Update README.md

README.md

@@ -1,3 +1,65 @@
----
-license: mit
----
+---
+license: mit
+language:
+- en
+base_model:
+- CompVis/stable-diffusion-v1-4
+pipeline_tag: text-to-image
+library_name: diffusers
+tags:
+- stable-diffusion
+- fine-tuned
+- unet
+- flickr8k
+- generative-ai
+- text-to-image
+---
+
+# 🖼️ Fine-Tuned U-Net (Flickr8k) — Stable Diffusion
+
+This model contains a **fine-tuned U-Net** from the `CompVis/stable-diffusion-v1-4` Stable Diffusion pipeline, trained using natural English captions from the **Flickr8k** dataset. It enhances generation quality for everyday, human-centered scenarios like actions, objects, and environmental scenes.
+
+> ✅ Only the U-Net was fine-tuned. The VAE, tokenizer, and text encoder remain from the original base model.
+
+---
+
+## 📊 Training Details
+
+- **Base model**: [`CompVis/stable-diffusion-v1-4`](https://huggingface.co/CompVis/stable-diffusion-v1-4)  
+- **Fine-tuned on**: [Flickr8k Kaggle Dataset](https://www.kaggle.com/datasets/adityajn105/flickr8k)  
+- **Components fine-tuned**: `unet` only  
+- **Frozen**: text encoder, VAE, and tokenizer  
+- **Epochs**: 10  
+- **Learning rate**: 1e-6  
+- **Batch size**: 1 (with gradient accumulation = 16 → effective batch size ≈ 16)  
+- **Image resolution**: 256×256  
+- **Training size**: 1000 image-caption pairs  
+- **Mixed precision**: FP16  
+- **Gradient Accumulation Steps**: 16  
+- **Trained on**: Kaggle GPU (Tesla T4, 16GB VRAM)  
+- **Seed**: 42  
+- **Checkpointing**: every 200 steps
+
+---
+
+## 🧠 Usage
+
+This U-Net can be loaded into a standard Stable Diffusion pipeline to enhance image generation on descriptive prompts:
+
+```python
+from diffusers import StableDiffusionPipeline, UNet2DConditionModel
+import torch
+
+# Load fine-tuned U-Net
+unet = UNet2DConditionModel.from_pretrained("srishticrai/unet-flickr8k")
+
+# Load pipeline with original components + fine-tuned U-Net
+pipe = StableDiffusionPipeline.from_pretrained(
+    "CompVis/stable-diffusion-v1-4",
+    unet=unet,
+    torch_dtype=torch.float16
+).to("cuda")
+
+# Generate image
+image = pipe("A child blowing bubbles in a park at sunset").images[0]
+image.show()