Modelpublic

abirmondalind/story2dialogue-SODA-BART-Large-LoRA

sourceHugging Faceapache-2.0updated 8mo agoView on Hugging Face

0likes7downloads

Model Card

story2dialogue-SODA-BART-Large-LoRA

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

—Loss: 1.3667
—Rouge1: 0.2848
—Rouge2: 0.1241
—Rougel: 0.2545
—Rougelsum: 0.2543
—Bleu: 0.0720
—Bleu1: 0.3624
—Bleu2: 0.1387
—Bleu3: 0.0789
—Bleu4: 0.0503
—Meteor: 0.2572
—Avg Distinct 1: 0.9633
—Avg Distinct 2: 0.9975
—Avg Distinct 3: 0.9930
—Avg Jaccard: 0.1273
—Gen Length: 10.7279

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: 0.0001
—trainbatchsize: 16
—evalbatchsize: 16
—seed: 42
—optimizer: Use OptimizerNames.ADAMWTORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizerargs=No additional optimizer arguments
—lrschedulertype: linear
—lrschedulerwarmup_steps: 500
—num_epochs: 3
—mixedprecisiontraining: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Bleu	Bleu1	Bleu2	Bleu3	Bleu4	Meteor	Avg Distinct 1	Avg Distinct 2	Avg Distinct 3	Avg Jaccard	Gen Length
1.7823	0.2004	1000	1.5746	0.2044	0.0646	0.1778	0.1776	0.0345	0.2822	0.0726	0.0343	0.0181	0.1810	0.9572	0.9978	0.9991	0.0760	10.4520
1.7107	0.4009	2000	1.5063	0.2327	0.0818	0.2037	0.2035	0.0461	0.3092	0.0926	0.0456	0.0248	0.2068	0.9526	0.9975	0.9997	0.0937	10.8824
1.6482	0.6013	3000	1.4627	0.2427	0.0899	0.2127	0.2127	0.0512	0.3183	0.1007	0.0513	0.0294	0.2168	0.9543	0.9967	0.9995	0.0997	10.9731
1.6283	0.8018	4000	1.4502	0.2498	0.0928	0.2192	0.2191	0.0535	0.3297	0.1065	0.0552	0.0325	0.2238	0.9642	0.9977	0.9990	0.1026	10.6925
1.6046	1.0022	5000	1.4270	0.2548	0.1001	0.2250	0.2251	0.0580	0.3366	0.1125	0.0599	0.0355	0.2324	0.9587	0.9979	0.9987	0.1079	10.8118
1.5787	1.2026	6000	1.4217	0.2592	0.1029	0.2304	0.2302	0.0613	0.3428	0.1180	0.0640	0.0389	0.2353	0.9664	0.9980	0.9977	0.1099	10.7235
1.5661	1.4031	7000	1.4012	0.2724	0.1124	0.2425	0.2423	0.0648	0.3513	0.1251	0.0686	0.0418	0.2476	0.9637	0.9977	0.9965	0.1171	10.7619
1.5572	1.6035	8000	1.3973	0.2738	0.1133	0.2428	0.2426	0.0668	0.3476	0.1269	0.0697	0.0426	0.2463	0.9553	0.9968	0.9963	0.1181	11.0247
1.5365	1.8040	9000	1.3891	0.2787	0.1192	0.2488	0.2487	0.0684	0.3534	0.1326	0.0740	0.0462	0.2507	0.9590	0.9972	0.9932	0.1222	10.8450
1.5298	2.0044	10000	1.3805	0.2767	0.1180	0.2477	0.2476	0.0678	0.3587	0.1336	0.0745	0.0464	0.2510	0.9628	0.9976	0.9935	0.1217	10.5836
1.5201	2.2049	11000	1.3786	0.2823	0.1213	0.2524	0.2522	0.0708	0.3578	0.1350	0.0755	0.0474	0.2557	0.9614	0.9975	0.9953	0.1246	10.8934
1.5322	2.4053	12000	1.3780	0.2842	0.1235	0.2538	0.2536	0.0705	0.3597	0.1372	0.0774	0.0490	0.2553	0.9604	0.9974	0.9940	0.1256	10.7393
1.5279	2.6057	13000	1.3695	0.2834	0.1228	0.2533	0.2532	0.0715	0.3591	0.1364	0.0779	0.0494	0.2564	0.9656	0.9980	0.9937	0.1259	10.8033
1.5146	2.8062	14000	1.3667	0.2848	0.1241	0.2545	0.2543	0.0720	0.3624	0.1387	0.0789	0.0503	0.2572	0.9633	0.9975	0.9930	0.1273	10.7279

Framework versions

—PEFT 0.15.2
—Transformers 4.52.4
—Pytorch 2.6.0+cu124
—Datasets 3.6.0
—Tokenizers 0.21.2