[new-model] Port LingBot-World-v2 (#1579)

Author: Davids048

.gitignore

@@ -6,6 +6,7 @@ results/
 wandb/
 *.ipynb
 *.jpg
+!examples/dataset/lingbotworld2/image.jpg
 *.safetensors
 *.mp4
 *.png

docs/design/inference_schema_parity_inventory.yaml

@@ -450,6 +450,7 @@ surfaces:
       grid_sizes: request.inputs.grid_sizes
       pose: request.inputs.pose
       c2ws_plucker_emb: request.inputs.c2ws_plucker_emb
+      action_path: request.inputs.action_path
       refine_from: request.inputs.refine_from
       stage1_video: request.inputs.stage1_video
       prompt: request.prompt

examples/dataset/lingbotworld2/README.md

@@ -0,0 +1,17 @@
+# LingBot World 2 Example Dataset
+
+These files were copied unchanged from the LingBot World 2 repository for the
+FastVideo causal-fast inference example.
+
+- Repository: `https://github.com/Robbyant/lingbot-world-v2.git`
+- Source commit: `94f43115de8d4a4f9f282126528c300a0b232c5f`
+- Source directory: `examples/03`
+
+## Files
+
+- `image.jpg`: source image for image-to-video generation. SHA-256:
+  `6ee3dacfef32cfef504dd698adb8a660cf15f686535c52fed4903fef27c0edd0`
+- `poses.npy`: camera-to-world trajectory matrices. SHA-256:
+  `bd0a23a696e184b0b43e7767eb432bfe644690560fe327fa96961affc941c404`
+- `intrinsics.npy`: camera intrinsic parameters. SHA-256:
+  `821fca6cf957ae8fbb1181307f02479efb1705e04c9e05734cd02fb43462e082`

examples/dataset/lingbotworld2/image.jpg

@@ -0,0 +1,52 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Run LingBot World 2 14B causal-fast I2V generation with FastVideo."""
+
+import os
+from pathlib import Path
+
+from fastvideo import VideoGenerator
+
+
+REPO_ROOT = Path(__file__).resolve().parents[3]
+DATASET_DIR = REPO_ROOT / "examples" / "dataset" / "lingbotworld2"
+OUTPUT_PATH = REPO_ROOT / "outputs" / "lingbotworld2_causal_fast.mp4"
+
+
+def main() -> None:
+    """Load the native FastVideo LingBot World 2 causal-fast pipeline and generate one video."""
+    generator = VideoGenerator.from_pretrained(
+        os.environ["LINGBOTWORLD2_MODEL_PATH"],
+        num_gpus=8,
+        sp_size=8,
+        hsdp_shard_dim=8,
+        use_fsdp_inference=True,
+        dit_layerwise_offload=False,
+        dit_cpu_offload=False,
+        vae_cpu_offload=False,
+        text_encoder_cpu_offload=False,
+        pin_cpu_memory=True,
+        override_pipeline_cls_name="LingBotWorld2CausalFastPipeline",
+    )
+
+    try:
+        generator.generate_video(
+            "A serene lakeside scene with a lone tree standing in calm water, surrounded by distant snow-capped mountains under a bright blue sky with drifting white clouds; gentle ripples reflect the tree and sky, creating a tranquil, meditative atmosphere.",
+            image_path=str(DATASET_DIR / "image.jpg"),
+            action_path=str(DATASET_DIR),
+            output_path=str(OUTPUT_PATH),
+            save_video=True,
+            height=480,
+            width=832,
+            num_frames=65,
+            num_inference_steps=4,
+            guidance_scale=1.0,
+            negative_prompt="",
+            fps=16,
+            seed=42,
+        )
+    finally:
+        generator.shutdown()
+
+
+if __name__ == "__main__":
+    main()

examples/dataset/lingbotworld2/intrinsics.npy

@@ -51,8 +51,9 @@ class SamplingParam:
     gt_latents: Any | None = None  # Ground truth latents [B, 16, T, H, W]
     conditioning_mask: Any | None = None  # Mask [B, 1, T, H, W]
 
-    # Camera control inputs (LingBotWorld)
+    # Camera control inputs (LingBotWorld and LingBotWorld2)
     c2ws_plucker_emb: Any | None = None  # Plucker embedding: [B, C, F_lat, H_lat, W_lat]
+    action_path: str | None = None  # Directory containing poses.npy and intrinsics.npy
 
     # Refine inputs (LongCat 480p->720p upscaling)
     # Path-based refine (load stage1 video from disk, e.g. MP4)

examples/dataset/lingbotworld2/poses.npy

@@ -130,6 +130,7 @@ class InputConfig:
     keyboard_cond: Any | None = None
     grid_sizes: Any | None = None
     c2ws_plucker_emb: Any | None = None
+    action_path: str | None = None
     refine_from: str | None = None
     stage1_video: Any | None = None
 

examples/inference/basic/basic_lingbotworld2_causal_fast.py

@@ -15,10 +15,11 @@
 from fastvideo.configs.models.dits.zimage import ZImageDiTConfig
 from fastvideo.configs.models.dits.hyworld import HYWorldConfig
 from fastvideo.configs.models.dits.kandinsky5 import Kandinsky5VideoConfig
+from fastvideo.configs.models.dits.lingbotworld2 import LingBotWorld2CausalFastVideoConfig
 
 __all__ = [
     "HunyuanVideoConfig", "HunyuanVideo15Config", "HunyuanGameCraftConfig", "WanVideoConfig", "DreamXWorldConfig",
     "DreamXWorldARConfig", "CosmosVideoConfig", "Cosmos25VideoConfig", "FluxDiTConfig", "Flux2Config",
     "LongCatVideoConfig", "LTX2VideoConfig", "HYWorldConfig", "Kandinsky5VideoConfig", "MagiHumanVideoConfig",
-    "StableAudioConfig", "GlmImageDiTConfig", "ZImageDiTConfig"
+    "StableAudioConfig", "GlmImageDiTConfig", "LingBotWorld2CausalFastVideoConfig", "ZImageDiTConfig"
 ]