davanstrien/rt-detr-newspaper-photo-detection

rt-detr-britain-handbooks-photos

This model is a fine-tuned version of PekingU/rtdetr_r50vd on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 4.9066

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: 4
• eval_batch_size: 4
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 90
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
No log	1.0	9	370.5898
542.4309	2.0	18	371.3819
522.3717	3.0	27	337.3348
459.2125	4.0	36	287.6073
395.5238	5.0	45	230.7955
286.8508	6.0	54	176.7677
228.066	7.0	63	136.6897
161.5375	8.0	72	99.1186
117.6188	9.0	81	78.7769
91.4598	10.0	90	60.9189
91.4598	11.0	99	48.0892
74.3667	12.0	108	41.1282
62.515	13.0	117	34.5201
52.9375	14.0	126	32.2538
44.9663	15.0	135	29.8093
40.1683	16.0	144	26.0296
36.8552	17.0	153	25.6356
31.1489	18.0	162	20.8665
28.862	19.0	171	18.5289
26.1241	20.0	180	16.8950
26.1241	21.0	189	16.1161
23.8333	22.0	198	16.3788
20.779	23.0	207	14.8389
19.1705	24.0	216	13.2360
18.1938	25.0	225	12.2007
16.0306	26.0	234	11.2442
15.0287	27.0	243	10.4789
14.8518	28.0	252	9.8145
13.2291	29.0	261	9.2670
12.8187	30.0	270	9.0118
12.8187	31.0	279	8.4731
11.8135	32.0	288	7.8816
11.8473	33.0	297	7.9214
10.3618	34.0	306	7.2866
11.291	35.0	315	6.7351
9.5297	36.0	324	6.5649
9.1887	37.0	333	6.3402
9.3662	38.0	342	6.8995
8.8807	39.0	351	6.7314
8.871	40.0	360	6.3205
8.871	41.0	369	6.2681
8.4285	42.0	378	6.1505
8.0272	43.0	387	5.7252
7.7106	44.0	396	5.7598
7.6973	45.0	405	5.8607
7.4816	46.0	414	5.8354
7.7731	47.0	423	6.1469
7.3577	48.0	432	5.7994
6.9262	49.0	441	5.7957
6.7817	50.0	450	5.3749
6.7817	51.0	459	5.4785
6.4421	52.0	468	5.5997
6.6866	53.0	477	5.6428
6.2332	54.0	486	5.2460
6.033	55.0	495	5.0726
6.4574	56.0	504	5.0716
5.8004	57.0	513	5.1538
6.3994	58.0	522	5.2739
5.4569	59.0	531	4.9320
6.8395	60.0	540	4.8320
6.8395	61.0	549	4.8788
5.6847	62.0	558	5.0752
6.0732	63.0	567	5.1419
5.5805	64.0	576	5.0951
5.5178	65.0	585	5.1321
5.1452	66.0	594	4.9179
5.4156	67.0	603	4.9118
5.5653	68.0	612	4.9279
5.2236	69.0	621	4.8551
5.3366	70.0	630	5.1762
5.3366	71.0	639	4.9683
5.2791	72.0	648	5.0238
5.5227	73.0	657	4.9368
4.9659	74.0	666	4.9490
5.1077	75.0	675	4.9549
5.1386	76.0	684	4.9591
5.2915	77.0	693	4.9849
4.9752	78.0	702	5.1533
5.2248	79.0	711	4.9699
5.2362	80.0	720	5.0121
5.2362	81.0	729	4.9954
5.1962	82.0	738	5.0266
4.9936	83.0	747	4.9814
4.9174	84.0	756	5.0074
5.1872	85.0	765	5.2822
5.3053	86.0	774	4.9256
4.7016	87.0	783	4.9502
5.1849	88.0	792	5.0745
5.2689	89.0	801	5.0272
4.9275	90.0	810	4.9066

Framework versions

• Transformers 4.57.1
• Pytorch 2.9.1+cu128
• Datasets 4.4.1
• Tokenizers 0.22.1