Instructions to use lablup/GLM-5.2-NVFP4-DFlash-ko with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use lablup/GLM-5.2-NVFP4-DFlash-ko with Transformers:

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

pipe = pipeline("text-generation", model="lablup/GLM-5.2-NVFP4-DFlash-ko", trust_remote_code=True)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModel

tokenizer = AutoTokenizer.from_pretrained("lablup/GLM-5.2-NVFP4-DFlash-ko", trust_remote_code=True)
model = AutoModel.from_pretrained("lablup/GLM-5.2-NVFP4-DFlash-ko", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use lablup/GLM-5.2-NVFP4-DFlash-ko with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "lablup/GLM-5.2-NVFP4-DFlash-ko"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "lablup/GLM-5.2-NVFP4-DFlash-ko",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/lablup/GLM-5.2-NVFP4-DFlash-ko

SGLang

How to use lablup/GLM-5.2-NVFP4-DFlash-ko 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 "lablup/GLM-5.2-NVFP4-DFlash-ko" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "lablup/GLM-5.2-NVFP4-DFlash-ko",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "lablup/GLM-5.2-NVFP4-DFlash-ko" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "lablup/GLM-5.2-NVFP4-DFlash-ko",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use lablup/GLM-5.2-NVFP4-DFlash-ko with Docker Model Runner:
```
docker model run hf.co/lablup/GLM-5.2-NVFP4-DFlash-ko
```

GLM-5.2-NVFP4-DFlash-ko

A DFlash block-diffusion draft model for speculative decoding, paired with the nvidia/GLM-5.2-NVFP4 target (753B-param MoE, 40B active). Specialized for Korean-query agentic (tool-calling) and coding traffic — trained on captured multi-turn agent trajectories, not benchmark suites.

Load with trust_remote_code=True; the draft is a 1.05B DFlashDraftModel (5 layers, block size 16) that consumes concatenated hidden states from target layers [1, 20, 38, 56, 75] and predicts a block of masked tokens in parallel.

Serving (vLLM, TP8 + expert parallel)

vllm serve nvidia/GLM-5.2-NVFP4 \
  --tensor-parallel-size 8 --enable-expert-parallel \
  --max-model-len 262144 \
  --tool-call-parser glm47 --enable-auto-tool-choice \
  --reasoning-parser glm45 \
  --speculative-config '{"method": "dflash", "model": "lablup/GLM-5.2-NVFP4-DFlash-ko", "num_speculative_tokens": 15, "attention_backend": "triton_attn"}'

Serve with CUDA graphs (not --enforce-eager) for full throughput. k=15 is optimal on this stack — a k-sweep confirmed shorter blocks lose despite higher accept ratio.

Results (checkpoint at 33,971 steps)

Measured on vLLM TP8, CUDA graphs, against nvidia/GLM-5.2-NVFP4. MAL = mean acceptance length (1 + accepted/drafts); speedup is measured decode tok/s vs the same server with speculation off (~118 tok/s baseline, batch-1).

Paper suites (greedy, non-thinking):

suite	MAL	decode tok/s	speedup
HumanEval	4.87	400	3.38×
MATH-500	2.88	242	2.07×
MBPP	2.65	224	1.90×
GSM8K	2.50	212	1.79×
MT-Bench	1.94	167	1.42×

Production traffic (temperature 1.0 / top-p 0.95, the serving distribution):

workload	MAL	decode tok/s	speedup
Agentic multi-turn (tool-calling)	2.71	198	1.72×
Korean coding queries	2.77	220	1.92×

On real Korean agentic traffic this draft holds MAL flat across agentic and Korean (2.71 / 2.77), where an English-trained draft drops ~0.6 MAL between the two.

Training

Warm-started from a preview draft, then trained on 135,887 captured agentic pairs at 16K context (1 epoch, 33,971 steps, lr 7e-5 cosine).
Data: real coding-agent sessions (OpenCode / Codex / Claude Code harnesses) driving a GLM-5.2-NVFP4 endpoint over 17.6K Korean-localized GitHub issues across 619 permissive-license repos, with responses captured server-side (exact target token ids).
4 nodes × 8 B200, ~6.9 days.

License

MIT, following the nvidia/GLM-5.2-NVFP4 target. The draft weights are original (HF random init, trained from scratch). Verify upstream dataset/target licenses for your own redistribution.