	200000	1.5598	0.0013	4.7580	2.2503	0.5424	0.7652	0.4188	0.6411	0.5339	0.7796	0.5744	0.7708	0.4461	0.6948	0.3397	0.6138	0.5024	0.7430	0.5046	0.7437	0.4778	0.6585	0.3314	0.5551	0.4977	0.6963	0.4902	0.7065	0.5365	0.8041	0.3187	0.7456	0.5171	0.7895	0.2577	0.5526	0.6165	0.4770	0.5211	0.6022	0.8101	0.7717	0.7609	0.8128

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.0+cu118
• Datasets 2.18.0
• Tokenizers 0.19.1