Instructions to use magiccodingman/Qwen3.8-27B-heretic-ara-DFlash2-fp8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use magiccodingman/Qwen3.8-27B-heretic-ara-DFlash2-fp8 with Transformers:

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

pipe = pipeline("text-generation", model="magiccodingman/Qwen3.8-27B-heretic-ara-DFlash2-fp8")

# Load model directly
from transformers import AutoTokenizer, AutoModel

tokenizer = AutoTokenizer.from_pretrained("magiccodingman/Qwen3.8-27B-heretic-ara-DFlash2-fp8")
model = AutoModel.from_pretrained("magiccodingman/Qwen3.8-27B-heretic-ara-DFlash2-fp8", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use magiccodingman/Qwen3.8-27B-heretic-ara-DFlash2-fp8 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "magiccodingman/Qwen3.8-27B-heretic-ara-DFlash2-fp8"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "magiccodingman/Qwen3.8-27B-heretic-ara-DFlash2-fp8",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/magiccodingman/Qwen3.8-27B-heretic-ara-DFlash2-fp8

SGLang

How to use magiccodingman/Qwen3.8-27B-heretic-ara-DFlash2-fp8 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 "magiccodingman/Qwen3.8-27B-heretic-ara-DFlash2-fp8" \
    --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": "magiccodingman/Qwen3.8-27B-heretic-ara-DFlash2-fp8",
		"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 "magiccodingman/Qwen3.8-27B-heretic-ara-DFlash2-fp8" \
        --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": "magiccodingman/Qwen3.8-27B-heretic-ara-DFlash2-fp8",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use magiccodingman/Qwen3.8-27B-heretic-ara-DFlash2-fp8 with Docker Model Runner:
```
docker model run hf.co/magiccodingman/Qwen3.8-27B-heretic-ara-DFlash2-fp8
```

Qwen3.8-27B-heretic-ara-DFlash2-fp8 / fp8_target_conditioned_validation.json

magiccodingman

Upload folder using huggingface_hub

b6bbe8a verified about 1 month ago

Raw History Blame Contribute Delete

1.25 kB

	{
	"status": "pass",
	"comparison": "BF16 DFlash2 vs selective block-FP8 weights dequantized to BF16",
	"samples": 8,
	"draft_positions": 56,
	"proposal_kl_mean": 0.0018929995239503794,
	"proposal_kl_median": 0.001594925176043174,
	"proposal_kl_p95": 0.00300999847240746,
	"proposal_kl_max": 0.008521616478903316,
	"proposal_top1_agreement": 0.9642857313156128,
	"selector_path_token_agreement": 0.8571429252624512,
	"selector_full_path_agreement": 0.75,
	"top16_candidate_overlap": 0.9765625,
	"draft_hidden_relative_rmse": 0.03063499420217177,
	"draft_hidden_cosine_similarity": 0.9995341897010803,
	"teacher_mean_forward_seconds": 0.20252476473979186,
	"student_mean_forward_seconds": 0.5098142223869218,
	"per_sample_selector_token_agreement": [
	1.0,
	1.0,
	1.0,
	1.0,
	0.0,
	1.0,
	0.8571429252624512,
	1.0
	],
	"gates": {
	"proposal_kl_mean_max": 0.02,
	"proposal_top1_agreement_min": 0.9,
	"selector_path_token_agreement_min": 0.85,
	"top16_candidate_overlap_min": 0.95
	},
	"limitation": "RTX 3090 dequantizes FP8 weights to BF16. This test uses real target hidden states and the exact target LM head, but it is not an end-to-end native W8A8 throughput or acceptance benchmark."
	}