MSc_llama2_finetuned_model_secondData10

This model is a fine-tuned version of meta-llama/Llama-2-7b-chat-hf on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.7118

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

The following bitsandbytes quantization config was used during training:

• quant_method: bitsandbytes
• _load_in_8bit: False
• _load_in_4bit: True
• llm_int8_threshold: 6.0
• llm_int8_skip_modules: None
• llm_int8_enable_fp32_cpu_offload: False
• llm_int8_has_fp16_weight: False
• bnb_4bit_quant_type: nf4
• bnb_4bit_use_double_quant: True
• bnb_4bit_compute_dtype: bfloat16
• load_in_4bit: True
• load_in_8bit: False

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 3e-05
• train_batch_size: 32
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.03
• training_steps: 250

Training results

Training Loss	Epoch	Step	Validation Loss
3.8835	1.33	10	3.4352
2.9529	2.67	20	2.3780
1.991	4.0	30	1.6911
1.5061	5.33	40	1.2670
1.0666	6.67	50	0.8670
0.8464	8.0	60	0.8088
0.7622	9.33	70	0.7478
0.6869	10.67	80	0.7055
0.6336	12.0	90	0.6840
0.5789	13.33	100	0.6749
0.5518	14.67	110	0.6685
0.5159	16.0	120	0.6657
0.4894	17.33	130	0.6743
0.4674	18.67	140	0.6720
0.4496	20.0	150	0.6806
0.4292	21.33	160	0.6883
0.421	22.67	170	0.6910
0.4088	24.0	180	0.6956
0.3988	25.33	190	0.7014
0.3898	26.67	200	0.7065
0.3827	28.0	210	0.7091
0.3819	29.33	220	0.7104
0.3778	30.67	230	0.7117
0.3803	32.0	240	0.7126
0.3804	33.33	250	0.7118

Framework versions

• PEFT 0.4.0
• Transformers 4.38.2
• Pytorch 2.4.0+cu121
• Datasets 2.13.1
• Tokenizers 0.15.2