MSc_llama2_finetuned_model_secondData8

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: 0.6809

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
3.9839	1.33	10	3.6565
3.3873	2.67	20	2.9922
2.6379	4.0	30	2.2276
1.9614	5.33	40	1.7433
1.6082	6.67	50	1.4612
1.3139	8.0	60	1.1369
0.9676	9.33	70	0.8804
0.8286	10.67	80	0.8272
0.7654	12.0	90	0.7804
0.7181	13.33	100	0.7474
0.6712	14.67	110	0.7286
0.6433	16.0	120	0.7128
0.6169	17.33	130	0.6997
0.5964	18.67	140	0.6971
0.5806	20.0	150	0.6893
0.5669	21.33	160	0.6867
0.5472	22.67	170	0.6861
0.5538	24.0	180	0.6819
0.535	25.33	190	0.6816
0.5322	26.67	200	0.6825
0.5333	28.0	210	0.6818
0.5203	29.33	220	0.6813
0.5259	30.67	230	0.6812
0.5264	32.0	240	0.6812
0.5209	33.33	250	0.6809

Framework versions

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