Publish LingBot as a LoRA-only adapter

Remove merged transformer files and update the model card and integrity manifest. Adapter weights and configuration are unchanged.

README.md

@@ -17,7 +17,7 @@ tags:
 
 # WORL-Lingbot
 
-**WORL-Lingbot** is a camera-conditioned video world-model policy trained with
+**WORL-Lingbot** is a camera-conditioned video world-model LoRA adapter trained with
 WORL on top of [LingBot-World-Fast](https://huggingface.co/robbyant/lingbot-world-fast).
 It generates video continuations conditioned on an initial image, text and
 camera trajectories.
@@ -35,15 +35,13 @@ and [WORL-Helios](https://huggingface.co/Efficient-Large-Model/WORL-Helios).
 | Base revision | `b7a1db2da9bd3ba675ee01c228f41db0309b981e` |
 | Base subfolder | `lingbot_world_fast` |
 | Architecture | Camera-conditioned `WanModelFast` transformer |
-| Release formats | BF16 PEFT LoRA adapter and merged BF16 transformer |
+| Release format | BF16 PEFT LoRA adapter |
 | LoRA rank / alpha | 128 / 128 |
-| Merged parameter count | 18,544,332,864 |
 | Output geometry documented by the source release | 81 frames, 832 × 464 pixels, 16 FPS |
 
-The two formats represent the same released policy. The merged transformer
-has the adapter incorporated and does not need PEFT for inference. Both forms
-still require the compatible text encoder, tokenizer, VAE, scheduler and
-LingBot runtime, which are not included in this repository.
+Load this adapter on top of the pinned LingBot-World-Fast base transformer.
+The compatible base transformer, text encoder, tokenizer, VAE, scheduler and
+LingBot runtime are required separately.
 
 ## Inputs and outputs
 
@@ -73,35 +71,12 @@ dataset is distributed separately from these weights.
 
 ## Download and load
 
-For the adapter only:
-
-```bash
-hf download Efficient-Large-Model/WORL-Lingbot \
-  --exclude 'merged/*' \
-  --local-dir artifacts/WORL-Lingbot
-```
-
-For the full repository, including the merged transformer:
-
 ```bash
 hf download Efficient-Large-Model/WORL-Lingbot --local-dir artifacts/WORL-Lingbot
 ```
 
-With a compatible LingBot runtime installed, load the merged transformer:
-
-```python
-import torch
-from wan.modules.model_fast import WanModelFast
-
-model = WanModelFast.from_pretrained(
-    "Efficient-Large-Model/WORL-Lingbot",
-    subfolder="merged",
-    torch_dtype=torch.bfloat16,
-    control_type="cam",
-).eval()
-```
-
-Alternatively, attach the adapter to the pinned base transformer:
+With a compatible LingBot runtime installed, attach the adapter to the pinned
+base transformer:
 
 ```python
 import torch
@@ -122,7 +97,7 @@ model = PeftModel.from_pretrained(
 ).eval()
 ```
 
-These examples load the transformer. Integrate it with the remaining compatible
+This example loads the base transformer and adapter. Integrate them with the remaining compatible
 pipeline components for complete video generation.
 
 ## Files and provenance
@@ -131,19 +106,14 @@ pipeline components for complete video generation.
 | --- | --- |
 | `adapter_model.safetensors` | Approximately 1.23 GB BF16 adapter |
 | `adapter_config.json` | PEFT configuration |
-| `merged/` | Eight indexed BF16 transformer shards, approximately 37.09 GB, and config |
-| `merged/SHA256SUMS` | Checksums for the merged files |
 | `RELEASE_MANIFEST.json` | Source revision and SHA-256 inventory |
 
-Weights and model configurations are unchanged from
+Adapter weights and configuration are unchanged from
 [qyoo/lingbot-worl at d3e449c675f5](https://huggingface.co/qyoo/lingbot-worl/tree/d3e449c675f5a8bfb3afd29e30f1675e4445b9c3).
-Documentation and release manifests were updated for this repository. Original
-export, merge and load records remain available at that source revision.
-
-The source records report a safe PEFT merge, matching sampled merged weights,
-and an H100 reload with no remaining LoRA state keys. Those checks establish
-merge/load behavior, not a new training run or a general performance claim.
-No new GPU benchmark was performed for this publication.
+Documentation and release manifests were updated for this repository's
+adapter-only release. The original export record remains available at the
+pinned source revision. File integrity is checked against that source snapshot;
+no new GPU benchmark was performed for this publication.
 
 ## Intended use and limitations
 

RELEASE_MANIFEST.json

@@ -27,8 +27,8 @@
     },
     {
       "path": "README.md",
-      "size_bytes": 5828,
-      "sha256": "a1597b03b840eb11b05ac7993a566ac0bf125cced7437a08a041bbac4fa768d7"
+      "size_bytes": 4884,
+      "sha256": "33507fb8cc826f364239a04039add527bfe5da82094e31cb32ff8925eb6e5868"
     },
     {
       "path": "adapter_config.json",
@@ -49,66 +49,6 @@
       "path": "assets/worl-teaser.webp",
       "size_bytes": 273380,
       "sha256": "4cce61f77c912cfdbeac7a2b6d8626c3d1b77099744fee068aa5e5d2d98c89b3"
-    },
-    {
-      "path": "merged/README.md",
-      "size_bytes": 702,
-      "sha256": "7b34dc542821a8d5867978b392fd6c086353600970c5a14968864199188ed581"
-    },
-    {
-      "path": "merged/SHA256SUMS",
-      "size_bytes": 1203,
-      "sha256": "eb9ccb4a64139c585c31c92a265da7791a357d63d0491b56e9de31e5e5224428"
-    },
-    {
-      "path": "merged/config.json",
-      "size_bytes": 532,
-      "sha256": "8e162023892651c30de45c41d029b2d05767585a2eec25a74fa9becd872e3267"
-    },
-    {
-      "path": "merged/diffusion_pytorch_model-00001-of-00008.safetensors",
-      "size_bytes": 4991624976,
-      "sha256": "97870275ccec0fa64a213e7909d3bae3d77cb2021ec37609d4c29127f1f05263"
-    },
-    {
-      "path": "merged/diffusion_pytorch_model-00002-of-00008.safetensors",
-      "size_bytes": 4982358760,
-      "sha256": "3a16767352aea3e7bbd58bc4e92488aac3e24d456131da7c553662ea7769fe3f"
-    },
-    {
-      "path": "merged/diffusion_pytorch_model-00003-of-00008.safetensors",
-      "size_bytes": 4861666096,
-      "sha256": "fad09faf0c9353f6803c5852fc2aaec7d28d4e5dc049b3db05c6c65221d10288"
-    },
-    {
-      "path": "merged/diffusion_pytorch_model-00004-of-00008.safetensors",
-      "size_bytes": 4966568072,
-      "sha256": "14cec26b8afe32622f885d3fd32f6732e17eedbee7466a2fe7478d51b2d94703"
-    },
-    {
-      "path": "merged/diffusion_pytorch_model-00005-of-00008.safetensors",
-      "size_bytes": 4929878840,
-      "sha256": "2411c8bfe88a18816f2195e859910e42786ee4791a5e1c63cfc50ba8d01a5f46"
-    },
-    {
-      "path": "merged/diffusion_pytorch_model-00006-of-00008.safetensors",
-      "size_bytes": 4950774384,
-      "sha256": "3a7db7a6e36fe047fa66c1f86e67004f6cc3ecb2a5e5ff0066694257ac98fae6"
-    },
-    {
-      "path": "merged/diffusion_pytorch_model-00007-of-00008.safetensors",
-      "size_bytes": 4982358936,
-      "sha256": "c3dd20329b34c13e79a3c5f6b23171434fea59131d98c37625a23bf339630707"
-    },
-    {
-      "path": "merged/diffusion_pytorch_model-00008-of-00008.safetensors",
-      "size_bytes": 2423583064,
-      "sha256": "48cc5e471d323e03bcb96ed39f8ea44b307a9f96b84327847c2656f40949e01e"
-    },
-    {
-      "path": "merged/diffusion_pytorch_model.safetensors.index.json",
-      "size_bytes": 127685,
-      "sha256": "2444bb18169f99786e3aafb3080a8cbd4998359300851c6570f90be57493e6ff"
     }
   ]
 }

merged/README.md

@@ -1,15 +0,0 @@
-# WORL-Lingbot: merged transformer
-
-This directory contains the BF16 camera-conditioned `WanModelFast`
-transformer with the WORL adapter merged. It has 18,544,332,864
-parameters stored in eight indexed Safetensors shards.
-
-See the [main model card](../README.md) for loading examples, inputs,
-limitations and provenance. The text encoder, tokenizer, VAE, scheduler
-and compatible LingBot runtime are required separately.
-
-Weight files and configuration are unchanged from the source release.
-This documentation and `SHA256SUMS` were updated for WORL-Lingbot.
-Run `sha256sum -c SHA256SUMS` from this directory to check its files.
-
-License: [Apache 2.0](../LICENSE.txt). See [NOTICE.txt](../NOTICE.txt).

merged/SHA256SUMS

@@ -1,11 +0,0 @@
-7b34dc542821a8d5867978b392fd6c086353600970c5a14968864199188ed581  README.md
-8e162023892651c30de45c41d029b2d05767585a2eec25a74fa9becd872e3267  config.json
-97870275ccec0fa64a213e7909d3bae3d77cb2021ec37609d4c29127f1f05263  diffusion_pytorch_model-00001-of-00008.safetensors
-3a16767352aea3e7bbd58bc4e92488aac3e24d456131da7c553662ea7769fe3f  diffusion_pytorch_model-00002-of-00008.safetensors
-fad09faf0c9353f6803c5852fc2aaec7d28d4e5dc049b3db05c6c65221d10288  diffusion_pytorch_model-00003-of-00008.safetensors
-14cec26b8afe32622f885d3fd32f6732e17eedbee7466a2fe7478d51b2d94703  diffusion_pytorch_model-00004-of-00008.safetensors
-2411c8bfe88a18816f2195e859910e42786ee4791a5e1c63cfc50ba8d01a5f46  diffusion_pytorch_model-00005-of-00008.safetensors
-3a7db7a6e36fe047fa66c1f86e67004f6cc3ecb2a5e5ff0066694257ac98fae6  diffusion_pytorch_model-00006-of-00008.safetensors
-c3dd20329b34c13e79a3c5f6b23171434fea59131d98c37625a23bf339630707  diffusion_pytorch_model-00007-of-00008.safetensors
-48cc5e471d323e03bcb96ed39f8ea44b307a9f96b84327847c2656f40949e01e  diffusion_pytorch_model-00008-of-00008.safetensors
-2444bb18169f99786e3aafb3080a8cbd4998359300851c6570f90be57493e6ff  diffusion_pytorch_model.safetensors.index.json

merged/config.json

@@ -1,18 +0,0 @@
-{
-  "_class_name": "WanModelFast",
-  "_diffusers_version": "0.35.2",
-  "_name_or_path": "/lustre/fs1/portfolios/nvr/projects/nvr_elm_llm/users/qiyuanh/benchmark_experiments/recovery/rev_wbench_candidate_6c391161_parity_20260809/source/checkpoint/models/lingbot_world_fast/base",
-  "control_type": "cam",
-  "dim": 5120,
-  "eps": 1e-06,
-  "ffn_dim": 13824,
-  "freq_dim": 256,
-  "in_dim": 36,
-  "local_attn_size": -1,
-  "model_type": "i2v",
-  "num_heads": 40,
-  "num_layers": 40,
-  "out_dim": 16,
-  "sink_size": 0,
-  "text_len": 512
-}

merged/diffusion_pytorch_model-00001-of-00008.safetensors

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:97870275ccec0fa64a213e7909d3bae3d77cb2021ec37609d4c29127f1f05263
-size 4991624976

merged/diffusion_pytorch_model-00002-of-00008.safetensors

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:3a16767352aea3e7bbd58bc4e92488aac3e24d456131da7c553662ea7769fe3f
-size 4982358760

merged/diffusion_pytorch_model-00003-of-00008.safetensors

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:fad09faf0c9353f6803c5852fc2aaec7d28d4e5dc049b3db05c6c65221d10288
-size 4861666096

merged/diffusion_pytorch_model-00004-of-00008.safetensors

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:14cec26b8afe32622f885d3fd32f6732e17eedbee7466a2fe7478d51b2d94703
-size 4966568072

merged/diffusion_pytorch_model-00005-of-00008.safetensors

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:2411c8bfe88a18816f2195e859910e42786ee4791a5e1c63cfc50ba8d01a5f46
-size 4929878840

merged/diffusion_pytorch_model-00006-of-00008.safetensors

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:3a7db7a6e36fe047fa66c1f86e67004f6cc3ecb2a5e5ff0066694257ac98fae6
-size 4950774384

merged/diffusion_pytorch_model-00007-of-00008.safetensors

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:c3dd20329b34c13e79a3c5f6b23171434fea59131d98c37625a23bf339630707
-size 4982358936

merged/diffusion_pytorch_model-00008-of-00008.safetensors

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:48cc5e471d323e03bcb96ed39f8ea44b307a9f96b84327847c2656f40949e01e
-size 2423583064