Instructions to use BirdToast/gemma4-31b-confetti-adapter with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use BirdToast/gemma4-31b-confetti-adapter with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("/home/aibox/models/gemma-4-31B-pt-embed-it")
model = PeftModel.from_pretrained(base_model, "BirdToast/gemma4-31b-confetti-adapter")

Transformers

How to use BirdToast/gemma4-31b-confetti-adapter with Transformers:

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

pipe = pipeline("text-generation", model="BirdToast/gemma4-31b-confetti-adapter")

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("BirdToast/gemma4-31b-confetti-adapter", dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use BirdToast/gemma4-31b-confetti-adapter with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "BirdToast/gemma4-31b-confetti-adapter"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "BirdToast/gemma4-31b-confetti-adapter",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/BirdToast/gemma4-31b-confetti-adapter

SGLang

How to use BirdToast/gemma4-31b-confetti-adapter with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "BirdToast/gemma4-31b-confetti-adapter" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "BirdToast/gemma4-31b-confetti-adapter",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "BirdToast/gemma4-31b-confetti-adapter" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "BirdToast/gemma4-31b-confetti-adapter",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use BirdToast/gemma4-31b-confetti-adapter with Docker Model Runner:
```
docker model run hf.co/BirdToast/gemma4-31b-confetti-adapter
```

gemma4-31b-pt-embed-r32a8-textcomp

This model was fine-tuned using SFT.

W&B run: https://wandb.ai/cooawoo-personal/Gemma4-31B/runs/lkdzz1dj

Training procedure

Hyperparameters

Parameter	Value
Learning rate	`1e-05`
LR scheduler	rex (custom; max_lr=1e-5, min_lr=1e-6, warmup_ratio=0.05)
Per-device batch size	1
Gradient accumulation	4
Effective batch size	4
Epochs	2
Max sequence length	4096
Optimizer	OptimizerNames.PAGED_ADAMW_8BIT
Warmup ratio	0.05
Max gradient norm	1.0
Precision	bf16
Loss type	nll
Chunked cross-entropy	yes

LoRA configuration

Parameter	Value
Rank (r)	32
Alpha	8
Target modules	.*language_model.layers.\d+.(self_attn.(q
rsLoRA	yes
Quantization	4-bit (nf4)

Dataset statistics

Dataset	Samples	Total tokens	Trainable tokens
brainrot_chatlog.jsonl	3,232	1,115,996	1,115,996
erotica_quality_cleaned_20pct_seed42.json	575	2,212,468	2,212,468
worm_chapters.json	658	2,268,255	2,268,255
counter_signal_training_balanced.jsonl	1,038	4,173,666	4,173,666
Total	5,503	9,770,385	9,770,385

Training config

model_name_or_path: Columbidae/gemma4-31b-pt-embed-it
data_config: data.yaml
prepared_dataset: prepared_packed
output_dir: gemma4-31b-pt-embed-r32a8-textcomp
attn_implementation: flex_attention
bf16: true
gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: false
use_cce: true
chunked_mlp: true
chunked_mlp_chunks: 16
dataloader_num_workers: 2
dataloader_pin_memory: true
model_parallel: true
max_memory:
  0: 16GiB
  1: 24GiB
max_length: 4096
per_device_train_batch_size: 1
gradient_accumulation_steps: 4
pad_to_multiple_of: 128
use_peft: true
load_in_4bit: true
bnb_4bit_quant_type: nf4
lora_r: 32
lora_alpha: 8
lora_dropout: 0.0
use_rslora: true
lora_target_modules: .*language_model\.layers\.\d+\.(self_attn\.(q|k|v|o)_proj|mlp\.(gate|up|down)_proj)$
learning_rate: 1.0e-05
lr_scheduler_type: rex
warmup_ratio: 0.05
weight_decay: 0.0
max_grad_norm: 1.0
optim: paged_adamw_8bit
num_train_epochs: 2
saves_per_epoch: 2
save_total_limit: 4
rolling_save_steps: 30
rolling_save_total_limit: 1
logging_steps: 1
disable_tqdm: false
report_to: wandb
run_name: g4-31b-pt-embed-r32a8-textcomp

Data config

datasets:
- path: erotica_quality_cleaned_20pct_seed42.json
  type: text
  truncation_strategy: split
- path: counter_signal_training_balanced.jsonl
  type: text
  truncation_strategy: split
- path: worm_chapters.json
  type: text
  truncation_strategy: split
- path: brainrot_chatlog.jsonl
  type: text
  truncation_strategy: split
shuffle_datasets: true
shuffle_combined: true
shuffle_seed: 42
eval_split: 0
split_seed: 42
assistant_only_loss: false

Framework versions

PEFT 0.18.1
Loft: 0.1.0
Transformers: 5.5.4
Pytorch: 2.6.0+cu124
Datasets: 4.6.1
Tokenizers: 0.22.2

Downloads last month: 55

Model tree for BirdToast/gemma4-31b-confetti-adapter

Base model

Columbidae/gemma4-31b-pt-embed-it

Adapter

(1)

this model