Release 0.12.0 (#282)

CHANGELOG.md

@@ -5,6 +5,34 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.12.0] - 2025-11-27
+
+# MFLUX v.0.12.0 Release Notes
+
+### 🎨 New Model Support
+
+- **Bria FIBO Support**: Added support for [FIBO](https://huggingface.co/briaai/FIBO), the first open-source JSON-native text-to-image model from [Bria.ai](https://bria.ai)
+- **Three operation modes**: Generate (text-to-image with VLM expansion), Refine (structured prompt editing), and Inspire (image-to-prompt extraction)
+- **New commands**:
+  - `mflux-generate-fibo` - Generate images from text prompts with VLM-guided JSON expansion
+  - `mflux-refine-fibo` - Refine images using structured JSON prompts for targeted attribute editing
+  - `mflux-inspire-fibo` - Extract structured prompts from reference images for style transfer and remixing
+- **VLM-guided JSON prompting**: Automatically expands short text prompts into 1,000+ word structured schemas using a fine-tuned Qwen3-VL model
+
+### 🔧 Restructure and 🔄 Breaking Changes
+
+- **Common module reorganization**: Moved shared functionality to `models/common/` for better code reuse
+  - Unified latent creators across model families
+  - Centralized scheduler implementations
+  - Common quantization utilities
+  - Shared model saving functionality
+
+### 👩‍💻 Contributors
+
+- **Filip Strand (@filipstrand)**: FIBO model implementation, architecture, core development
+
+---
+
 ## [0.11.1] - 2025-11-13
 
 # MFLUX v.0.11.1 Release Notes

pyproject.toml

@@ -17,7 +17,7 @@ source-exclude = [
 
 [project]
 name = "mflux"
-version = "0.12.0.dev0"
+version = "0.12.0"
 description = "A MLX port of FLUX based on the Huggingface Diffusers implementation."
 readme = "README.md"
 keywords = ["diffusers", "flux", "mlx"]
@@ -30,7 +30,7 @@ dependencies = [
     "filelock>=3.18.0",
     "huggingface-hub>=0.24.5,<1.0",
     "matplotlib>=3.9.2,<4.0",
-    "mlx>=0.27.0,<0.30.0",
+    "mlx>=0.27.0,<0.31.0",
     "numpy>=2.0.1,<3.0",
     "opencv-python>=4.10.0,<5.0",
     "piexif>=1.1.3,<2.0",

src/mflux/models/fibo/variants/txt2img/fibo.py

@@ -88,18 +88,18 @@ class FIBO(nn.Module):
         for t in time_steps:
             try:
                 # 3.t Predict the noise
-                noise_pred = self.transformer(
+                noise = self.transformer(
                     t=t,
                     config=runtime_config,
                     hidden_states=latents,
                     encoder_hidden_states=encoder_hidden_states,
                     text_encoder_layers=text_encoder_layers,
                 )
-                noise_pred = FIBO._apply_classifier_free_guidance(noise_pred, runtime_config.guidance)
+                noise = FIBO._apply_classifier_free_guidance(noise, runtime_config.guidance)
 
                 # 4.t Take one denoise step
                 latents = runtime_config.scheduler.step(
-                    model_output=noise_pred,
+                    model_output=noise,
                     timestep=t,
                     sample=latents,
                 )
@@ -155,10 +155,10 @@ class FIBO(nn.Module):
         )
 
     @staticmethod
-    def _apply_classifier_free_guidance(noise_pred: mx.array, guidance: float) -> mx.array:
-        half = noise_pred.shape[0] // 2
-        noise_uncond = noise_pred[:half]
-        noise_text = noise_pred[half:]
+    def _apply_classifier_free_guidance(noise: mx.array, guidance: float) -> mx.array:
+        half = noise.shape[0] // 2
+        noise_uncond = noise[:half]
+        noise_text = noise[half:]
         return noise_uncond + guidance * (noise_text - noise_uncond)
 
     @staticmethod