[None][feat] Add GlmImage text-to-image pipeline support to VisualGen

docs/source/models/visual-generation.md

@@ -47,6 +47,7 @@ TensorRT-LLM **VisualGen** provides a unified inference stack for diffusion mode
 | `nvidia/Cosmos3-Edge` | Text-to-Image, Text-to-Video, Image-to-Video (Nemotron-dense backbone, 480p-native) |
 | `hunyuanvideo-community/HunyuanVideo-1.5-Diffusers-480p_t2v` | Text-to-Video |
 | `hunyuanvideo-community/HunyuanVideo-1.5-Diffusers-720p_t2v` | Text-to-Video |
+| `zai-org/GLM-Image` | Text-to-Image |
 
 
 Models are auto-detected from the checkpoint directory. Diffusers-format models are detected via `model_index.json`; LTX-2 monolithic safetensors checkpoints are detected via embedded metadata. The `AutoPipeline` registry selects the appropriate pipeline class automatically.
@@ -67,6 +68,7 @@ Models are auto-detected from the checkpoint directory. Diffusers-format models
 | **Qwen-Image-Edit-2511** | Yes | Yes | No | No | Yes | No | No | Yes | Yes | No | No | No | No | No |
 | **Cosmos3** | Yes | Yes | No | No | Yes | Yes | Yes | Yes | Yes | Yes | No | No | Yes | No |
 | **HunyuanVideo 1.5** | Yes | Yes | No | No | No | No | No | No | No | Yes | No | No | No | No |
+| **GlmImage** | Yes | Yes | No | No | No | No | No | No | No | Yes | No | No | No | No |
 
 [^1]: FLUX models use embedded guidance and do not have a separate negative prompt path, so CFG parallelism is not applicable.
 

examples/visual_gen/README.md

@@ -25,6 +25,7 @@ python models/cosmos3_ti2v.py --prompt "A robot arm picks fruit in a grocery sto
 python models/qwen_image.py
 python models/qwen_image_layered.py --image /path/to/image.png
 python models/qwen_image_edit.py --image /path/to/source.png --prompt "Make the image look like a watercolor painting"
+python models/glm_image.py
 python models/hunyuan_t2v.py
 
 # With engine config (quant, parallelism, etc.)

examples/visual_gen/models/glm_image.py

@@ -0,0 +1,81 @@
+# Copyright (c) 2026, NVIDIA CORPORATION.  All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""GlmImage text-to-image generation.
+
+Usage:
+    python glm_image.py
+"""
+
+import argparse
+from pathlib import Path
+
+from tensorrt_llm import VisualGen, VisualGenArgs
+
+
+def _output_paths(output_path: str, num_images: int) -> str | list[str]:
+    if num_images == 1:
+        return output_path
+
+    path = Path(output_path)
+    return [str(path.with_name(f"{path.stem}_{index}{path.suffix}")) for index in range(num_images)]
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument(
+        "--model",
+        default="zai-org/GLM-Image",
+        help="Hugging Face model id or local checkpoint path.",
+    )
+    parser.add_argument(
+        "--visual_gen_args",
+        "--extra_visual_gen_options",
+        dest="visual_gen_args",
+        help="Optional VisualGenArgs YAML file.",
+    )
+    parser.add_argument(
+        "--prompt",
+        default="A serene mountain lake at sunrise, watercolor style, highly detailed",
+        help="Text prompt for image generation.",
+    )
+    parser.add_argument(
+        "--num_images_per_prompt",
+        type=int,
+        default=1,
+        help="Number of images to generate for the prompt.",
+    )
+    parser.add_argument(
+        "--output_path",
+        default="glm_image_output.png",
+        help="Image output path. Multiple images append an index before the suffix.",
+    )
+    return parser.parse_args()
+
+
+def main() -> None:
+    args = parse_args()
+    if args.num_images_per_prompt < 1:
+        raise ValueError("--num_images_per_prompt must be >= 1")
+    extra_args = VisualGenArgs.from_yaml(args.visual_gen_args) if args.visual_gen_args else None
+    visual_gen = VisualGen(model=args.model, args=extra_args)
+    params = visual_gen.default_params
+    params.num_images_per_prompt = args.num_images_per_prompt
+    output = visual_gen.generate(inputs=args.prompt, params=params)
+    saved = output.save(_output_paths(args.output_path, args.num_images_per_prompt))
+    print(f"Saved image(s) to {saved}")
+
+
+if __name__ == "__main__":
+    main()

tensorrt_llm/_torch/visual_gen/models/__init__.py

@@ -35,6 +35,7 @@
 from ..pipeline_registry import AutoPipeline, register_pipeline
 from .cosmos3 import Cosmos3OmniMoTPipeline
 from .flux import Flux2Pipeline, FluxPipeline
+from .glm_image import GlmImagePipeline
 from .hunyuan_video1_5 import HunyuanVideo15Pipeline
 from .ltx2 import LTX2Pipeline  # noqa: F401
 from .qwen_image import QwenImageEditPlusPipeline, QwenImagePipeline
@@ -46,6 +47,7 @@
     "BasePipeline",
     "FluxPipeline",
     "Flux2Pipeline",
+    "GlmImagePipeline",
     "QwenImageEditPlusPipeline",
     "QwenImageLayeredPipeline",
     "QwenImagePipeline",

tensorrt_llm/_torch/visual_gen/models/glm_image/__init__.py

@@ -0,0 +1,19 @@
+# SPDX-FileCopyrightText: Copyright (c) 2022-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from .pipeline_glm_image import GlmImagePipeline
+from .transformer_glm_image import GlmImageAttention, GlmImageTransformer2DModel
+
+__all__ = ["GlmImageAttention", "GlmImagePipeline", "GlmImageTransformer2DModel"]