dfm_indirect_speech

This model is a fine-tuned version of KennethEnevoldsen/dfm-sentence-encoder-large-exp2-no-lang-align on an unknown dataset. It achieves the following results on the evaluation set:

• Accuracy: 0.9158
• Precision: 0.9150
• Recall: 0.9158
• F1: 0.9099
• Loss: 0.6465

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Accuracy	Precision	Recall	F1	Validation Loss
No log	1.0	13	0.8602	0.8484	0.8602	0.8492	0.4979
No log	2.0	26	0.8809	0.8636	0.8809	0.8718	0.4677
No log	3.0	39	0.9062	0.8906	0.9062	0.8973	0.3820
No log	4.0	52	0.9290	0.9110	0.9290	0.9194	0.4183
No log	5.0	65	0.9018	0.8924	0.9018	0.8958	0.3585
No log	6.0	78	0.8939	0.8984	0.8939	0.8855	0.5402
No log	7.0	91	0.9171	0.9105	0.9171	0.9131	0.4350
No log	8.0	104	0.9255	0.9198	0.9255	0.9196	0.4858
No log	9.0	117	0.9242	0.9259	0.9242	0.9161	0.5378
No log	10.0	130	0.9177	0.9153	0.9177	0.9114	0.5865
No log	11.0	143	0.9181	0.9169	0.9181	0.9120	0.6122
No log	12.0	156	0.9163	0.9154	0.9163	0.9102	0.6229
No log	13.0	169	0.9165	0.9146	0.9165	0.9106	0.6222
No log	14.0	182	0.9154	0.9127	0.9154	0.9096	0.6227
No log	15.0	195	0.9146	0.9141	0.9146	0.9088	0.6326
No log	16.0	208	0.9154	0.9146	0.9154	0.9095	0.6406
No log	17.0	221	0.9154	0.9146	0.9154	0.9095	0.6462
No log	18.0	234	0.9158	0.9150	0.9158	0.9099	0.6458
No log	18.48	240	0.9158	0.9150	0.9158	0.9099	0.6465

Framework versions

• Transformers 4.48.2
• Pytorch 2.5.1+cu124
• Tokenizers 0.21.0