Instructions to use rovangju/Swift-Qwen3.8-27b-W8A8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use rovangju/Swift-Qwen3.8-27b-W8A8 with Transformers:

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

pipe = pipeline("image-text-to-text", model="rovangju/Swift-Qwen3.8-27b-W8A8")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("rovangju/Swift-Qwen3.8-27b-W8A8")
model = AutoModelForMultimodalLM.from_pretrained("rovangju/Swift-Qwen3.8-27b-W8A8", device_map="auto")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.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(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use rovangju/Swift-Qwen3.8-27b-W8A8 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "rovangju/Swift-Qwen3.8-27b-W8A8"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "rovangju/Swift-Qwen3.8-27b-W8A8",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/rovangju/Swift-Qwen3.8-27b-W8A8

SGLang

How to use rovangju/Swift-Qwen3.8-27b-W8A8 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 "rovangju/Swift-Qwen3.8-27b-W8A8" \
    --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": "rovangju/Swift-Qwen3.8-27b-W8A8",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

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 "rovangju/Swift-Qwen3.8-27b-W8A8" \
        --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": "rovangju/Swift-Qwen3.8-27b-W8A8",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use rovangju/Swift-Qwen3.8-27b-W8A8 with Docker Model Runner:
```
docker model run hf.co/rovangju/Swift-Qwen3.8-27b-W8A8
```

rovangju commited on 7 days ago

Commit

bf37ab8

verified ·

1 Parent(s): 43ddb52

Update README.md

Browse files

Bit of cleanup of llm fluff.

Files changed (1) hide show

README.md +3 -22

README.md CHANGED Viewed

@@ -6,7 +6,7 @@ library_name: transformers
 base_model: ukisai/Swift-Qwen3.8-27b
 quantized_with:
   name: llm-compressor
-  version: "0.14"
 tags:
 - compressed-tensors
 - w8a8
@@ -75,11 +75,6 @@ Two things were handled deliberately during quantization:
   original index, per-channel scale shapes checked, and a hard failure if any
   activation scale tensors were written) passed for all shards.
-## Requirements
-- vLLM with compressed-tensors W8A8 INT8 support (compute capability **≥ 7.5**)
-- Fits on a single 64 GB GPU (loaded at ~28.5 GiB)
 ## Usage — vLLM
 ```bash
@@ -120,16 +115,6 @@ are wide; treat them as a sanity gate, not a final benchmark. The strict-match
 GSM8K score is a known artifact of the harness's strict filter on this
 template, not a model failure.
-## Reproducing the quantization
-```bash
-CUDA_VISIBLE_DEVICES=1 .venv/bin/python scripts/quantize_swift_w8a8.py
-```
-- Base model resident in RAM (~95 GB free needed); quant ops stream each
-  module onto a free GPU (needs ~10 GB free on `cuda:0`)
-- ~45 GB free disk; ~5–10 min wall time
 ## Appendix — benchmark serve script
 The vLLM run used for the benchmark numbers above (identical for both
@@ -137,12 +122,8 @@ models; the only difference between the two runs was the model path):
 ```bash
 #!/usr/bin/env bash
-# 03-w8a8-champ: kv-cache-dtype removed, attn backend -> FLASH_ATTN, K -> 5
-# generated from study.yaml by bench/study.py — do not hand-edit;
-# edit the run entry in study.yaml and rerun `python bench/study.py materialize`.
-set -euo pipefail
-export CUDA_VISIBLE_DEVICES=0
 vllm serve <MODEL> \
   --dtype bfloat16 \
@@ -169,4 +150,4 @@ Swift distribution.
 `swift-open-license-1.0`, inherited from the base model
 [ukisai/Swift-Qwen3.8-27b](https://huggingface.co/ukisai/Swift-Qwen3.8-27b) —
 see the [base model's LICENSE](https://huggingface.co/ukisai/Swift-Qwen3.8-27b/blob/main/LICENSE)
-for the exact terms (this quantization adds no restrictions of its own).

 base_model: ukisai/Swift-Qwen3.8-27b
 quantized_with:
   name: llm-compressor
+  version: '0.14'
 tags:
 - compressed-tensors
 - w8a8
   original index, per-channel scale shapes checked, and a hard failure if any
   activation scale tensors were written) passed for all shards.
 ## Usage — vLLM
 ```bash
 GSM8K score is a known artifact of the harness's strict filter on this
 template, not a model failure.
 ## Appendix — benchmark serve script
 The vLLM run used for the benchmark numbers above (identical for both
 ```bash
 #!/usr/bin/env bash
+set -euo pipefail
 vllm serve <MODEL> \
   --dtype bfloat16 \
 `swift-open-license-1.0`, inherited from the base model
 [ukisai/Swift-Qwen3.8-27b](https://huggingface.co/ukisai/Swift-Qwen3.8-27b) —
 see the [base model's LICENSE](https://huggingface.co/ukisai/Swift-Qwen3.8-27b/blob/main/LICENSE)
+for the exact terms (this quantization adds no restrictions of its own).