library_name, extra_gated_heading, extra_gated_prompt, extra_gated_button_content, license, license_name, license_link, base_model, datasets

library_name	extra_gated_heading	extra_gated_prompt	extra_gated_button_content	license	license_name	license_link	base_model	datasets
transformers	Access Gemma on Hugging Face	To access Gemma on Hugging Face, you’re required to review and agree to Google’s usage license. To do this, please ensure you’re logged-in to Hugging Face and click below. Requests are processed immediately.	Acknowledge license	other	gemma-terms-of-use	https://ai.google.dev/gemma/terms	google/gemma-7b	vicgalle/alpaca-gpt4

Gemmalpaca-7B

This is gemma-7b model supervised fine-tuned on the vicgalle/alpaca-gpt4 dataset. It outperforms gemma-7b-it, Google's chat version, on Nous' benchmark suite.

It's mostly a test to see how fine-tuning works with Gemma models on a well-known dataset.

🔍 Applications

This model has a context length of 8k. I recommend using it with the Alpaca chat template and NOT the Gemma Instruct template (works perfectly with LM Studio). You also want to add </s> as a stop token.

🏆 Evaluation

Nous

Gemmalpaca-7B outperforms gemma-7b and gemma-7b-it on Nous' benchmark suite (evaluation performed using LLM AutoEval). See the entire leaderboard here.

Model	Average	AGIEval	GPT4All	TruthfulQA	Bigbench
mlabonne/Gemmalpaca-7B 📄	34.45	21.6	40.87	44.85	30.49
google/gemma-2b 📄	34.26	22.7	43.35	39.96	31.03
google/gemma-7b 📄	33.56	20.64	38.49	46.61	28.51
google/gemma-7b-it 📄	33.53	21.33	40.84	41.7	30.25

🧩 Configuration

It was trained using Axolotl with the following configuration.

base_model: alpindale/gemma-7b
model_type: AutoModelForCausalLM
tokenizer_config: philschmid/gemma-tokenizer-chatml
tokenizer_type: AutoTokenizer
tokenizer_use_fast: true

load_in_8bit: false
load_in_4bit: true
strict: false

datasets:
  - path: vicgalle/alpaca-gpt4
    type: alpaca

dataset_prepared_path:
val_set_size: 0.01
output_dir: ./out

sequence_len: 2048
sample_packing: true
pad_to_sequence_len: true

adapter: qlora
lora_model_dir:
lora_r: 32
lora_alpha: 64
lora_dropout: 0.05
lora_target_linear: true

wandb_project: axolotl
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 2
micro_batch_size: 4
num_epochs: 3
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_steps: 10
evals_per_epoch: 10
eval_table_size:
eval_table_max_new_tokens: 128
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.1
fsdp:
fsdp_config:
special_tokens: