LTX 2.3 Full model (42GB) works on a 5090. How?
Migrated from LTX-2.3 notes: 2 extracted reference links.
Original discussion: https://www.reddit.com/r/StableDiffusion/comments/1rna2ts/ltx_23_full_model_42gb_works_on_a_5090_how/
Referenced links:
- https://civitai.com/models/2304098/ltx-2-19b-gguf-12gb-comfyui-workflows-5-total-t2vi2vv2via2vta2v — Haven’t played around with 2.3 yet. I was using the 19B GGUF model. The same variant as this Workflow. I think it was Q4. …
- https://old.reddit.com/r/StableDiffusion/comments/1p7bs1o/vram_ram_offloading_performance_benchmark_with/ — With current image/video model architectures you’re not limited by bandwidth like in LLM’s but compute. Some benchmarks here: …