git-large-r-coco-IDB_ADv1_COCOv6-r

This model is a fine-tuned version of microsoft/git-large-r-coco on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.0568
• Meteor Score: {'meteor': 0.7213608939819159}

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: 3e-05
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 15
• num_epochs: 150
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Meteor Score
91.9877	1.4815	5	11.4608	{'meteor': 0.046723581008416726}
84.6168	2.9630	10	9.8170	{'meteor': 0.043886991456607405}
77.4092	4.4444	15	9.1821	{'meteor': 0.06849055586443556}
71.3318	5.9259	20	8.3193	{'meteor': 0.061564918769236095}
65.2006	7.4074	25	7.6103	{'meteor': 0.0754162033563066}
60.1084	8.8889	30	7.0968	{'meteor': 0.1078676449812926}
56.3449	10.3704	35	6.6624	{'meteor': 0.16409628582839803}
53.0194	11.8519	40	6.2707	{'meteor': 0.17640875038111917}
49.885	13.3333	45	5.9076	{'meteor': 0.20852663016362413}
47.1535	14.8148	50	5.5612	{'meteor': 0.21135425345585648}
44.3251	16.2963	55	5.2211	{'meteor': 0.25163567787871793}
41.4631	17.7778	60	4.8845	{'meteor': 0.36102118410159373}
38.8862	19.2593	65	4.5516	{'meteor': 0.40677731182809435}
36.2603	20.7407	70	4.2228	{'meteor': 0.42841296399868456}
33.5914	22.2222	75	3.8960	{'meteor': 0.4442706614076698}
31.0193	23.7037	80	3.5716	{'meteor': 0.44790686728484286}
28.4252	25.1852	85	3.2513	{'meteor': 0.4634788207419503}
25.927	26.6667	90	2.9379	{'meteor': 0.46009853575600396}
23.4406	28.1481	95	2.6312	{'meteor': 0.4702576268818571}
21.0641	29.6296	100	2.3319	{'meteor': 0.4724561426359612}
18.6236	31.1111	105	2.0445	{'meteor': 0.4759864205500001}
16.4033	32.5926	110	1.7706	{'meteor': 0.4905426072931855}
14.1917	34.0741	115	1.5119	{'meteor': 0.48899836960829424}
12.1305	35.5556	120	1.2729	{'meteor': 0.5241780827869041}
10.2674	37.0370	125	1.0567	{'meteor': 0.5097833762351499}
8.5593	38.5185	130	0.8637	{'meteor': 0.4997929450485001}
7.013	40.0	135	0.6974	{'meteor': 0.5105554674131357}
5.666	41.4815	140	0.5584	{'meteor': 0.5152670866288012}
4.5962	42.9630	145	0.4444	{'meteor': 0.5396604542862918}
3.6125	44.4444	150	0.3520	{'meteor': 0.5617093146804203}
2.9293	45.9259	155	0.2796	{'meteor': 0.5886004681488384}
2.259	47.4074	160	0.2219	{'meteor': 0.553105842439049}
1.8309	48.8889	165	0.1785	{'meteor': 0.5814978557330325}
1.4494	50.3704	170	0.1458	{'meteor': 0.6231566268972263}
1.1158	51.8519	175	0.1248	{'meteor': 0.6099620120504717}
0.9661	53.3333	180	0.1060	{'meteor': 0.6546451210304227}
0.7271	54.8148	185	0.0880	{'meteor': 0.6338177371529071}
0.6177	56.2963	190	0.0775	{'meteor': 0.6518057184133086}
0.4962	57.7778	195	0.0698	{'meteor': 0.6193332334649128}
0.4415	59.2593	200	0.0652	{'meteor': 0.7079004093008636}
0.3629	60.7407	205	0.0632	{'meteor': 0.6940804620155375}
0.3219	62.2222	210	0.0595	{'meteor': 0.6912091264820688}
0.2679	63.7037	215	0.0545	{'meteor': 0.7030913652832623}
0.234	65.1852	220	0.0537	{'meteor': 0.6676437817167338}
0.2026	66.6667	225	0.0534	{'meteor': 0.6811270764539213}
0.1843	68.1481	230	0.0515	{'meteor': 0.6779206905106651}
0.1662	69.6296	235	0.0511	{'meteor': 0.715690856984691}
0.1432	71.1111	240	0.0502	{'meteor': 0.7156957872649886}
0.1331	72.5926	245	0.0495	{'meteor': 0.6543089583737782}
0.1156	74.0741	250	0.0494	{'meteor': 0.6594307549550437}
0.1021	75.5556	255	0.0485	{'meteor': 0.6576803078677578}
0.092	77.0370	260	0.0487	{'meteor': 0.6811188428011016}
0.0809	78.5185	265	0.0498	{'meteor': 0.7035441241921582}
0.0723	80.0	270	0.0506	{'meteor': 0.7038905922785134}
0.0649	81.4815	275	0.0502	{'meteor': 0.7421547726024114}
0.0608	82.9630	280	0.0507	{'meteor': 0.730981066765409}
0.0491	84.4444	285	0.0518	{'meteor': 0.7443890531705978}
0.0458	85.9259	290	0.0534	{'meteor': 0.7176754184986865}
0.0388	87.4074	295	0.0523	{'meteor': 0.7176410382626139}
0.0334	88.8889	300	0.0534	{'meteor': 0.7026151957465668}
0.026	90.3704	305	0.0520	{'meteor': 0.7362294347533668}
0.0241	91.8519	310	0.0531	{'meteor': 0.7236102999038347}
0.0211	93.3333	315	0.0542	{'meteor': 0.7152270797326922}
0.0182	94.8148	320	0.0547	{'meteor': 0.7175103144573044}
0.0175	96.2963	325	0.0547	{'meteor': 0.7140333983909092}
0.0158	97.7778	330	0.0558	{'meteor': 0.7230279458318826}
0.0153	99.2593	335	0.0560	{'meteor': 0.7216506562988318}
0.0133	100.7407	340	0.0558	{'meteor': 0.7089953285495003}
0.0137	102.2222	345	0.0557	{'meteor': 0.7050918755935003}
0.0127	103.7037	350	0.0561	{'meteor': 0.7140423279140877}
0.0118	105.1852	355	0.0563	{'meteor': 0.716693515553002}
0.0121	106.6667	360	0.0559	{'meteor': 0.7213321747823379}
0.0115	108.1481	365	0.0559	{'meteor': 0.7244157912903824}
0.0107	109.6296	370	0.0559	{'meteor': 0.7258048026400388}
0.0115	111.1111	375	0.0562	{'meteor': 0.7249978244263323}
0.0107	112.5926	380	0.0564	{'meteor': 0.7261256725269803}
0.0103	114.0741	385	0.0565	{'meteor': 0.7222910388476055}
0.0102	115.5556	390	0.0567	{'meteor': 0.7209011950646473}
0.0099	117.0370	395	0.0567	{'meteor': 0.7221559565497234}
0.0101	118.5185	400	0.0567	{'meteor': 0.7234189947956161}
0.0096	120.0	405	0.0567	{'meteor': 0.7231619412067093}
0.0097	121.4815	410	0.0567	{'meteor': 0.7224063072061223}
0.0103	122.9630	415	0.0567	{'meteor': 0.7209850209026843}
0.0101	124.4444	420	0.0567	{'meteor': 0.7208576227726795}
0.0095	125.9259	425	0.0567	{'meteor': 0.7215621466948617}
0.0093	127.4074	430	0.0568	{'meteor': 0.7221874032851583}
0.01	128.8889	435	0.0568	{'meteor': 0.7207488272286875}
0.0097	130.3704	440	0.0568	{'meteor': 0.7213694074696705}
0.0096	131.8519	445	0.0568	{'meteor': 0.7213044564985146}
0.0099	133.3333	450	0.0568	{'meteor': 0.7213608939819159}

Framework versions

• Transformers 4.46.1
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.20.2