CheritiA/distilbert-sentiment-5class

checkpoints

This model is a fine-tuned version of distilbert-base-uncased trained on processed Amazon review datasets from multiple categories. It is designed for sentiment classification across five classes. It achieves the following results on the evaluation set:

• Loss: 0.9056
• Accuracy: 0.6087
• F1 Macro: 0.6101
• F1 Weighted: 0.6085
• Precision Class 1: 0.6762
• Precision Class 2: 0.4900
• Precision Class 3: 0.5177
• Precision Class 4: 0.6173
• Precision Class 5: 0.7535
• Recall Class 1: 0.6781
• Recall Class 2: 0.5293
• Recall Class 3: 0.5049
• Recall Class 4: 0.5606
• Recall Class 5: 0.7781
• F1 Class 1: 0.6772
• F1 Class 2: 0.5089
• F1 Class 3: 0.5112
• F1 Class 4: 0.5876
• F1 Class 5: 0.7656

Model description

This model is a distilled, lightweight transformer optimized for multi-class sentiment analysis. It is capable of processing text efficiently while maintaining competitive performance. The model was fine-tuned on Amazon product reviews spanning different categories such as electronics and home goods... It can classify reviews into five sentiment levels, from highly negative to highly positive.

Intended uses & limitations

Automated sentiment classification for Amazon reviews or similar e-commerce datasets.

Analysis of customer feedback trends to inform product decisions.

Training and evaluation data

The model was trained on a processed subset of Amazon review datasets covering multiple product categories. Text data was cleaned, tokenized, and labeled into five sentiment classes. Evaluation was performed on a held-out test set representing the same categories to ensure balanced coverage.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 32
• eval_batch_size: 64
• 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: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 4
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1 Macro	F1 Weighted	Precision Class 1	Precision Class 2	Precision Class 3	Precision Class 4	Precision Class 5	Recall Class 1	Recall Class 2	Recall Class 3	Recall Class 4	Recall Class 5	F1 Class 1	F1 Class 2	F1 Class 3	F1 Class 4	F1 Class 5
1.0806	0.16	500	1.0584	0.5392	0.5333	0.5316	0.5485	0.4359	0.4347	0.5297	0.7385	0.7797	0.3506	0.4185	0.4576	0.6959	0.6440	0.3886	0.4264	0.4910	0.7166
0.9979	0.32	1000	0.9753	0.5727	0.5741	0.5726	0.7303	0.4461	0.4691	0.5915	0.7095	0.5433	0.5949	0.4908	0.4406	0.8016	0.6231	0.5099	0.4797	0.5050	0.7528
0.9773	0.48	1500	0.9464	0.5869	0.5832	0.5814	0.6321	0.5040	0.4675	0.5748	0.7381	0.7328	0.3800	0.5110	0.5416	0.7785	0.6787	0.4333	0.4883	0.5577	0.7578
0.9311	0.64	2000	0.9356	0.5876	0.5923	0.5909	0.7000	0.4546	0.5063	0.5627	0.8035	0.6272	0.6081	0.4236	0.6077	0.6754	0.6616	0.5203	0.4613	0.5843	0.7339
0.9332	0.8	2500	0.9305	0.5924	0.5832	0.5815	0.6279	0.4773	0.5509	0.5754	0.6969	0.7421	0.4949	0.3492	0.5432	0.8437	0.6802	0.4860	0.4274	0.5589	0.7633
0.9193	0.96	3000	0.9068	0.6037	0.6024	0.6008	0.6696	0.4849	0.5144	0.6016	0.7372	0.6956	0.5188	0.4514	0.5519	0.8097	0.6823	0.5013	0.4809	0.5757	0.7717
0.8735	1.12	3500	0.9177	0.6032	0.6029	0.6012	0.6766	0.5004	0.4973	0.5865	0.7445	0.6859	0.4791	0.4817	0.5848	0.7941	0.6812	0.4895	0.4894	0.5857	0.7685
0.8656	1.28	4000	0.9069	0.6042	0.6011	0.5996	0.6823	0.4792	0.5197	0.6217	0.7149	0.6830	0.5532	0.4448	0.4968	0.8526	0.6826	0.5136	0.4794	0.5523	0.7777
0.8667	1.44	4500	0.9034	0.6059	0.6051	0.6035	0.6859	0.4958	0.5050	0.5923	0.7384	0.6877	0.4941	0.4807	0.5640	0.8124	0.6868	0.4950	0.4925	0.5778	0.7736
0.879	1.6	5000	0.9001	0.6067	0.6086	0.6071	0.6592	0.4874	0.5043	0.6120	0.7918	0.7239	0.5192	0.4876	0.5703	0.7381	0.6901	0.5028	0.4958	0.5904	0.7640
0.8709	1.76	5500	0.9005	0.6039	0.6045	0.6029	0.7008	0.4802	0.5250	0.5652	0.7736	0.6591	0.5636	0.4078	0.6366	0.7603	0.6793	0.5186	0.4590	0.5988	0.7669
0.8565	1.92	6000	0.9076	0.6084	0.6053	0.6036	0.6409	0.5094	0.5011	0.6234	0.7513	0.7529	0.4305	0.5253	0.5368	0.8057	0.6924	0.4666	0.5129	0.5769	0.7776
0.794	2.08	6500	0.9080	0.6091	0.6103	0.6086	0.6960	0.4837	0.5061	0.6135	0.7610	0.6749	0.5568	0.4676	0.5521	0.8022	0.6853	0.5177	0.4861	0.5812	0.7810
0.7857	2.24	7000	0.9131	0.6058	0.6062	0.6046	0.6521	0.4940	0.5055	0.5987	0.7828	0.7342	0.4886	0.4795	0.5853	0.7482	0.6907	0.4913	0.4922	0.5919	0.7651
0.7956	2.4	7500	0.9092	0.6074	0.6105	0.6090	0.6897	0.4941	0.4919	0.6113	0.7813	0.6776	0.5317	0.5251	0.5475	0.7613	0.6836	0.5122	0.5080	0.5776	0.7712
0.7721	2.56	8000	0.9153	0.6057	0.6080	0.6065	0.6641	0.5053	0.4927	0.5936	0.7982	0.7196	0.4777	0.5280	0.5911	0.7177	0.6908	0.4911	0.5098	0.5924	0.7558
0.7974	2.7200	8500	0.9019	0.6095	0.6110	0.6094	0.6790	0.4909	0.5019	0.6212	0.7680	0.6981	0.5208	0.5058	0.5483	0.7819	0.6884	0.5054	0.5038	0.5825	0.7749
0.7884	2.88	9000	0.9119	0.6074	0.6054	0.6037	0.6467	0.5007	0.5033	0.6101	0.7592	0.7431	0.4593	0.4868	0.5608	0.7958	0.6915	0.4791	0.4949	0.5844	0.7771
0.7359	3.04	9500	0.9220	0.6066	0.6100	0.6085	0.6829	0.4916	0.4923	0.6094	0.7899	0.6889	0.5139	0.5286	0.5661	0.7414	0.6859	0.5025	0.5098	0.5869	0.7649
0.7453	3.2	10000	0.9298	0.6092	0.6064	0.6047	0.6534	0.4958	0.5115	0.6233	0.7431	0.7369	0.4930	0.4696	0.5315	0.8238	0.6927	0.4944	0.4897	0.5738	0.7814

Framework versions

• Transformers 4.57.3
• Pytorch 2.9.0+cu126
• Datasets 4.0.0
• Tokenizers 0.22.1