bhasini_hindi_llm

This model is a fine-tuned version of unsloth/tinyllama-chat-bnb-4bit on the None dataset. It achieves the following results on the evaluation set:

• Loss: 10.1097

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: 0.0003
• 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
• lr_scheduler_warmup_steps: 2
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
16.0558	0.2	2	10.1148
16.0447	0.4	4	10.1148
16.0365	0.6	6	10.1148
16.0966	0.8	8	10.1148
15.9258	1.0	10	10.1148
15.9849	1.2	12	10.1148
16.0978	1.4	14	10.1148
16.0488	1.6	16	10.1149
9.6559	1.8	18	10.1151
10.152	2.0	20	10.1152
10.0189	2.2	22	10.1152
9.9467	2.4	24	10.1153
8.1757	2.6	26	10.1155
6.956	2.8	28	10.1155
6.3082	3.0	30	10.1156
5.9322	3.2	32	10.1156
5.5365	3.4	34	10.1155
5.5109	3.6	36	10.1155
5.41	3.8	38	10.1154
5.3045	4.0	40	10.1154
5.2227	4.2	42	10.1153
5.2854	4.4	44	10.1151
5.0931	4.6	46	10.1148
4.9491	4.8	48	10.1144
4.9003	5.0	50	10.1142
4.7583	5.2	52	10.1140
4.7297	5.4	54	10.1136
4.6986	5.6	56	10.1131
4.678	5.8	58	10.1127
4.6374	6.0	60	10.1123
4.6267	6.2	62	10.1119
4.6493	6.4	64	10.1115
4.5741	6.6	66	10.1112
4.6336	6.8	68	10.1109
4.6337	7.0	70	10.1108
4.5785	7.2	72	10.1107
4.6021	7.4	74	10.1106
4.6085	7.6	76	10.1105
4.5954	7.8	78	10.1105
4.5919	8.0	80	10.1104
4.5622	8.2	82	10.1104
4.4947	8.4	84	10.1103
4.5473	8.6	86	10.1102
4.5789	8.8	88	10.1101
4.5734	9.0	90	10.1100
4.4943	9.2	92	10.1099
4.5666	9.4	94	10.1099
4.5599	9.6	96	10.1098
4.5575	9.8	98	10.1097
4.5705	10.0	100	10.1097

Framework versions

• PEFT 0.14.0
• Transformers 4.48.3
• Pytorch 2.6.0+cu124
• Datasets 3.3.1
• Tokenizers 0.21.0