67f78e5609d10f79d93334947eb4ea81

This model is a fine-tuned version of google/mt5-small on the Helsinki-NLP/opus_books [fr-nl] dataset. It achieves the following results on the evaluation set:

• Loss: 1.9072
• Data Size: 1.0
• Epoch Runtime: 146.8319
• Bleu: 7.6933

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: 5e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 32
• total_eval_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: constant
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Data Size	Epoch Runtime	Bleu
No log	0	0	25.0059	0	12.7692	0.0048
No log	1	1000	21.2792	0.0078	13.7547	0.0046
No log	2	2000	15.9712	0.0156	15.1095	0.0047
No log	3	3000	12.4554	0.0312	17.7115	0.0063
0.5906	4	4000	7.8840	0.0625	22.7222	0.0151
7.1888	5	5000	4.8394	0.125	31.0653	0.0323
0.2984	6	6000	3.5904	0.25	47.6058	0.8191
0.3735	7	7000	3.2006	0.5	81.2562	1.8199
3.6568	8.0	8000	2.9083	1.0	146.8896	2.8294
3.4029	9.0	9000	2.7560	1.0	146.8793	3.3878
3.2491	10.0	10000	2.6562	1.0	144.8004	3.7127
3.1471	11.0	11000	2.5820	1.0	143.9108	4.0387
3.0497	12.0	12000	2.5156	1.0	143.9953	4.2739
2.9515	13.0	13000	2.4650	1.0	143.0167	4.5165
2.8894	14.0	14000	2.4179	1.0	143.5409	4.6757
2.8164	15.0	15000	2.3775	1.0	146.5068	4.8487
2.7565	16.0	16000	2.3439	1.0	142.7527	5.0430
2.668	17.0	17000	2.3097	1.0	143.3021	5.2109
2.6807	18.0	18000	2.2790	1.0	141.4281	5.3746
2.6175	19.0	19000	2.2513	1.0	140.5433	5.5266
2.5615	20.0	20000	2.2301	1.0	142.2420	5.6386
2.5557	21.0	21000	2.2066	1.0	141.5709	5.7679
2.4982	22.0	22000	2.1906	1.0	141.4199	5.8798
2.4646	23.0	23000	2.1690	1.0	142.7218	5.9994
2.4235	24.0	24000	2.1503	1.0	144.2434	6.0590
2.448	25.0	25000	2.1291	1.0	142.9854	6.1774
2.3732	26.0	26000	2.1164	1.0	142.8238	6.2716
2.3499	27.0	27000	2.1017	1.0	142.3916	6.3534
2.3031	28.0	28000	2.0889	1.0	142.5359	6.4245
2.3186	29.0	29000	2.0757	1.0	142.3031	6.5032
2.2609	30.0	30000	2.0675	1.0	143.6074	6.5912
2.2772	31.0	31000	2.0480	1.0	143.4460	6.6842
2.2382	32.0	32000	2.0414	1.0	142.8258	6.7297
2.1964	33.0	33000	2.0333	1.0	142.5789	6.8041
2.1787	34.0	34000	2.0263	1.0	141.3301	6.8929
2.1484	35.0	35000	2.0118	1.0	140.3920	6.9442
2.1415	36.0	36000	1.9954	1.0	141.8709	7.0288
2.1151	37.0	37000	1.9950	1.0	142.0263	7.0444
2.1257	38.0	38000	1.9811	1.0	141.7408	7.1191
2.0678	39.0	39000	1.9765	1.0	141.5414	7.1762
2.0599	40.0	40000	1.9726	1.0	142.8618	7.2139
2.0554	41.0	41000	1.9567	1.0	142.9126	7.2727
2.0399	42.0	42000	1.9563	1.0	142.8270	7.3510
2.0209	43.0	43000	1.9488	1.0	142.8891	7.3960
2.0283	44.0	44000	1.9399	1.0	140.6617	7.4480
1.9838	45.0	45000	1.9342	1.0	145.7035	7.4922
1.9968	46.0	46000	1.9319	1.0	142.4968	7.5003
1.9656	47.0	47000	1.9235	1.0	141.3154	7.5467
1.9808	48.0	48000	1.9177	1.0	142.4049	7.5940
1.9234	49.0	49000	1.9177	1.0	142.8916	7.6179
1.9233	50.0	50000	1.9072	1.0	146.8319	7.6933

Framework versions

• Transformers 4.57.0
• Pytorch 2.8.0+cu128
• Datasets 4.2.0
• Tokenizers 0.22.1