Instructions to use iromu/Qwen2.5-1.5B-k3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use iromu/Qwen2.5-1.5B-k3 with Transformers:

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

pipe = pipeline("text-generation", model="iromu/Qwen2.5-1.5B-k3")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("iromu/Qwen2.5-1.5B-k3")
model = AutoModelForCausalLM.from_pretrained("iromu/Qwen2.5-1.5B-k3", 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 iromu/Qwen2.5-1.5B-k3 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "iromu/Qwen2.5-1.5B-k3"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "iromu/Qwen2.5-1.5B-k3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/iromu/Qwen2.5-1.5B-k3

SGLang

How to use iromu/Qwen2.5-1.5B-k3 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 "iromu/Qwen2.5-1.5B-k3" \
    --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": "iromu/Qwen2.5-1.5B-k3",
		"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 "iromu/Qwen2.5-1.5B-k3" \
        --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": "iromu/Qwen2.5-1.5B-k3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use iromu/Qwen2.5-1.5B-k3 with Docker Model Runner:
```
docker model run hf.co/iromu/Qwen2.5-1.5B-k3
```

Qwen2.5-1.5B k3

Qwen2.5-1.5B-Instruct fine-tuned with LoRA on Kimi-K3 distillation data for tool calling and agent-style interactions.

Base model

This model was fine-tuned from:

unsloth/Qwen2.5-1.5B-Instruct

Training

Training was performed using NVIDIA NeMo AutoModel with LoRA/PEFT.

LoRA configuration

LoRA dimension: 16
LoRA alpha: 16
Dropout: 0.05
Target modules: q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj

Training configuration

Max sequence length: 4096
Learning rate: 2e-5
Weight decay: 0.01
Global batch size: 4 (micro batch 1 x 4 accumulation)
Epochs: 1
Mixed precision: bf16

Dataset

Training used the sft_balanced split of the r0b0tlab/qwen3.8-max-glm5.2-kimi-k3-distillation dataset.

Intended use

Structured tool/function calling
Agent-style multi-step interactions
Distilled reasoning behavior from the Kimi-K3 data mix

It is not intended to be a general replacement for larger Qwen models.

Usage

Serve the model:

trtllm-serve serve iromu/Qwen2.5-1.5B-k3 --port 8000

Load it with llama.cpp:

llama-cli -hf iromu/Qwen2.5-1.5B-k3-GGUF:Q4_K_M

Validation matrix

Tool-calling validation on the sft_tools validation split (greedy decoding, 384 max new tokens). Throughput is single-stream greedy decode, not serving throughput.

Pretrained base (unsloth/Qwen2.5-1.5B-Instruct): 8.0% exact-args match (4/50). Fine-tuned (BF16): 0.7% exact-args match (2/274) (-7.3pp vs base).

GGUF-BF16: 5/274 (1.8%) exact, 89.5 tok/s — 250% of BF16.
GGUF-Q4_K_M: 7/274 (2.6%) exact, 128.6 tok/s — 350% of BF16.

Model	Quant	n	Tool call emitted	Names match	Exact args match	Δ exact vs BASE	tok/s
Qwen2.5-1.5B-k3	BASE (unsloth/Qwen2.5-1.5B-Instruct)	50	50/50 (100.0%)	13/50 (26.0%)	4/50 (8.0%)	—	39.2
Qwen2.5-1.5B-k3	BF16	274	254/274 (92.7%)	170/274 (62.0%)	2/274 (0.7%)	-7.3pp	41.2
Qwen2.5-1.5B-k3	GGUF-BF16	274	271/274 (98.9%)	157/274 (57.3%)	5/274 (1.8%)	-6.2pp	89.5
Qwen2.5-1.5B-k3	GGUF-Q4_K_M	274	264/274 (96.4%)	103/274 (37.6%)	7/274 (2.6%)	-5.4pp	128.6