Hugging Face's logo

kloodia
/
lora-8b-code

PEFT
TensorBoard
Safetensors
llama
Generated from Trainer
8-bit precision
bitsandbytes
Model card Files Metrics Community
lora-8b-code / README.md
kloodia's picture
kloodia
Upload folder using huggingface_hub
3305bc0 verified
preview code
|
raw
history blame contribute delete
3.48 kB
metadata
license: other
library_name: peft
tags:
  - generated_from_trainer
base_model: meta-llama/Meta-Llama-3-8B
model-index:
  - name: lora-out
    results: []

Built with Axolotl

See axolotl config

axolotl version: 0.4.0

base_model: meta-llama/Meta-Llama-3-8B
model_type: LlamaForCausalLM
tokenizer_type: AutoTokenizer

load_in_8bit: true
load_in_4bit: false
strict: false

datasets:
  - path: kloodia/raw_github
    type: oasst
dataset_prepared_path:
val_set_size: 0.05
output_dir: ./lora-out

sequence_len: 4096
sample_packing: true
pad_to_sequence_len: true

adapter: lora
lora_model_dir:
lora_r: 32
lora_alpha: 16
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:

wandb_project:
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 4
micro_batch_size: 1
num_epochs: 3
optimizer: adamw_bnb_8bit
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
s2_attention:

warmup_steps: 10
evals_per_epoch: 4
eval_table_size:
eval_max_new_tokens: 128
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
   pad_token: <|end_of_text|>

lora-out

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: 0.8585

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: 1
  • eval_batch_size: 1
  • seed: 42
  • distributed_type: multi-GPU
  • num_devices: 4
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 16
  • total_eval_batch_size: 4
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_steps: 10
  • num_epochs: 3

Training results

Training Loss Epoch Step Validation Loss
0.8982 0.0 1 0.9102
0.7637 0.25 243 0.8141
0.7688 0.5 486 0.8165
0.7058 0.75 729 0.8200
0.6952 1.0 972 0.8218
0.701 1.23 1215 0.8342
0.6826 1.48 1458 0.8373
0.6102 1.74 1701 0.8371
0.6007 1.99 1944 0.8391
0.6088 2.22 2187 0.8578
0.6148 2.47 2430 0.8587
0.5864 2.72 2673 0.8585

Framework versions

  • PEFT 0.10.0
  • Transformers 4.40.0.dev0
  • Pytorch 2.1.2+cu118
  • Datasets 2.15.0
  • Tokenizers 0.15.0