See axolotl config

axolotl version: 0.4.0

# use google/gemma-7b if you have access
#base_model: mhenrichsen/gemma-7b
base_model: google/gemma-7b
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

hub_model_id: MaziyarPanahi/gemma-7b-alpaca-52k-v0.1
hf_use_auth_token: true

load_in_8bit: false
load_in_4bit: true
strict: false

# huggingface repo
datasets:
  - path: tatsu-lab/alpaca
    type: alpaca
val_set_size: 0.1
output_dir: ./qlora-gemma-7b-alpaca

adapter: qlora
lora_r: 32
lora_alpha: 16
lora_dropout: 0.05
lora_target_linear: true

sequence_len: 4096
sample_packing: false
pad_to_sequence_len: false

wandb_project:
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:


gradient_accumulation_steps: 3
micro_batch_size: 2
num_epochs: 1
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

warmup_ratio: 0.1
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:

gemma-7b-alpaca-52k-v0.1

This model is a fine-tuned version of google/gemma-7b on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.1468

How to use

PEFT

from peft import PeftModel, PeftConfig
from transformers import AutoModelForCausalLM

model_id = "MaziyarPanahi/gemma-7b-alpaca-52k-v0.1"

config = PeftConfig.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained("google/gemma-7b")
model = PeftModel.from_pretrained(model, model_id)

Transformers

# Use a pipeline as a high-level helper
from transformers import pipeline

model_id = "MaziyarPanahi/gemma-7b-alpaca-52k-v0.1"

pipe = pipeline("text-generation", model=model_id)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id)

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
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 3
• total_train_batch_size: 24
• 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: 48
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss
1.5395	0.0	1	1.4186
1.099	0.25	488	1.1994
1.2188	0.5	976	1.1751
1.0511	0.75	1464	1.1468

Framework versions

• PEFT 0.8.2
• Transformers 4.39.0.dev0
• Pytorch 2.2.0+cu121
• Datasets 2.17.0
• Tokenizers 0.15.0