Instructions to use google/diffusiongemma-26B-A4B-it with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use google/diffusiongemma-26B-A4B-it with Transformers:

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

pipe = pipeline("image-text-to-text", model="google/diffusiongemma-26B-A4B-it")
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("google/diffusiongemma-26B-A4B-it")
model = AutoModelForMultimodalLM.from_pretrained("google/diffusiongemma-26B-A4B-it", 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 google/diffusiongemma-26B-A4B-it with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "google/diffusiongemma-26B-A4B-it"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "google/diffusiongemma-26B-A4B-it",
		"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/google/diffusiongemma-26B-A4B-it

SGLang

How to use google/diffusiongemma-26B-A4B-it 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 "google/diffusiongemma-26B-A4B-it" \
    --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": "google/diffusiongemma-26B-A4B-it",
		"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 "google/diffusiongemma-26B-A4B-it" \
        --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": "google/diffusiongemma-26B-A4B-it",
		"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 google/diffusiongemma-26B-A4B-it with Docker Model Runner:
```
docker model run hf.co/google/diffusiongemma-26B-A4B-it
```

fix: chat_template, close model turn across tool results + keep tool-call reasoning

#25

by dimondev - opened 4 days ago

base: refs/heads/main

←

from: refs/pr/25

Discussion Files changed

-2

dimondev

4 days ago

Summary

Two defects in chat_template.jinja, both verified with ChatLint against 49 conversation shapes. The patched file is attached / inlined as this PR.

	Hub `main`	this PR
score	260/294 (88%)	283/294 (96%)
errors	25	2

The remaining 2 are the preamble-hoist issue also present on E4B/31B (assistant text emitted after <tool_response|>); not addressed here.

Fix 1: misplaced `endif` (22 `marker_balance` errors)

ns.prev_non_tool_role was assigned outside the role != 'tool' block, so a tool message overwrote it to 'tool'. The next assistant turn then failed continue_same_model_turn and opened a second <|turn>model without closing the first:

broken:  ...<tool_response|><|turn>model\nIt is 18C...
fixed:   ...<tool_response|>It is 18C...<turn|>\n<|turn>user\n

This is the same class of turn-tag imbalance the Jul 15 fix cleared on gemma-4-31B-it / gemma-4-E4B-it. Those repos already have the endif in the right place; Diffusion Gemma did not get that change.

         {%- endif -%}
-    {%- endif -%}
-
     {#- Track previous non-tool role for next iteration (avoids O(n) backward scan) -#}
     {%- set ns.prev_non_tool_role = message['role'] -%}
+    {%- endif -%}
 {%- endfor -%}

Fix 2: tool-call reasoning dropped after the next user turn

- {%- set thinking_gate = (loop.index0 > ns_turn.last_user_idx) or (preserve_thinking and message.get('tool_calls')) -%}
+ {%- set thinking_gate = (loop.index0 > ns_turn.last_user_idx) or message.get('tool_calls') or preserve_thinking -%}

With the default preserve_thinking=false, an assistant turn that called a tool lost its reasoning_content as soon as a later user message existed, blowing ~27% of the cached prefix on every agent step after the first tool call.

Reproduce

pip install chatlint
# before
curl -sL https://huggingface.co/google/diffusiongemma-26B-A4B-it/raw/main/chat_template.jinja \
  | chatlint check /dev/stdin
# after (this PR's file)
chatlint check chat_template.jinja

Ask

If this lands, consider adding ChatLint's CI workflow so a sibling repo cannot drift from the Jul 15 fix again. The original Gemma 4 turn-tag regression survived 96 days; this file shows the same class of bug can reappear on a fork that nobody re-gated.

fix: chat_template, close model turn across tool results + keep tool-call reasoningc080de4d

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

Ready to merge

This branch is ready to get merged automatically.

· Sign up or log in to comment

fix: chat_template, close model turn across tool results + keep tool-call reasoning

Summary

Fix 1: misplaced endif (22 marker_balance errors)

Fix 2: tool-call reasoning dropped after the next user turn

Reproduce

Ask

Fix 1: misplaced `endif` (22 `marker_balance` errors)