Instructions to use weihao1/MuMath-Code-L-70B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use weihao1/MuMath-Code-L-70B with Transformers:

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

pipe = pipeline("text-generation", model="weihao1/MuMath-Code-L-70B")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("weihao1/MuMath-Code-L-70B")
model = AutoModelForCausalLM.from_pretrained("weihao1/MuMath-Code-L-70B", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use weihao1/MuMath-Code-L-70B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "weihao1/MuMath-Code-L-70B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "weihao1/MuMath-Code-L-70B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/weihao1/MuMath-Code-L-70B

SGLang

How to use weihao1/MuMath-Code-L-70B 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 "weihao1/MuMath-Code-L-70B" \
    --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": "weihao1/MuMath-Code-L-70B",
		"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 "weihao1/MuMath-Code-L-70B" \
        --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": "weihao1/MuMath-Code-L-70B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use weihao1/MuMath-Code-L-70B with Docker Model Runner:
```
docker model run hf.co/weihao1/MuMath-Code-L-70B
```

MuMath-Code-L-70B

File size: 1,393 Bytes

{
    "num_layers": 80,
    "hidden_size": 8192,
    "ffn_hidden_size": 28672,
    "layernorm_epsilon": 1e-05,
    "max_sequence_length": 2048,
    "num_attention_heads": 64,
    "num_key_value_heads": 8,
    "max_position_embeddings": 4096,
    "tensor_model_parallel_size": 1,
    "pipeline_model_parallel_size": 1,
    "data_parallel_size": 1,
    "use_distributed_optimizer": true,
    "use_rotary_position_embeddings": true,
    "add_embedding_padding": true,
    "make_vocab_size_divisible_by": 128,
    "vocab_size": 32000,
    "transpose_mlp_dense": true,
    "transpose_query_key_value": true,
    "load_platform": "megatron",
    "save_platform": "huggingface",
    "load_ckpt_path": "mumath_code_stage_2_pot_bs8/iter_0025128",
    "save_ckpt_path": "mumath_code_stage_2_pot_bs8/iter_0025128_hf",
    "common_config_path": "config/llama-2-70b.json",
    "no_load_optim": true,
    "no_save_optim": false,
    "megatron_path": "/workspace/Megatron-LM/",
    "_name_or_path": "meta-llama/Llama-2-70b-hf",
    "architectures": [
        "LlamaForCausalLM"
    ],
    "bos_token_id": 1,
    "eos_token_id": 2,
    "hidden_act": "silu",
    "intermediate_size": 28672,
    "initializer_range": 0.02,
    "model_type": "llama",
    "num_hidden_layers": 80,
    "pad_token_id": 0,
    "rms_norm_eps": 1e-05,
    "tie_word_embeddings": false,
    "pretraining_tp": 1,
    "use_cache": true
}