Hugging Face's logo

dvdmrs09
/
peft-gemma-2b-vXI

Text Generation
PEFT
Safetensors
gemma
Generated from Trainer
4-bit precision
bitsandbytes
Model card Files Community
peft-gemma-2b-vXI / README.md
dvdmrs09's picture
dvdmrs09
Update README.md
51b258b verified
preview code
|
raw
history blame contribute delete
3.37 kB
metadata
license: gemma
library_name: peft
tags:
  - generated_from_trainer
base_model: google/gemma-2b
model-index:
  - name: out1v
    results: []
pipeline_tag: text-generation

Built with Axolotl

See axolotl config

axolotl version: 0.4.0

# use google/gemma-7b if you have access
base_model: google/gemma-2b
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer
#lora_model_dir: 
load_in_8bit: false
load_in_4bit: true
strict: false

# huggingface repo
datasets:
  - path: ./python-oasst/cleaned-dataset.jsonl
    type: oasst
val_set_size: 0.20
output_dir: ./out1v

adapter: lora
lora_r: 8
lora_alpha: 16
lora_dropout: 0.05
lora_target_linear: true
gptq: false
sequence_len: 4096
sample_packing: false
pad_to_sequence_len: true

wandb_project: gemma-2b-it
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:


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

train_on_inputs: true
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_ratio: 0.1
evals_per_epoch: 4
eval_table_size:
eval_max_new_tokens: 128
saves_per_epoch: 1
debug:
deepspeed: deepspeed_configs/zero3_bf16.json
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:

out1v

This model is a fine-tuned version of google/gemma-2b on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 1.1243

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
  • distributed_type: multi-GPU
  • num_devices: 2
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 32
  • total_eval_batch_size: 8
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_steps: 1045
  • num_epochs: 2

Training results

Training Loss Epoch Step Validation Loss
1.9865 0.0 1 2.2808
1.1033 0.25 2613 1.1338
1.0035 0.5 5226 1.1308
1.0089 0.75 7839 1.1272
0.9816 1.0 10452 1.1238
0.8727 1.25 13065 1.1270
0.9951 1.5 15678 1.1254
1.01 1.75 18291 1.1242
1.0677 2.0 20904 1.1243

Framework versions

  • PEFT 0.10.0
  • Transformers 4.40.0.dev0
  • Pytorch 2.2.2+cu121
  • Datasets 2.18.0
  • Tokenizers 0.15.0