Document fl2va_pruned_nvfp4

README.md

@@ -14,7 +14,8 @@ base_model_relation: quantized
 
 # MiniMax H3 — NVFP4
 
-NVFP4 quantizations of the **MiniMax-H3** ref2va diffusion transformer for ComfyUI.
+NVFP4 quantizations of the **MiniMax-H3** diffusion transformer for ComfyUI — both task
+variants, `ref2va` (reference images → video) and `fl2va` (first/last frame → video).
 
 - Original model: https://huggingface.co/MiniMaxAI/MiniMax-H3
 - ComfyUI repackage these derive from: https://huggingface.co/Comfy-Org/MiniMax-H3
@@ -24,15 +25,26 @@ Hopper or older the NVFP4 path is emulated — use Comfy-Org's `int8_convrot` fi
 
 ## Which file do I want?
 
+First pick the **task**, then the file:
+
+- **`ref2va`** — up to 9 reference images (+ videos/audio) → video. Identity-driven.
+- **`fl2va`** — first and/or last frame → video. Interpolates between endpoints; also how
+  you chain clips past one generation by feeding the previous clip's last frame in.
+
 | file | size | s/it | VRAM (DiT) | notes |
 |---|---|---|---|---|
-| **`minimax_h3_ref2va_pruned_nvfp4.safetensors`** | **12.5 GB** | **1.90** | 11.9 GB | **recommended** |
+| **`minimax_h3_ref2va_pruned_nvfp4.safetensors`** | **12.5 GB** | **1.90** | 11.9 GB | **recommended, ref2va** |
+| **`minimax_h3_fl2va_pruned_nvfp4.safetensors`** | **12.5 GB** | — | ~11.9 GB | **recommended, fl2va** |
 | `minimax_h3_ref2va_nvfp4_mixed.safetensors` | 24.4 GB | 1.92 | ~20 GB | from unpruned bf16 |
 | `minimax_h3_ref2va_nvfp4_full.safetensors` | 18.7 GB | 1.91 | ~16 GB | experimental |
 
-Take **`pruned_nvfp4`** unless you have a specific reason not to: half the size of the
+Take a **`pruned_nvfp4`** file unless you have a specific reason not to: half the size of the
 alternatives at identical speed, and it leaves the modulation path at full precision.
 
+The two pruned files are the same size and were produced by the same script over the same
+200 quantized layers — they differ only in task head. `fl2va` s/it is unmeasured but should
+match `ref2va`; the architectures are identical.
+
 ## Why the pruned base is the right one to quantize
 
 Comfy-Org's `pruned` checkpoint is not lossily pruned — it is a **structural refactor of
@@ -156,7 +168,7 @@ reference images, and 3-4 varied shots hold identity far better than one.
 | `vae/minimax_h3_audio_vae_fp32.safetensors` | FP32 | **Comfy-Org, unmodified** |
 
 **The NVFP4 text encoder is Comfy-Org's work, not ours.** Only the
-`minimax_h3_ref2va_*nvfp4*.safetensors` diffusion models here are new.
+`minimax_h3_*_nvfp4*.safetensors` diffusion models here are new.
 
 VAEs are deliberately **not** quantized: they are small, run once per generation rather than
 per step, and decode straight to pixels and audio samples where error is immediately
@@ -167,7 +179,7 @@ visible. The text encoder being NVFP4 buys VRAM, not speed — it also runs once
 ```
 📂 ComfyUI/models/
 ├── 📂 diffusion_models/
-│   └── minimax_h3_ref2va_pruned_nvfp4.safetensors
+│   └── minimax_h3_ref2va_pruned_nvfp4.safetensors   (or the fl2va file)
 ├── 📂 text_encoders/
 │   └── qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors
 └── 📂 vae/
@@ -177,18 +189,21 @@ visible. The text encoder being NVFP4 buys VRAM, not speed — it also runs once
 
 Recommended stack total: **33.7 GB**. Use the official
 [R2V template](https://github.com/Comfy-Org/workflow_templates/blob/main/templates/video_minimax_h3_r2v.json),
-swapping the diffusion model. Requires ComfyUI >= 0.30.0 (native H3 support in
+swapping the diffusion model — or the
+[FL2V template](https://github.com/Comfy-Org/workflow_templates/blob/main/templates/video_minimax_h3_fl2v.json)
+for the `fl2va` file. Requires ComfyUI >= 0.30.0 (native H3 support in
 `comfy/ldm/minimax/`). `CLIPLoader` type must be `minimax`; sampler `res_multistep`;
 frame `length` must satisfy 17n+5.
 
 ## Honest limitations
 
 - Quality was compared against `pruned_int8_convrot` at three matched seeds with no visible
-  degradation, but this is **not a rigorous evaluation** — no FVD, no human study, no
-  long-duration or 2K testing.
+  degradation, but this is **not a rigorous evaluation** — no FVD, no human study, no 2K
+  testing. 15s (362 frames) generates fine at 480x864.
+- The **`fl2va` file is unbenchmarked** — it was produced by the same script over the same
+  layer set as `ref2va`, but has not been run end-to-end.
 - `pruned_nvfp4` is **doubly quantized** (bf16 -> int8_convrot by Comfy-Org -> NVFP4 here).
   Error from both passes compounds. It held up in testing, but that is a real caveat.
-- Only **ref2va** is converted. `fl2va` is not included.
 - Benchmarks are single-GPU, one card, one resolution.
 
 Failure cases are welcome in the discussions tab — concrete artifacts beat aggregate scores.