Tiny dummy models
Collection
Randomly initialized tiny models for debugging/testing purpose • 176 items • Updated • 6
How to use from
SGLangUse 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 "yujiepan/opt-tiny-random" \
--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": "yujiepan/opt-tiny-random",
"prompt": "Once upon a time,",
"max_tokens": 512,
"temperature": 0.5
}'Quick Links
This model is randomly initialized, using the config from facebook/opt-30b but with smaller size. Note the model is in float16.
Codes:
import torch
import transformers
import os
from optimum.intel.openvino import OVModelForCausalLM
save_path = '/tmp/yujiepan/opt-tiny-random'
repo_id = 'yujiepan/opt-tiny-random'
config = transformers.AutoConfig.from_pretrained('facebook/opt-30b')
config.ffn_dim = 32
config.hidden_size = 8
config.num_attention_heads = 2
config.num_hidden_layers = 2
config.word_embed_proj_dim = 8
model = transformers.AutoModelForCausalLM.from_config(config, torch_dtype=torch.float16)
model = model.half()
model.save_pretrained(save_path)
tokenizer = transformers.AutoTokenizer.from_pretrained('facebook/opt-30b')
tokenizer.save_pretrained(save_path)
ovmodel = OVModelForCausalLM.from_pretrained(save_path, export=True)
ovmodel = ovmodel.half()
ovmodel.save_pretrained(save_path)
os.system(f'ls -alh {save_path}')
from huggingface_hub import create_repo, upload_folder
create_repo(repo_id, exist_ok=True)
upload_folder(repo_id=repo_id, folder_path=save_path)
- Downloads last month
- 5,350
Install from pip and serve model
# Install SGLang from pip: pip install sglang# Start the SGLang server: python3 -m sglang.launch_server \ --model-path "yujiepan/opt-tiny-random" \ --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": "yujiepan/opt-tiny-random", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'