<h1 id="full-checkpoint-with-improved-te-do-not-load-additional-clipte-0h7bq339k">Full Checkpoint with improved TE do not load additional CLIP/TE</h1><h2 id="sd-3.5-medium-+-google-flan-6odx5z8r9">SD 3.5 Medium + Google FLAN</h2><ul><li>Update your Comfy UI before using(10/29/2024)</li><li>FP16 merge with Google FLAN (From FP32)</li></ul><ul><li>SGM Uniform with Dpmpp_2M at 40 Steps or Euler Normal/Simple at 20 steps working well</li><li>Do not use negatives above 0.2 timestamp - If you do not understand this line load any image as a workflow. (The same instructions as base SD 3.5)</li></ul><ul><li>I posted a FP8 version but even with a 8GB card I do not see an improvement in speed so I still recommend the FP16 (NF4 support in Forge would be great)</li></ul>In my testing on a low VRAM card this model runs several times faster then the <a target="_blank" rel="ugc" href="https://civarchive.com/models/882666/sd35-large-google-flan">Large Model</a>Per the <a target="_blank" rel="ugc" href="https://www.apache.org/licenses/LICENSE-2.0">Apache 2.0 license</a> FLAN is attributed to Google

Files with SHA256: 41934cb384f58a3337f4c9b60b55563f2e148db8c01f880898fdfd4b3b9a877e