Shivik-2B-Reasoning-SAFE
A reasoning-optimized language model with Chain-of-Thought (CoT) capabilities using <think> tags.
Model Details
| Property | Value |
|---|---|
| Parameters | 2.67B |
| Hidden Size | 2048 |
| Layers | 32 |
| Context Length | 131072 |
| CoT Support | ✅ Yes (<think> tags) |
Usage
from transformers import AutoModelForCausalLM, AutoTokenizer
model_id = "shivash/Shivik-2B-Reasoning-SAFE"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
model_id,
torch_dtype="auto",
device_map="auto"
)
# For reasoning tasks, the model uses <think> tags
prompt = "Solve this step by step: What is 15% of 80?"
messages = [
{"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(text, return_tensors="pt").to(model.device)
outputs = model.generate(
**inputs,
max_new_tokens=512,
temperature=0.7,
do_sample=True,
)
response = tokenizer.decode(outputs[0], skip_special_tokens=False)
print(response)
Chain-of-Thought Format
The model uses <think> tags for internal reasoning:
<think>
Let me work through this step by step...
15% means 15/100 = 0.15
0.15 × 80 = 12
</think>
The answer is 12.
Training
This model was trained on reasoning datasets with Chain-of-Thought demonstrations.
License
Apache 2.0