Hugging Face's logo

gardner
/
TinyLlama-1.1B-SlimOrca

PEFT
PyTorch
Safetensors
English
llama
axolotl
Generated from Trainer
8-bit precision
bitsandbytes
Model card Files Community
1
TinyLlama-1.1B-SlimOrca / README.md
gardner's picture
gardner
LoRA weights and merged
44c8dea verified
preview code
|
raw
history blame
3.8 kB
metadata
license: apache-2.0
library_name: peft
tags:
  - axolotl
  - generated_from_trainer
base_model: gardner/TinyLlama-1.1B-Instruct-3T
model-index:
  - name: TinyLlama-1.1B-SlimOrca-LoRA
    results: []

Built with Axolotl

See axolotl config

axolotl version: 0.4.0

base_model: gardner/TinyLlama-1.1B-Instruct-3T
model_type: LlamaForCausalLM
tokenizer_type: LlamaTokenizer
is_llama_derived_model: true

load_in_8bit: true
load_in_4bit: false
strict: false

datasets:
  - path: Open-Orca/SlimOrca-Dedup
    type: sharegpt
    split: train

dataset_prepared_path: ./dsprepare/Open-Orca/SlimOrca-Dedup
val_set_size: 0.05
output_dir: ./tinyllama-1.1b-slimorca-lora
hub_model_id: gardner/TinyLlama-1.1B-SlimOrca-LoRA

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: tinyllama
wandb_entity: gardner
wandb_name: tinyllama-slimorca

gradient_accumulation_steps: 4
micro_batch_size: 2
num_epochs: 4
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0002

train_on_inputs: false
group_by_length: false
bf16: true
fp16: false
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
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:

TinyLlama-1.1B-SlimOrca-LoRA

This model is a fine-tuned version of gardner/TinyLlama-1.1B-Instruct-3T on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 0.5636

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: 2
  • eval_batch_size: 2
  • seed: 42
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 8
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_steps: 10
  • num_epochs: 4

Training results

Training Loss Epoch Step Validation Loss
1.2902 0.0 1 0.9116
1.0653 0.25 1126 0.6458
1.0279 0.5 2252 0.6187
0.8918 0.75 3378 0.6042
0.9362 1.0 4504 0.5924
0.8138 1.23 5630 0.5863
0.9669 1.48 6756 0.5814
1.019 1.73 7882 0.5742
0.9232 1.98 9008 0.5695
0.8507 2.22 10134 0.5700
0.7542 2.47 11260 0.5662
0.8325 2.72 12386 0.5639
0.7913 2.97 13512 0.5617
0.8372 3.2 14638 0.5648
0.8984 3.45 15764 0.5638
0.7898 3.7 16890 0.5636

Framework versions

  • PEFT 0.7.1
  • Transformers 4.37.0
  • Pytorch 2.1.2+cu121
  • Datasets 2.16.1
  • Tokenizers 0.15.0