Upload folder using huggingface_hub

README.md

@@ -1,133 +0,0 @@
----
-base_model:
-  - Qwen/Qwen-Image-Edit-2511
-base_model_relation: quantized
-tags:
-- dfloat11
-- df11
-- lossless compression
-- 70% size, 100% accuracy
----
-
-# DFloat11 Compressed Model: `Qwen/Qwen-Image-Edit-2511`
-
-This is a **DFloat11 losslessly compressed** version of the original `Qwen/Qwen-Image-Edit-2511` model. It reduces model size by **32%** compared to the original BFloat16 model, while maintaining **bit-identical outputs** and supporting **efficient GPU inference**.
-
-🔥🔥🔥 Thanks to DFloat11 compression, Qwen-Image-Edit-2511 can now run on **a single 32GB GPU**, or on **a single 24GB GPU with CPU offloading**, while maintaining full model quality. 🔥🔥🔥
-
-### Qwen-Image-Edit-2511
-
-Thanks, DFloat11 team - I have recoverted the weights for the 2511 model using the DFloat11 package, and done minimal testing which produced valid images. The rest of this readme is taken from https://huggingface.co/DFloat11/Qwen-Image-Edit-2509-DF11/blob/main/README.md
-
-### 📊 Performance Comparison
-
-(Note: numbers are pulled from DFloat11/Qwen-Image-Edit-2509 but not re-validated for the -2511 version)
-
-| Model                                               | Model Size | Peak GPU Memory (1024x1024 image generation) | Image Editing Time (A100 GPU) |
-|-----------------------------------------------------|------------|----------------------------------------------|-------------------------------|
-| Qwen-Image-Edit-2511 (BFloat16)                     | ~41 GB     | OOM                                          | -                             |
-| Qwen-Image-Edit-2511 (DFloat11)                     | 28.43 GB   | 30.20 GB                                     | 102 seconds                   |
-
-### 🔧 How to Use
-
-1. Install or upgrade the DFloat11 pip package *(installs the CUDA kernel automatically; requires a CUDA-compatible GPU and PyTorch installed)*:
-
-    ```bash
-    pip install -U dfloat11[cuda12]
-    ```
-
-2. Install or upgrade diffusers:
-
-    ```bash
-    pip install git+https://github.com/huggingface/diffusers
-    ```
-
-3. Save the following code to a Python file `qwen_image_edit.py`:
-
-    ```python
-    import os
-    import torch
-    import argparse
-    from diffusers import QwenImageEditPlusPipeline
-    from diffusers.utils import load_image
-    from dfloat11 import DFloat11Model
-
-    parser = argparse.ArgumentParser(description="Qwen Image Edit with DFloat11")
-    parser.add_argument("--image", default="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/cat.png", help="Image URL or path")
-    parser.add_argument("--prompt", default="Make this cat an astronaut gazing at planet earth from space", help="Edit prompt")
-    parser.add_argument("--output", default="qwen_image_edit_output.png", help="Output image path")
-    parser.add_argument("--steps", type=int, default=40, help="Number of inference steps")
-    parser.add_argument("--seed", type=int, default=42, help="Random seed")
-    parser.add_argument("--true_cfg_scale", type=float, default=4.0, help="True CFG scale")
-    parser.add_argument("--negative_prompt", default=" ", help="Negative prompt")
-    parser.add_argument("--guidance_scale", type=float, default=1.0, help="Guidance scale")
-    parser.add_argument("--cpu_offload", action="store_true", help="Enable CPU offloading")
-    parser.add_argument("--cpu_offload_blocks", type=int, default=20, help="Number of blocks to offload to CPU for block swapping")
-    parser.add_argument("--cpu_offload_no_pin_memory", action="store_true", help="Disable memory pinning for CPU offloading")
-    args = parser.parse_args()
-
-    pipeline = QwenImageEditPlusPipeline.from_pretrained("Qwen/Qwen-Image-Edit-2511", torch_dtype=torch.bfloat16)
-    DFloat11Model.from_pretrained(
-        "DFloat11/Qwen-Image-Edit-2511-DF11",
-        bfloat16_model=pipeline.transformer,
-        device="cpu",
-        cpu_offload=args.cpu_offload,
-        cpu_offload_blocks=args.cpu_offload_blocks,
-        pin_memory=not args.cpu_offload_no_pin_memory,
-    )
-    pipeline.enable_model_cpu_offload()
-
-    image = load_image(args.image)
-    inputs = {
-        "image": [image],
-        "prompt": args.prompt,
-        "generator": torch.manual_seed(args.seed),
-        "true_cfg_scale": args.true_cfg_scale,
-        "negative_prompt": args.negative_prompt,
-        "num_inference_steps": args.steps,
-        "guidance_scale": args.guidance_scale,
-        "num_images_per_prompt": 1,
-    }
-    with torch.inference_mode():
-        output = pipeline(**inputs)
-        output_image = output.images[0]
-        output_image.save(args.output)
-        print("Image saved at", os.path.abspath(args.output))
-
-    max_memory = torch.cuda.max_memory_allocated()
-    print(f"Max memory: {max_memory / (1000 ** 3):.2f} GB")
-    ```
-
-4. To run without CPU offloading (32GB VRAM required):
-    ```bash
-    python qwen_image_edit.py
-    ```
-
-    To run with CPU offloading (24GB VRAM required):
-    ```bash
-    python qwen_image_edit.py --cpu_offload
-    ```
-
-    If you are getting out-of-CPU-memory errors, try limiting the number of offloaded blocks or disabling memory-pinning:
-    ```bash
-    # Offload only 16 blocks (offloading more blocks uses less GPU memory and more CPU memory; offloading less blocks is faster):
-    python qwen_image_edit.py --cpu_offload --cpu_offload_blocks 16
-
-    # Disable memory-pinning (the most memory efficient way, but could be slower):
-    python qwen_image_edit.py --cpu_offload --no_pin_memory
-    ```
-
-
-### 🔍 How It Works
-
-We apply **Huffman coding** to losslessly compress the exponent bits of BFloat16 model weights, which are highly compressible (their 8 bits carry only ~2.6 bits of actual information). To enable fast inference, we implement a highly efficient CUDA kernel that performs on-the-fly weight decompression directly on the GPU.
-
-The result is a model that is **~32% smaller**, delivers **bit-identical outputs**, and achieves performance **comparable to the original** BFloat16 model.
-
-Learn more in our [research paper](https://arxiv.org/abs/2504.11651).
-
-### 📄 Learn More
-
-* **Paper**: [70% Size, 100% Accuracy: Lossless LLM Compression for Efficient GPU Inference via Dynamic-Length Float](https://arxiv.org/abs/2504.11651)
-* **GitHub**: [https://github.com/LeanModels/DFloat11](https://github.com/LeanModels/DFloat11)
-* **HuggingFace**: [https://huggingface.co/DFloat11](https://huggingface.co/DFloat11)

config.json

@@ -33,6 +33,5 @@
       ],
       "proj_out": []
     }
-  },
-  "model_type": "qwen2_5_vl"
+  }
 }

model-00001-of-00004.safetensors → model-00001-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8b23efd536428c19b48ac16d25595260b885dc860344355038d143834f0c5cd0
-size 10639517248
+oid sha256:78c7f6e9b3a70201edbfc7a15cf7d80462055754b383a0bfb442451f15cc7bad
+size 10457086553

model-00002-of-00004.safetensors → model-00002-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ff83b048d072a291d23c1cc668fd3b31fcf96a1ecfd48aded2e780d2b6214302
-size 10723635456
+oid sha256:3e64c21039a3a42409d56e9155360ebfd3a1004320265f67f62b6ddfedca1182
+size 10447227491

model-00004-of-00004.safetensors → model-00003-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d5052f220e2f9d16cd6f2fb6044a484079dad8e9364452a2becfd6810ab863ac
-size 8748967038
+oid sha256:1f7f3506365b20a45659121f17746e77186a43f19ea829ba11ea5ec58d007e36
+size 7498715375

model-00003-of-00004.safetensors

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:b580f17b8fdd62ce3f65ae9de60eef209d72c5c30f558cb4a2b69cb04fa04433
-size 10723614872