library_name: diffusers (match 4-step card, enable download tracking); remove provenance.json and private-era docs

Documentation/metadata only: no weights, scheduler configs, modular_model_index.json or fastvideo_inference.json changed.

INFERENCE.md

@@ -1,53 +0,0 @@
-# FastH3 8-Step V2 inference
-
-## Compatibility status
-
-FastVideo [PR #1852](https://github.com/hao-ai-lab/FastVideo/pull/1852) adds
-the runtime support this checkpoint needs: scheduler shifts are read from the
-checkpoint (video 10, audio 3 here; base H3 keeps 12/3), the eight-rung DMD
-ladder is loaded from `fastvideo_inference.json`, and a new
-`examples/inference/basic/basic_fasth3_8step.py` example pins the recipe.
-Public `main` before that PR hard-codes video shift 12 and cannot run this
-checkpoint unchanged.
-
-## Exact sampling contract
-
-- Unshifted student inputs: `[999, 874, 749, 624, 500, 375, 250, 125]`.
-- The terminal clean endpoint is zero: eight model evaluations, nine grid nodes.
-- Video shift: `10`; audio shift: `3`; guidance: `1.0`.
-- Backend: `VIDEO_SPARSE_ATTN_H3`; sparsity: `0.8`; tile size: `64`.
-- Sampling: ODE updates, without stochastic re-noising.
-- Full H3 video and audio VAEs; preserve the output audio track.
-
-`--steps 9` alone is not equivalent to the explicit trained ladder.
-The environment variable below supplies the exact integer rungs; the two
-scheduler shifts must also be applied. `fastvideo_inference.json` records
-the same settings; verify that your launcher applies them rather than
-assuming its defaults match this file.
-
-## Command
-
-On the PR #1852 branch (or `main` once merged), on a GPU worker:
-
-```bash
-python examples/inference/basic/basic_fasth3_8step.py \
-  --prompt 'integrated_multimodal_description: A red fox runs through fresh snow at dawn. overall_soundscape: Fast pawsteps in snow, winter wind, and distant birds.' \
-  --height 768 --width 1344 --num-frames 124 \
-  --num-gpus 4 --vsa-kernel sm100a \
-  --profile strict --no-inference-torch-compile --no-compile-vae \
-  --output outputs/fasth3-v2 \
-  --repeats 1
-```
-
-No environment variables are needed: the example reads the ladder and shifts
-from the checkpoint and rejects any `--steps` other than 9.
-
-This example requests 1344×768, 124 frames at 24 fps (about 5.17 seconds),
-and writes MP4s with a stereo 32 kHz audio track under `outputs/fasth3-v2/`. The `sm100a` route needs a compatible
-Blackwell GPU and kernel build. Triton is a different kernel option, not a
-license to change the trained sparsity, tile geometry, or shifts.
-
-Verified on the PR #1852 head on 4x GB200: eight student forwards,
-video/audio shifts 10/3, VSA 0.8 tile 64, and a playable MP4 with audio
-(832x480, 124 frames). Report any latency with the exact hardware,
-resolution, frame count, attention path and warmup policy.

README.md

@@ -3,7 +3,7 @@ license: other
 license_name: minimax-h3-community
 license_link: LICENSE
 base_model: MiniMaxAI/MiniMax-H3
-library_name: fastvideo
+library_name: diffusers
 pipeline_tag: text-to-video
 tags:
 - text-to-video

RELEASE_CHECKLIST.md

@@ -1,49 +0,0 @@
-# FastH3 8-Step V2 release checklist
-
-Prepared 2026-09-08; updated 2026-09-15 after the repository was made public and FastVideo PR #1852 was opened.
-
-## Verified packaging
-
-- [x] Canonical repo: `FastVideo/FastVideo-FastH3-8-Step-V2`; private visibility retained.
-- [x] Model card identifies the step-1300 V14 data-free DMD2 model, VSA80/tile64,
-  eight forwards and video/audio shifts 10/3.
-- [x] FastVideo logo, blog, code, method citations and linked acknowledgements included.
-- [x] All indexed shards exist: 14 transformer, 14 text-encoder, 3 video-VAE;
-  the single-file audio VAE is also present.
-- [x] All 14 student shard sizes and SHA-256 values in `checkpoint_content.json`
-  match the Hub file/LFS metadata. No large weight download was required.
-- [x] The preserved `checkpoint_metadata.json` SHA-256 matches provenance.
-- [x] `LICENSE` matches the pinned MiniMax H3 base revision exactly.
-- [x] Added the MiniMax redistribution `NOTICE` and Qwen3-VL Apache-2.0 license.
-- [x] Model card and embedded repository references use the V2 name.
-
-The audited package revision before these documentation changes was
-`a001692841117d7d04707951db28b12187f81c66` (67 files, 147,870,260,063 bytes).
-The student contains 688 bf16 tensors. Documentation updates must not alter
-checkpoint weights, scheduler settings or original provenance hashes.
-
-## Required before publishing
-
-- [x] Land or provide a reviewed public FastVideo inference path for video shift
-  10 and the exact eight-rung ladder. Open as FastVideo PR #1852 (not yet merged). Public `main` at
-  `a943220c115228ade5d57b3bab9a6a87fd600a10` still hard-codes video shift 12;
-  the old README's scheduler-shift CLI options are absent there.
-- [x] Replace the compatibility hold with a verified public quickstart
-  (`basic_fasth3_8step.py`, PR #1852).
-- [x] Run one end-to-end GPU generation using the PR #1852 head; confirmed
-  eight forwards, shifts 10/3, VSA80/tile64 and playable video with audio
-  (4x GB200, 832x480x124f). Re-verify on `main` after merge.
-- [ ] Inspect representative high-motion samples. Do not copy the four-step
-  models' evaluation results or speed claims onto this checkpoint.
-- [ ] Complete author/legal review of the MiniMax H3 territory, redistribution
-  and commercial terms, including any required separate authorization.
-  Adding license notices is not legal clearance for worldwide distribution.
-- [ ] Review the historical cluster paths and run identifiers in the preserved
-  provenance files before publication. No credentials were found in the
-  reviewed metadata, but provenance sanitization would require updating the
-  corresponding integrity references deliberately.
-- [x] Repository made public by the author on 2026-09-15.
-
-Qwen license source:
-`https://github.com/QwenLM/Qwen3-VL/blob/96588727e44c78b25ba03ea03b8e12f7e64fd0da/LICENSE`.
-Runtime details and the compatible private-review command are in `INFERENCE.md`.

provenance.json

@@ -1,33 +0,0 @@
-{
-  "attention_backend": "VIDEO_SPARSE_ATTN_H3",
-  "audio_scheduler_shift": 3.0,
-  "base_model": "hf://MiniMaxAI/MiniMax-H3@9bfb6693f2cf6de171db46d1aa586f67d773a1da",
-  "checkpoint_content_sha256": "516323fa396fa5dff4e82669d4e9a08a5791692a3d3b98ff6bc3de3fc6a33d11",
-  "checkpoint_metadata_sha256": "ca9f2d609c05742ba465d24989981ec02cca26acb6ca2f163dc0f6dc8d11c27b",
-  "checkpoint_step": 1300,
-  "data_commit": "b92be606a439c67006220459df236d18c48fd38c3e11f4771d21fe0d078a3371",
-  "dmd_denoising_steps": [
-    999,
-    874,
-    749,
-    624,
-    500,
-    375,
-    250,
-    125
-  ],
-  "fastvideo_commit": "24bbe7fddd05ca6f2c34b3dbed06ac1c75b72086",
-  "lineage_parent_run_id": "20260829T023805Z-h3-8step-datafree-vsa80-vshift8",
-  "model_id": "FastVideo/FastVideo-FastH3-8-Step-V2",
-  "schema_version": "fasth3-checkpoint-provenance-v1",
-  "source_config_sha256": "92a4c5f2ece19d71929903d7cc094eecada3a768c9a8f846001b6610e02ff6df",
-  "student_dtype": "bfloat16",
-  "student_tensor_bytes": 70099502592,
-  "student_tensor_count": 688,
-  "training_mode": "data-free",
-  "training_run_id": "20260829T210445Z-h3-v14-8step-datafree-vsa80-vshift10",
-  "video_scheduler_shift": 10.0,
-  "vsa_sparsity": 0.8,
-  "vsa_tile_size": 64,
-  "wandb_url": "https://wandb.ai/wlsaidhi/h3-dmd2-vsa/runs/euce1zuw"
-}