Instructions to use Gaoxp-USTC/CastReasoner-Model with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Gaoxp-USTC/CastReasoner-Model with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="Gaoxp-USTC/CastReasoner-Model") messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# pip install -U transformers accelerate # Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("Gaoxp-USTC/CastReasoner-Model") model = AutoModelForCausalLM.from_pretrained("Gaoxp-USTC/CastReasoner-Model", 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=256) print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use Gaoxp-USTC/CastReasoner-Model with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "Gaoxp-USTC/CastReasoner-Model" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Gaoxp-USTC/CastReasoner-Model", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/Gaoxp-USTC/CastReasoner-Model
- SGLang
How to use Gaoxp-USTC/CastReasoner-Model 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 "Gaoxp-USTC/CastReasoner-Model" \ --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": "Gaoxp-USTC/CastReasoner-Model", "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 "Gaoxp-USTC/CastReasoner-Model" \ --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": "Gaoxp-USTC/CastReasoner-Model", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use Gaoxp-USTC/CastReasoner-Model with Docker Model Runner:
docker model run hf.co/Gaoxp-USTC/CastReasoner-Model
CastReasoner-9B
Merged inference weights from the last 9B post-training run. This is the selected step-180 checkpoint of that run, not the final optimizer iteration (step 250). The matching SFT/RL data are released separately as a Dataset repository.
Contents: BF16 model.safetensors, model/generation configurations, tokenizer, and the original chat template. This is a native text-only Qwen3_5ForCausalLM export. Distributed training shards, optimizer state, private logs and endpoint credentials are excluded.
from transformers import AutoTokenizer, AutoModelForCausalLM
repo = "Gaoxp-USTC/CastReasoner-Model"
tokenizer = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(repo, dtype="auto", device_map="auto")
Use a Transformers version supporting qwen3_5_text (the local export metadata was validated with Transformers 5.3). Use the original forecasting prompts/chat template from the matching data release for forecasting. The package underwent configuration/tokenizer/header and file integrity checks; this packaging step does not claim a new GPU inference evaluation.
Checkpoint provenance: last-run validation selected step 180. Historical validation used the 720 test sample IDs, so the selection set was not an independent test holdout. This model card does not substitute a new evaluation claim for that history.
The upstream Qwen license is retained in LICENSE. Matching data: CastReasoner-Datasets.
- Downloads last month
- 24