llama3_question

This model was trained from scratch on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.8999

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: False
• bnb_4bit_compute_dtype: float16

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0002
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant
• lr_scheduler_warmup_ratio: 0.03
• num_epochs: 6

Training results

Training Loss	Epoch	Step	Validation Loss
2.9948	0.14	1	2.8184
2.8697	0.29	2	2.6592
2.6264	0.43	3	2.4946
2.625	0.57	4	2.3588
2.3888	0.71	5	2.2385
2.2949	0.86	6	2.1219
2.5261	1.0	7	2.0221
2.0264	1.14	8	1.9246
1.9661	1.29	9	1.8298
1.9106	1.43	10	1.7456
1.8448	1.57	11	1.6686
1.619	1.71	12	1.6050
1.5881	1.86	13	1.5468
1.6859	2.0	14	1.4939
1.4643	2.14	15	1.4453
1.4583	2.29	16	1.3949
1.4086	2.43	17	1.3441
1.3314	2.57	18	1.2914
1.3502	2.71	19	1.2400
1.226	2.86	20	1.1892
1.073	3.0	21	1.1445
1.1113	3.14	22	1.0995
1.1292	3.29	23	1.0570
1.0242	3.43	24	1.0164
0.9279	3.57	25	0.9826
0.8518	3.71	26	0.9617
1.0302	3.86	27	0.9491
1.1736	4.0	28	0.9418
0.8832	4.14	29	0.9352
0.9151	4.29	30	0.9301
0.7495	4.43	31	0.9256
0.8785	4.57	32	0.9220
0.8635	4.71	33	0.9180
0.9499	4.86	34	0.9150
0.8744	5.0	35	0.9125
0.8221	5.14	36	0.9093
0.7826	5.29	37	0.9064
0.8421	5.43	38	0.9047
0.8155	5.57	39	0.9029
0.9097	5.71	40	0.9010
0.7449	5.86	41	0.9003
0.9502	6.0	42	0.8999

Framework versions

• PEFT 0.5.0
• Transformers 4.37.2
• Pytorch 2.1.2
• Datasets 2.18.0
• Tokenizers 0.15.1