iitb_en_indic_robust_punctuation_model

This model is a fine-tuned version of ai4bharat/indictrans2-en-indic-dist-200M on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.2679
• Bleu: 12.6319
• Chrfpp: 35.424
• Comet: 0.549
• Bleurt: None
• Gen Len: 20.8721

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: 2e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 8

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Chrfpp	Comet	Bleurt	Gen Len
0.4295	0.2530	6000	0.4148	9.3672	31.5376	0.531	None	20.8761
0.4049	0.5059	12000	0.3887	9.7036	32.0364	0.5341	None	20.8717
0.408	0.7589	18000	0.3717	10.0257	32.395	0.5361	None	20.8778
0.3392	1.0119	24000	0.3609	10.2988	32.7193	0.5377	None	20.8761
0.3328	1.2649	30000	0.3508	10.4288	32.9579	0.5393	None	20.8767
0.3216	1.5178	36000	0.3408	10.7134	33.2408	0.5413	None	20.8727
0.3236	1.7708	42000	0.3333	10.7915	33.3923	0.5412	None	20.8762
0.2862	2.0238	48000	0.3286	10.8948	33.499	0.5414	None	20.8747
0.293	2.2768	54000	0.3217	11.0627	33.6756	0.5426	None	20.8708
0.2839	2.5297	60000	0.3168	11.1573	33.8408	0.5432	None	20.8743
0.2758	2.7827	66000	0.3109	11.3293	33.9424	0.5433	None	20.8745
0.2667	3.0357	72000	0.3075	11.3372	34.0924	0.5441	None	20.8726
0.2618	3.2886	78000	0.3040	11.5382	34.275	0.5452	None	20.871
0.2657	3.5416	84000	0.2997	11.6458	34.3695	0.5451	None	20.8744
0.2625	3.7946	90000	0.2961	11.7085	34.4786	0.5461	None	20.8709
0.2483	4.0476	96000	0.2931	11.8339	34.5484	0.5461	None	20.8726
0.2348	4.3005	102000	0.2897	11.8966	34.6524	0.5467	None	20.8713
0.2473	4.5535	108000	0.2870	11.9908	34.7337	0.5469	None	20.8717
0.2569	4.8065	114000	0.2829	12.0335	34.7971	0.5466	None	20.8748
0.2164	5.0594	120000	0.2822	12.1516	34.9439	0.5476	None	20.8711
0.2297	5.3124	126000	0.2802	12.2298	35.0421	0.548	None	20.8724
0.2261	5.5654	132000	0.2784	12.2563	35.0484	0.5481	None	20.8713
0.224	5.8184	138000	0.2749	12.3494	35.1261	0.5478	None	20.8722
0.2108	6.0713	144000	0.2750	12.3841	35.1321	0.5478	None	20.8729
0.2259	6.3243	150000	0.2726	12.4555	35.2263	0.5485	None	20.871
0.2172	6.5773	156000	0.2712	12.5084	35.2688	0.5484	None	20.8734
0.2139	6.8303	162000	0.2706	12.5631	35.3595	0.5489	None	20.8732
0.1997	7.0832	168000	0.2699	12.5653	35.367	0.5489	None	20.8736
0.2075	7.3362	174000	0.2690	12.5846	35.3799	0.5491	None	20.8711
0.199	7.5892	180000	0.2686	12.6322	35.4432	0.5493	None	20.8724
0.204	7.8421	186000	0.2679	12.6319	35.424	0.549	None	20.8721

Framework versions

• Transformers 4.53.2
• Pytorch 2.4.0a0+f70bd71a48.nv24.06
• Datasets 2.21.0
• Tokenizers 0.21.4