See axolotl config

axolotl version: 0.4.1

base_model: meta-llama/Meta-Llama-3-8B  # same model you originally used
# Load your previously fine-tuned model as a PEFT adapter
peft_model: ahmedelgebaly/llama-3.1-8b-squadv2
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

load_in_8bit: false
load_in_4bit: true
strict: false

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

test_datasets:
  - path: ahmedelgebaly/SciQ_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: 16
lora_dropout: 0.05
lora_target_modules:
lora_target_linear: true
lora_fan_in_fan_out:

wandb_project: llama-3.1-8b-squadv2_SciQ_e1
wandb_entity:
wandb_watch:
wandb_name: llama-3.1-8b-squadv2-v0_SciQ_e1
wandb_log_model:

hub_model_id: ahmedelgebaly/llama-3.1-8b-squadv2_SciQ_e1

gradient_accumulation_steps: 4
micro_batch_size: 4
num_epochs: 1
optimizer: paged_adamw_32bit
lr_scheduler: cosine
learning_rate: 0.0002

train_on_inputs: false
group_by_length: false
bf16: auto
fp16:
tf32: false

gradient_checkpointing: true
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true

warmup_steps: 10
evals_per_epoch: 4
eval_table_size:
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
  pad_token: "<|end_of_text|>"

llama-3.1-8b-squadv2_SciQ_e1

This model is a fine-tuned version of meta-llama/Meta-Llama-3-8B on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.9369

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: 10
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss
1.7866	0.0305	1	1.8420
1.1313	0.2443	8	1.0968
0.841	0.4885	16	0.9655
0.8722	0.7328	24	0.9415
0.8736	0.9771	32	0.9369

Framework versions

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