camembert-pcg-annotation-full

This model is a fine-tuned version of camembert-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.9741
• Accuracy: 0.5529
• Top3 Accuracy: 0.7549
• Top5 Accuracy: 0.8285
• Precision: 0.5827
• Recall: 0.5529
• F1 Weighted: 0.5500
• F1 Macro: 0.3164

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: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• 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_steps: 0.1
• num_epochs: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Top3 Accuracy	Top5 Accuracy	Precision	Recall	F1 Weighted	F1 Macro
10.4836	0.2000	5685	5.0665	0.1127	0.2548	0.3753	0.0669	0.1127	0.0599	0.0173
8.4373	0.4000	11370	4.0622	0.3031	0.4722	0.5786	0.2701	0.3031	0.2376	0.0689
7.1056	0.5999	17055	3.5299	0.3624	0.5438	0.6517	0.3858	0.3624	0.3307	0.1013
6.4190	0.7999	22740	3.1918	0.3942	0.5826	0.6804	0.4443	0.3942	0.3790	0.1313
6.0150	0.9999	28425	2.9816	0.4022	0.6066	0.7010	0.4656	0.4022	0.3937	0.1473
5.5633	1.1999	34110	2.7845	0.4255	0.6289	0.7215	0.4763	0.4255	0.4210	0.1830
5.3629	1.3999	39795	2.6584	0.4417	0.6436	0.7366	0.4848	0.4417	0.4327	0.1989
5.1547	1.5998	45480	2.5266	0.4505	0.6605	0.7506	0.5006	0.4505	0.4441	0.2065
4.8821	1.7998	51165	2.4337	0.4646	0.6726	0.7618	0.5107	0.4646	0.4597	0.2182
4.8468	1.9998	56850	2.3580	0.4751	0.6825	0.7700	0.5255	0.4751	0.4705	0.2366
4.5083	2.1998	62535	2.3354	0.4872	0.6988	0.7844	0.5255	0.4872	0.4839	0.2488
4.3747	2.3998	68220	2.2397	0.4968	0.7037	0.7875	0.5416	0.4968	0.4955	0.2542
4.3334	2.5997	73905	2.2009	0.5124	0.7196	0.7999	0.5468	0.5124	0.5076	0.2679
4.3103	2.7997	79590	2.1654	0.5165	0.7254	0.8035	0.5578	0.5165	0.5118	0.2719
4.1306	2.9997	85275	2.1243	0.529	0.7327	0.8120	0.5584	0.529	0.5244	0.2871
3.8964	3.1997	90960	2.0769	0.5324	0.7378	0.8155	0.5664	0.5324	0.5295	0.3021
3.9158	3.3996	96645	2.0659	0.5306	0.7429	0.8186	0.5684	0.5306	0.5301	0.2978
3.7772	3.5996	102330	2.0397	0.5421	0.7443	0.8209	0.5725	0.5421	0.5381	0.3020
3.8012	3.7996	108015	2.0179	0.5422	0.7487	0.8231	0.5769	0.5422	0.5408	0.3028
3.7211	3.9996	113700	2.0004	0.5481	0.7525	0.8265	0.5791	0.5481	0.5448	0.3097
3.5257	4.1996	119385	1.9928	0.5482	0.7527	0.8268	0.5821	0.5482	0.5466	0.3095
3.5389	4.3995	125070	1.9911	0.5499	0.7541	0.8279	0.5807	0.5499	0.5470	0.3101
3.6145	4.5995	130755	1.9784	0.5515	0.7541	0.828	0.5811	0.5515	0.5488	0.3141
3.6517	4.7995	136440	1.9743	0.5529	0.7549	0.8287	0.5826	0.5529	0.5500	0.3170
3.5229	4.9995	142125	1.9741	0.5529	0.7549	0.8285	0.5827	0.5529	0.5500	0.3164

Framework versions

• Transformers 5.2.0
• Pytorch 2.10.0+cu128
• Datasets 4.5.0
• Tokenizers 0.22.2