mistral-7b-instruct-v0.3-mimic4-adapt-l2r-multilabel-plantclassify

This model is a fine-tuned version of mistralai/Mistral-7B-Instruct-v0.3 on the None dataset. It achieves the following results on the evaluation set:

• F1 Micro: 0.0173
• F1 Macro: 0.0083
• Precision At 5: 0.4191
• Recall At 5: 0.0999
• Precision At 8: 0.3571
• Recall At 8: 0.1335
• Precision At 15: 0.2744
• Recall At 15: 0.1864
• Rare F1 Micro: 0.0097
• Rare F1 Macro: 0.0063
• Rare Precision: 0.0049
• Rare Recall: 0.6434
• Rare Precision At 5: 0.1862
• Rare Recall At 5: 0.0670
• Rare Precision At 8: 0.1513
• Rare Recall At 8: 0.0868
• Rare Precision At 15: 0.1107
• Rare Recall At 15: 0.1172
• Not Rare F1 Micro: 0.1361
• Not Rare F1 Macro: 0.1312
• Not Rare Precision: 0.0730
• Not Rare Recall: 0.9997
• Not Rare Precision At 5: 0.4268
• Not Rare Recall At 5: 0.2761
• Not Rare Precision At 8: 0.3573
• Not Rare Recall At 8: 0.3544
• Not Rare Precision At 15: 0.2682
• Not Rare Recall At 15: 0.4784
• Loss: -2.5098

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
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• 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: linear
• lr_scheduler_warmup_steps: 50
• num_epochs: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	F1 Micro	F1 Macro	Precision At 5	Recall At 5	Precision At 8	Recall At 8	Precision At 15	Recall At 15	Rare F1 Micro	Rare F1 Macro	Rare Precision	Rare Recall	Rare Precision At 5	Rare Recall At 5	Rare Precision At 8	Rare Recall At 8	Rare Precision At 15	Rare Recall At 15	Not Rare F1 Micro	Not Rare F1 Macro	Not Rare Precision	Not Rare Recall	Not Rare Precision At 5	Not Rare Recall At 5	Not Rare Precision At 8	Not Rare Recall At 8	Not Rare Precision At 15	Not Rare Recall At 15	Validation Loss
-2.7793	0.9981	262	0.0114	0.0060	0.3790	0.0897	0.3251	0.1194	0.2546	0.1696	0.0065	0.0040	0.0033	0.7152	0.1317	0.0472	0.1068	0.0602	0.0780	0.0804	0.1354	0.1308	0.0726	1.0	0.3727	0.2367	0.3162	0.3110	0.2446	0.4322	-2.3806
-3.0203	1.9981	524	0.0143	0.0070	0.4135	0.0996	0.3589	0.1343	0.2777	0.1888	0.0082	0.0050	0.0041	0.6913	0.1803	0.0641	0.1481	0.0829	0.1086	0.1116	0.1356	0.1309	0.0728	0.9999	0.3987	0.2581	0.3407	0.3388	0.2602	0.4678	-2.4964
-3.2253	2.9981	786	0.0122	0.0069	0.4385	0.1062	0.3724	0.1403	0.2868	0.1941	0.0071	0.0050	0.0036	0.7429	0.1983	0.0713	0.1572	0.0898	0.1142	0.1200	0.1355	0.1308	0.0727	1.0	0.4206	0.2727	0.3544	0.3536	0.2707	0.4839	-2.5382
-3.5758	3.9981	1048	0.0166	0.0082	0.4336	0.1040	0.3683	0.1379	0.2819	0.1919	0.0094	0.0062	0.0047	0.6661	0.1956	0.0704	0.1583	0.0905	0.1155	0.1224	0.1358	0.1310	0.0728	0.9999	0.4260	0.2760	0.3564	0.3548	0.2686	0.4780	-2.5326
-3.9755	4.9981	1310	0.0173	0.0083	0.4191	0.0999	0.3571	0.1335	0.2744	0.1864	0.0097	0.0063	0.0049	0.6434	0.1862	0.0670	0.1513	0.0868	0.1107	0.1172	0.1361	0.1312	0.0730	0.9997	0.4268	0.2761	0.3573	0.3544	0.2682	0.4784	-2.5098

Framework versions

• Transformers 4.49.0
• Pytorch 2.6.0
• Datasets 3.6.0
• Tokenizers 0.21.1