Instructions to use formalmathatepfl/qwen3-cpt with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use formalmathatepfl/qwen3-cpt with Transformers:

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

pipe = pipeline("text-generation", model="formalmathatepfl/qwen3-cpt")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("formalmathatepfl/qwen3-cpt")
model = AutoModelForCausalLM.from_pretrained("formalmathatepfl/qwen3-cpt", device_map="auto")
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 Settings

vLLM

How to use formalmathatepfl/qwen3-cpt with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "formalmathatepfl/qwen3-cpt"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "formalmathatepfl/qwen3-cpt",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/formalmathatepfl/qwen3-cpt

SGLang

How to use formalmathatepfl/qwen3-cpt 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 "formalmathatepfl/qwen3-cpt" \
    --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": "formalmathatepfl/qwen3-cpt",
		"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 "formalmathatepfl/qwen3-cpt" \
        --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": "formalmathatepfl/qwen3-cpt",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use formalmathatepfl/qwen3-cpt with Docker Model Runner:
```
docker model run hf.co/formalmathatepfl/qwen3-cpt
```

Qwen3-8B-cpt

This model is a fine-tuned version of Qwen/Qwen3-8B-Base on the cpt_data dataset. It achieves the following results on the evaluation set:

Loss: 0.7642

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 3e-06
train_batch_size: 2
eval_batch_size: 2
seed: 42
distributed_type: multi-GPU
num_devices: 8
total_train_batch_size: 16
total_eval_batch_size: 16
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.03
num_epochs: 2.0

Training results

Training Loss	Epoch	Step	Validation Loss
1.1321	0.0594	100	1.0979
1.0204	0.1188	200	1.0132
1.0155	0.1781	300	0.9683
0.9392	0.2375	400	0.9368
0.8472	0.2969	500	0.9136
0.9268	0.3563	600	0.8942
0.8568	0.4157	700	0.8777
0.9049	0.4751	800	0.8637
0.8401	0.5344	900	0.8517
0.8449	0.5938	1000	0.8408
0.8638	0.6532	1100	0.8310
0.8485	0.7126	1200	0.8226
0.8273	0.7720	1300	0.8150
0.8165	0.8314	1400	0.8077
0.824	0.8907	1500	0.8014
0.7944	0.9501	1600	0.7957
0.7492	1.0095	1700	0.7912
0.754	1.0689	1800	0.7873
0.7184	1.1283	1900	0.7835
0.7298	1.1876	2000	0.7802
0.7093	1.2470	2100	0.7777
0.7624	1.3064	2200	0.7744
0.7419	1.3658	2300	0.7723
0.7439	1.4252	2400	0.7704
0.7064	1.4846	2500	0.7688
0.7387	1.5439	2600	0.7675
0.7527	1.6033	2700	0.7664
0.7359	1.6627	2800	0.7656
0.7167	1.7221	2900	0.7650
0.7159	1.7815	3000	0.7646
0.7399	1.8409	3100	0.7644
0.6995	1.9002	3200	0.7642
0.7915	1.9596	3300	0.7642

Framework versions

Transformers 4.57.6
Pytorch 2.9.0+cu128
Datasets 4.0.0
Tokenizers 0.22.2

Downloads last month: 1,295

Safetensors

Model size

308k params

Tensor type

BF16

Model tree for formalmathatepfl/qwen3-cpt

Base model

Qwen/Qwen3-8B-Base

Finetuned

(531)

this model

Finetunes

1 model