Instructions to use cs-552-2026-claude-bots/group_model with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use cs-552-2026-claude-bots/group_model with Transformers:

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

pipe = pipeline("text-generation", model="cs-552-2026-claude-bots/group_model")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("cs-552-2026-claude-bots/group_model")
model = AutoModelForCausalLM.from_pretrained("cs-552-2026-claude-bots/group_model")
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

vLLM

How to use cs-552-2026-claude-bots/group_model with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "cs-552-2026-claude-bots/group_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": "cs-552-2026-claude-bots/group_model",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/cs-552-2026-claude-bots/group_model

SGLang

How to use cs-552-2026-claude-bots/group_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 "cs-552-2026-claude-bots/group_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": "cs-552-2026-claude-bots/group_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 "cs-552-2026-claude-bots/group_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": "cs-552-2026-claude-bots/group_model",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use cs-552-2026-claude-bots/group_model with Docker Model Runner:
```
docker model run hf.co/cs-552-2026-claude-bots/group_model
```

Group Model

This is a merge of pre-trained language models created using mergekit.

Merge Details

Merge Method

This model was merged using the TIES merge method using Qwen/Qwen3-1.7B as a base.

Models Merged

The following models were included in the merge:

Configuration

The following YAML configuration was used to produce this model:

# TIES merge configuration for 4 specialized Qwen fine-tunes
# Goal:
# - Preserve strong domain expertise from each model
# - Reduce destructive interference between skills
# - Keep general reasoning stability from the base model
#
# Recommended for:
# - Qwen2 / Qwen2.5 instruction-tuned variants
# - Same architecture + same parameter count
# - Same tokenizer + same base checkpoint lineage

models:
  # ---------------------------
  # Math specialist
  # ---------------------------
  - model: cs-552-2026-claude-bots/math_model
    parameters:
      # High density because math capabilities are usually sparse
      # and easily lost during merging
      density:
        - filter: self_attn
          value: 0.72
        - filter: mlp
          value: 0.82

      # Strong contribution in reasoning-heavy blocks
      weight:
        - filter: self_attn
          value: 1.25
        - filter: mlp
          value: 1.15
        - value: 1.10

  # ---------------------------
  # Knowledge / factual model
  # ---------------------------
  - model: cs-552-2026-claude-bots/general_knowledge_model
    parameters:
      # Moderate density:
      # factual tuning tends to be more distributed
      density: 0.58

      # Slightly lower than math to avoid overwriting reasoning
      weight:
        - filter: self_attn
          value: 1.00
        - filter: mlp
          value: 0.95
        - value: 0.95

  # ---------------------------
  # Multilingual specialist
  # ---------------------------
  - model: cs-552-2026-claude-bots/multilingual_model
    parameters:
      # Language capabilities are often spread broadly,
      # so use reasonably high density
      density:
        - filter: embed_tokens
          value: 0.90
        - filter: self_attn
          value: 0.68
        - filter: mlp
          value: 0.62

      # Stronger influence on embeddings and attention
      weight:
        - filter: embed_tokens
          value: 1.30
        - filter: self_attn
          value: 1.10
        - value: 1.00

  # ---------------------------
  # Safety / alignment model
  # ---------------------------
  - model: cs-552-2026-claude-bots/safety_model
    parameters:
      # Lower density prevents excessive refusals
      # while still preserving alignment behavior
      density: 0.34

      # Important but intentionally constrained
      weight:
        - filter: self_attn
          value: 0.82
        - filter: mlp
          value: 0.72
        - value: 0.75

merge_method: ties

# IMPORTANT:
# Use the ORIGINAL shared pretrained base model
# from which all four fine-tunes were derived.
base_model: Qwen/Qwen3-1.7B

parameters:
  # Critical for TIES stability
  normalize: true

  # Helps reduce memory usage and improves masking behavior
  int8_mask: true

  # Trim very small parameter deltas
  # Good default for 4-way merges
  prune_threshold: 0.015

dtype: bfloat16

# Optional:
# tokenizer_source: base
# chat_template: auto