Hugging Face's logo

Undi95
/
toxicqa-Llama2-7B-lora

PEFT
Safetensors
llama
Generated from Trainer
8-bit precision
bitsandbytes
Model card Files Community
toxicqa-Llama2-7B-lora / README.md
Undi95's picture
Undi95
Update README.md
38cf909
preview code
|
raw
history blame
3.65 kB
metadata
library_name: peft
tags:
  - generated_from_trainer
base_model: NousResearch/Llama-2-7b-hf
model-index:
  - name: NobodyExistsOnTheInternet/toxicqa
    results: []

Built with Axolotl

See axolotl config

axolotl version: 0.3.0

base_model: NousResearch/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: dataset
    type: sharegpt
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: 128
lora_alpha: 64
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:

wandb_project: toxicLlama-2-13B
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 1
micro_batch_size: 2
num_epochs: 2
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0002
eval_batch_size: 2

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
eval_table_size:
eval_table_max_new_tokens: 128
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
  bos_token: "<s>"
  eos_token: "</s>"
  unk_token: "<unk>"

NobodyExistsOnTheInternet/toxicqa

This model is a fine-tuned version of NousResearch/Llama-2-7b-hf on the NobodyExistsOnTheInternet/toxicqa dataset. It achieves the following results on the evaluation set:

  • Loss: 0.8100

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 0.0002
  • train_batch_size: 2
  • eval_batch_size: 2
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_steps: 10
  • num_epochs: 2

Training results

Training Loss Epoch Step Validation Loss
1.0748 0.0 1 1.1154
0.8635 0.25 176 0.8732
0.8284 0.5 352 0.8463
0.7928 0.75 528 0.8295
0.8313 1.0 704 0.8155
0.6694 1.23 880 0.8196
0.636 1.48 1056 0.8144
0.6842 1.73 1232 0.8105
0.6277 1.98 1408 0.8100

Framework versions

  • Transformers 4.36.2
  • Pytorch 2.0.1+cu118
  • Datasets 2.16.1
  • Tokenizers 0.15.0

Training procedure

The following bitsandbytes quantization config was used during training:

  • quant_method: bitsandbytes
  • load_in_8bit: True
  • load_in_4bit: False
  • llm_int8_threshold: 6.0
  • llm_int8_skip_modules: None
  • llm_int8_enable_fp32_cpu_offload: False
  • llm_int8_has_fp16_weight: False
  • bnb_4bit_quant_type: fp4
  • bnb_4bit_use_double_quant: False
  • bnb_4bit_compute_dtype: float32

Framework versions

  • PEFT 0.6.0