Instructions to use arcAman07/apex-coordinator-qwen1.5b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use arcAman07/apex-coordinator-qwen1.5b with Transformers:

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

pipe = pipeline("text-generation", model="arcAman07/apex-coordinator-qwen1.5b")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("arcAman07/apex-coordinator-qwen1.5b")
model = AutoModelForCausalLM.from_pretrained("arcAman07/apex-coordinator-qwen1.5b", 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 arcAman07/apex-coordinator-qwen1.5b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "arcAman07/apex-coordinator-qwen1.5b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "arcAman07/apex-coordinator-qwen1.5b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/arcAman07/apex-coordinator-qwen1.5b

SGLang

How to use arcAman07/apex-coordinator-qwen1.5b 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 "arcAman07/apex-coordinator-qwen1.5b" \
    --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": "arcAman07/apex-coordinator-qwen1.5b",
		"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 "arcAman07/apex-coordinator-qwen1.5b" \
        --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": "arcAman07/apex-coordinator-qwen1.5b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use arcAman07/apex-coordinator-qwen1.5b with Docker Model Runner:
```
docker model run hf.co/arcAman07/apex-coordinator-qwen1.5b
```

APEX Coordinator — Qwen2.5-1.5B

A small learned coordinator that plans a team of LLM agents on long-horizon professional-services tasks (the APEX-Agents benchmark). Given a task it emits one plan: a schedule of which agents run in which round (parallel vs serial) and the routing between them:

{"batches": [["corpus_librarian", "metric_extractor"], ["ratio_calculator"]],
 "routes": {"ratio_calculator": ["corpus_librarian"]}}

Training: Stage-1 SFT warm-start (behaviour-cloning a commitment-parallel plan) → Stage-2 GRPO with a floor-protected reward (each sampled plan is executed by the real agent team + graded by a rubric judge).

Results (64-task financial test set, same grader)

system (workers)	score	serial-depth
single-agent	0.29–0.32	1.0
sequential	0.71–0.73	4.4
trained_coordinator (this model)	0.760 (Opus) / 0.795 (GLM)	3.2

Highest accuracy of the compared systems, and beats the sequential baseline on both accuracy and critical-path depth. A 5-seed reproducibility run: 0.760 ± 0.035 (Opus) / 0.795 ± 0.016 (GLM).

Load

from transformers import AutoModelForCausalLM, AutoTokenizer
tok = AutoTokenizer.from_pretrained("arcAman07/apex-coordinator-qwen1.5b")
model = AutoModelForCausalLM.from_pretrained("arcAman07/apex-coordinator-qwen1.5b")

It is used inside a multi-agent harness that executes the emitted plan with worker agents (Opus / GLM) and grades the result. The coordinator does greedy decoding of a single plan per task.

Downloads last month: 328

Safetensors

Model size

2B params

Tensor type

BF16

Model tree for arcAman07/apex-coordinator-qwen1.5b

Base model

Qwen/Qwen2.5-1.5B

Finetuned

Qwen/Qwen2.5-1.5B-Instruct

Finetuned

(1769)

this model