How to use from the
Use from the
Transformers library
# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="code-critic-model/qwen3-4b-sft-prm")
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("code-critic-model/qwen3-4b-sft-prm")
model = AutoModelForCausalLM.from_pretrained("code-critic-model/qwen3-4b-sft-prm", 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]:]))
Quick Links

Duplicate repository. The weights here are byte-identical to Qwen3-4B-Critic-SFT, the 4B SFT critic from Steer, Don't Solve. Use that repository; it has the full model card. This copy is kept because the DPO training runs in this organization reference it as their base_model.

qwen3-4b-sft-prm

Qwen3-4B-Instruct-2507 fine-tuned on critic-sft-cwm-qwen. The auto-generated card that used to live here named the wrong training run; the weights were always the CWM + Qwen corpus model.

Downloads last month
16
Safetensors
Model size
4B params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for code-critic-model/qwen3-4b-sft-prm

Finetuned
(2,761)
this model
Finetunes
4 models

Paper for code-critic-model/qwen3-4b-sft-prm