Instructions to use mlfoundations-dev/a1_math_open2math with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use mlfoundations-dev/a1_math_open2math with Transformers:

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

pipe = pipeline("text-generation", model="mlfoundations-dev/a1_math_open2math")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("mlfoundations-dev/a1_math_open2math")
model = AutoModelForCausalLM.from_pretrained("mlfoundations-dev/a1_math_open2math")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use mlfoundations-dev/a1_math_open2math with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "mlfoundations-dev/a1_math_open2math"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "mlfoundations-dev/a1_math_open2math",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/mlfoundations-dev/a1_math_open2math

SGLang

How to use mlfoundations-dev/a1_math_open2math 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 "mlfoundations-dev/a1_math_open2math" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "mlfoundations-dev/a1_math_open2math",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "mlfoundations-dev/a1_math_open2math" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "mlfoundations-dev/a1_math_open2math",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use mlfoundations-dev/a1_math_open2math with Docker Model Runner:
```
docker model run hf.co/mlfoundations-dev/a1_math_open2math
```

sedrickkeh commited on Apr 16, 2025

Commit

113ed60

verified ·

1 Parent(s): de50131

Upload configs.yaml with huggingface_hub

Browse files

Files changed (1) hide show

configs.yaml +22 -24

configs.yaml CHANGED Viewed

@@ -1,36 +1,34 @@
 assistant_tag: gpt
-bf16: true
 content_tag: value
-cutoff_len: 16384
-dataloader_num_workers: 4
-dataloader_persistent_workers: true
-dataloader_pin_memory: true
 dataset: mlfoundations-dev/a1_math_open2math
 dataset_dir: ONLINE
-ddp_timeout: 180000000
-deepspeed: dcft/train/zero3.json
-do_train: true
-enable_liger_kernel: false
-eval_strategy: 'no'
 finetuning_type: full
 formatting: sharegpt
-global_batch_size: 128
-gradient_accumulation_steps: 4
 hub_model_id: mlfoundations-dev/a1_math_open2math
-include_hp: dcft/train/hp_settings/paper/reasoning_medium.yaml
-learning_rate: 4.0e-05
-logging_steps: 1
 lr_scheduler_type: cosine
 messages: conversations
 model_name_or_path: Qwen/Qwen2.5-7B-Instruct
-num_train_epochs: 5.0
-output_dir: /tmp/dcft_checkpoints/train/checkpoints/a1_math_open2math
-overwrite_cache: true
-per_device_train_batch_size: 1
-plot_loss: true
-preprocessing_num_workers: 4
-push_to_db: true
-push_to_hub: true
 report_to: wandb
 role_tag: from
 run_name: a1_math_open2math
@@ -38,4 +36,4 @@ save_strategy: epoch
 stage: sft
 template: qwen25
 user_tag: human
-warmup_ratio: 0.1

 assistant_tag: gpt
+bf16: 'True'
 content_tag: value
+cutoff_len: '16384'
+dataloader_num_workers: '4'
+dataloader_persistent_workers: 'True'
+dataloader_pin_memory: 'True'
 dataset: mlfoundations-dev/a1_math_open2math
 dataset_dir: ONLINE
+ddp_timeout: '180000000'
+deepspeed: /opt/ml/code/zero3.json
+do_train: 'True'
+enable_liger_kernel: 'True'
 finetuning_type: full
 formatting: sharegpt
+global_batch_size: '128'
+gradient_accumulation_steps: '16'
 hub_model_id: mlfoundations-dev/a1_math_open2math
+learning_rate: 4e-05
+logging_steps: '1'
 lr_scheduler_type: cosine
 messages: conversations
 model_name_or_path: Qwen/Qwen2.5-7B-Instruct
+num_train_epochs: '5.0'
+output_dir: /opt/ml/model
+overwrite_cache: 'True'
+per_device_train_batch_size: '1'
+plot_loss: 'True'
+preprocessing_num_workers: '16'
+push_to_db: 'True'
+push_to_hub: 'True'
 report_to: wandb
 role_tag: from
 run_name: a1_math_open2math
 stage: sft
 template: qwen25
 user_tag: human
+warmup_ratio: '0.1'