ayushgs's picture
Adding the dataset name
4a35ed1 verified
metadata
library_name: peft
tags:
  - generated_from_trainer
base_model: meta-llama/Llama-2-7b-hf
model-index:
  - name: models/run2
    results: []

Built with Axolotl

See axolotl config

axolotl version: 0.4.0

# This file is used by the training script in train.ipynb. You can read more about
# the format and see more examples at https://github.com/OpenAccess-AI-Collective/axolotl.
# One of the parameters you might want to play around with is `num_epochs`: if you have a
# smaller dataset size, making that large can have good results.

base_model: meta-llama/Llama-2-7b-hf
base_model_config: meta-llama/Llama-2-7b-hf
model_type: LlamaForCausalLM
tokenizer_type: LlamaTokenizer
is_llama_derived_model: true

load_in_8bit: true
load_in_4bit: false
strict: false

datasets:
  - path: ./resources/train.jsonl
    type: alpaca
dataset_prepared_path: ./resources/last_run_prepared
val_set_size: 0.05
output_dir: ./models/run2

sequence_len: 4096
sample_packing: 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:

# This will report stats from your training run to https://wandb.ai/. If you don't want to create a wandb account you can comment this section out.
wandb_project: google-boolq
wandb_entity:
wandb_watch:
wandb_run_id: run2
wandb_log_model:


gradient_accumulation_steps: 4
micro_batch_size: 2
num_epochs: 5
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: false

warmup_steps: 10
eval_steps: 20
save_steps: 60
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
  bos_token: "<s>"
  eos_token: "</s>"
  unk_token: "<unk>"

models/run2

This model is a fine-tuned version of meta-llama/Llama-2-7b-hf on the google/boolq dataset. It achieves the following results on the evaluation set:

  • Loss: 0.3248

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: 5

Training results

Training Loss Epoch Step Validation Loss
8.1402 0.02 1 8.4654
0.3619 0.3 20 0.3422
0.3432 0.6 40 0.3379
0.3227 0.9 60 0.3375
0.3315 1.18 80 0.3373
0.3204 1.48 100 0.3315
0.3291 1.79 120 0.3300
0.319 2.07 140 0.3277
0.3165 2.37 160 0.3280
0.3133 2.67 180 0.3388
0.3088 2.97 200 0.3263
0.3448 3.25 220 0.3252
0.3264 3.55 240 0.3273
0.2946 3.85 260 0.3310
0.3212 4.13 280 0.3244
0.3118 4.43 300 0.3245
0.3377 4.73 320 0.3248

Framework versions

  • PEFT 0.9.0
  • Transformers 4.40.0.dev0
  • Pytorch 2.1.2+cu121
  • Datasets 2.18.0
  • Tokenizers 0.15.0

Evaluation results

Model Accuracy Avg Time Avg Cost
gpt-4 0.874 0.624 0.00552
gpt-3.5-turbo 0.824 0.530 0.0000916
llama2-7b-ft-boolq-run2 0.856 0.0432 0.0000155

ft vs gpt4

  • Cost Improvement: 357x
  • Latency Improvement: 12x

ft vs gpt3.5-turbo

  • Cost Improvement: 6x
  • Latency Improvement: 14x