base_model: meta-llama/Llama-3.1-8B
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

load_in_8bit: false
load_in_4bit: true
strict: false

datasets:
  - path: ahmedelgebaly/SQuad_2_Alpaca
    type: alpaca
    split: train

test_datasets:
  - path: ahmedelgebaly/SQuad_2_Alpaca
    type: alpaca
    split: validation

dataset_prepared_path:
output_dir: ./outputs/qlora-out

adapter: qlora
lora_model_dir:

sequence_len: 4096
sample_packing: true
pad_to_sequence_len: true

lora_r: 32
lora_alpha: 64 #Before it was 16
lora_dropout: 0.05
lora_target_modules: #Before it was empty
  - q_proj
  - k_proj
  - v_proj
  - o_proj
  - gate_proj
  - up_proj
  - down_proj
lora_target_linear: true

wandb_project: llama-3.1-8b-squadv2_E3_V2
wandb_entity:
wandb_watch:
wandb_name: llama-3.1-8b-squadv2_E3_V2
wandb_log_model:

hub_model_id: ahmedelgebaly/llama-3.1-8b-squadv2_E3_V2

gradient_accumulation_steps: 4
micro_batch_size: 4
num_epochs: 3

optimizer: paged_adamw_32bit
lr_scheduler: cosine
learning_rate: 0.0002

train_on_inputs: false
group_by_length: true #Before it was false

bf16: auto
tf32: false

gradient_checkpointing: true
flash_attention: true

warmup_steps: 50 #Before it was 10
evals_per_epoch: 4
saves_per_epoch: 1

weight_decay: 0.0

special_tokens:
  pad_token: "<|end_of_text|>"

llama-3.1-8b-squadv2_E3_V2

This model is a fine-tuned version of meta-llama/Llama-3.1-8B on the None dataset. It achieves the following results on the evaluation set:

Loss: 1.0569

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: 0.0002
train_batch_size: 4
eval_batch_size: 4
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 16
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 50
num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.0033	1	1.5437
0.9335	0.2512	77	0.9273
0.8874	0.5024	154	0.9174
0.8652	0.7537	231	0.9162
0.8276	1.0024	308	0.9154
0.7331	1.2537	385	0.9453
0.7383	1.5049	462	0.9459
0.7348	1.7561	539	0.9466
0.6799	2.0041	616	0.9538
0.5361	2.2553	693	1.0466
0.527	2.5065	770	1.0517
0.5181	2.7577	847	1.0569

Framework versions

PEFT 0.13.2
Transformers 4.45.2
Pytorch 2.3.1+cu121
Datasets 3.0.1
Tokenizers 0.20.1