MSc_llama2_finetuned_model_secondData7

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: 0.6939

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
3.9823	1.33	10	3.6153
3.3928	2.67	20	2.9413
2.6305	4.0	30	2.1743
1.9546	5.33	40	1.7079
1.5996	6.67	50	1.4500
1.2984	8.0	60	1.1277
0.9632	9.33	70	0.8761
0.8296	10.67	80	0.8206
0.7589	12.0	90	0.7735
0.7063	13.33	100	0.7446
0.671	14.67	110	0.7278
0.6405	16.0	120	0.7091
0.6096	17.33	130	0.7021
0.5845	18.67	140	0.6986
0.5697	20.0	150	0.6938
0.5539	21.33	160	0.6936
0.5414	22.67	170	0.6913
0.5313	24.0	180	0.6920
0.522	25.33	190	0.6919
0.5168	26.67	200	0.6932
0.5191	28.0	210	0.6942
0.5079	29.33	220	0.6938
0.5132	30.67	230	0.6939
0.5085	32.0	240	0.6939
0.5079	33.33	250	0.6939

Framework versions

• PEFT 0.4.0
• Transformers 4.38.2
• Pytorch 2.3.1+cu121
• Datasets 2.13.1
• Tokenizers 0.15.2