Instructions to use stabilityai/japanese-stablelm-base-alpha-7b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use stabilityai/japanese-stablelm-base-alpha-7b with Transformers:

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

pipe = pipeline("text-generation", model="stabilityai/japanese-stablelm-base-alpha-7b", trust_remote_code=True)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("stabilityai/japanese-stablelm-base-alpha-7b", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use stabilityai/japanese-stablelm-base-alpha-7b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "stabilityai/japanese-stablelm-base-alpha-7b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "stabilityai/japanese-stablelm-base-alpha-7b",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/stabilityai/japanese-stablelm-base-alpha-7b

SGLang

How to use stabilityai/japanese-stablelm-base-alpha-7b 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 "stabilityai/japanese-stablelm-base-alpha-7b" \
    --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": "stabilityai/japanese-stablelm-base-alpha-7b",
		"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 "stabilityai/japanese-stablelm-base-alpha-7b" \
        --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": "stabilityai/japanese-stablelm-base-alpha-7b",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use stabilityai/japanese-stablelm-base-alpha-7b with Docker Model Runner:
```
docker model run hf.co/stabilityai/japanese-stablelm-base-alpha-7b
```

Fine tuning by LoRA

by konbraphat51 - opened Aug 19, 2023

Discussion

konbraphat51

Aug 19, 2023

I tried fine tune this model by PEFT library and Trainer from transformers library

I got this error:

japanese-stablelm-base-alpha-7b\69e1599948909cccca2369385fb6c82ef59086f4\modeling_japanese_stablelm_alpha.py", line 383, in apply_rotary_pos_emb
q_embed = (q * cos) + (rotate_half(q) * sin)
RuntimeError: The size of tensor a (32) must match the size of tensor b (64) at non-singleton dimension 3

The loading of the model is:

tokenizer = LlamaTokenizer.from_pretrained("novelai/nerdstash-tokenizer-v1", load_in_8bit=self.finetuner_properties.useint8, device_map="auto")

lm_model = AutoModelForCausalLM.from_pretrained("stabilityai/japanese-stablelm-base-alpha-7b", load_in_8bit=true, device_map="auto", trust_remote_code=True)

My arguments are:

#for LoraConfig
        lora_r:int = 8,
        lora_alpha:int = 16,
        lora_dropout:float = 0.05,
        lora_bias:str = "none",
        
#for Trainer
ta_epochs:int = 4,
        ta_logging_steps:int = 200,
        ta_save_steps:int = 100000,
        ta_save_total_limit:int = 3,
        ta_train_batch_size:int = 8,
        ta_warmup_steps:int = 200,
        ta_weight_decay:float = 0.1,
        ta_learning_rate:float = 5e-4,

What's wrong? This model unable to fine-tuning in int8 condition?

konbraphat51

Aug 19, 2023

I also tried variant="int8" but didn't solve

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment