Source: <a target="_blank" rel="ugc" href="https://huggingface.co/Freepik/flux.1-lite-8B-alpha">https://huggingface.co/Freepik/flux.1-lite-8B-alpha</a> from <a target="_blank" rel="ugc" href="https://huggingface.co/Freepik">Freepik</a> The alpha release of Flux.1 Lite, an 8B parameter transformer model distilled from the FLUX.1-dev model. This version uses 7 GB less RAM and runs 23% faster while maintaining the same precision (bfloat16) as the original model. <a target="_blank" rel="ugc" href="https://civarchive.com/models/889791/flux1-lite-8b-alpha-gguf-q3ks-q4q41q4ks-q5q51q5ks-q6k-q8-bf16">Download GGUF version here</a>.<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/79857332-8c1c-41a4-8edd-47f22b630cd4/width=525/79857332-8c1c-41a4-8edd-47f22b630cd4.jpeg" />☕ Buy me a coffee: <a target="_blank" rel="ugc" href="https://ko-fi.com/ralfingerai">https://ko-fi.com/ralfingerai</a> 🍺 Join my discord: <a target="_blank" rel="ugc" href="https://discord.com/invite/pAz4Bt3rqb">https://discord.com/invite/pAz4Bt3rqb</a>

Flux.1 Lite 8B Alpha (Freepik Company)

For the best results, we strongly recommend using a <code>guidance_scale</code> of 3.5 and setting <code>n_steps</code> between 22 and 30.

flux1Lite8BAlpha_flux1Lite8BAlpha.safetensors

flux.1-lite-8B-alpha

Comfy Workflow

Flux.1 LiteWe are thrilled to announce the alpha release of Flux.1 Lite, an 8B parameter transformer model distilled from the FLUX.1-dev model. This version uses 7 GB less RAM and runs 23% faster while maintaining the same precision (bfloat16) as the original model.Text-to-ImageFlux.1 Lite is ready to unleash your creativity! For the best results, we strongly recommend using a guidance_scale of 3.5 and setting n_steps between 22 and 30.MotivationInspired by <a href="https://ostris.com/2024/09/07/skipping-flux-1-dev-blocks/" rel="ugc noopener noreferrer nofollow" target="_blank">Ostris</a> findings, we analyzed the mean squared error (MSE) between the input and output of each block to quantify their contribution to the final result, revealing significant variability.As Ostris pointed out, not all blocks contribute equally. While skipping just one of the early MMDiT or late DiT blocks can significantly impact model performance, skipping any single block in between does not have a significant impact over the final image quality.Future workStay tuned! Our goal is to distill FLUX.1-dev further until it can run smoothly on 24 GB consumer-grade GPU cards, maintaining its original precision (bfloat16), and running even faster, making high-quality AI models accessible to everyone.ComfyUIWe've also crafted a ComfyUI workflow to make using Flux.1 Lite even more seamless! Find it in comfy/flux.1-lite_workflow.json.The safetensors checkpoint is available here: <a href="https://huggingface.co/Freepik/flux.1-lite-8B-alpha/blob/main/flux.1-lite-8B-alpha.safetensors" rel="ugc noopener noreferrer nofollow" target="_blank">flux.1-lite-8B-alpha.safetensors</a>Try it out at Freepik!Our <a href="https://www.freepik.com/pikaso/ai-image-generator" rel="ugc noopener noreferrer nofollow" target="_blank">AI generator</a> is now powered by Flux.1 Lite!🔥 News 🔥<ul><li>Oct 23, 2024. Alpha 8B checkpoint is publicly available on <a href="https://huggingface.co/Freepik/flux.1-lite-8B-alpha" rel="ugc noopener noreferrer nofollow" target="_blank">HuggingFace Repo</a>.</li><li></li></ul>CitationIf you find our work helpful, please cite it!Attribution noticeThe FLUX.1 [dev] Model is licensed by Black Forest Labs. Inc. under the FLUX.1 [dev] Non-Commercial License. Copyright Black Forest Labs. Inc.Our model weights are released under the FLUX.1 [dev] Non-Commercial License.

Files with SHA256: f80efd41b489d4f7035533a11fa94b45ad824ded2a256c0263a6da240f4872e4