FLUX on Forge – “You do not have CLIP state dict!” even with clip_g.safetensors (RTX 3060 12GB)

[damianmaz1992-gif]

Details:
Hi! I’m trying to run FLUX (flux1-dev.safetensors) on Stable Diffusion Forge and I keep getting:
AssertionError: You do not have CLIP state dict!
You do not have CLIP state dict!
What I’ve done:
Text encoders:
models\text_encoder\t5xxl_fp16.safetensors (selected in UI)
clip_g.safetensors (OpenCLIP ViT-G/14) copied to:
models\text_encoder\clip_g.safetensors
models\text_encoders\clip_g.safetensors
models\Stable-diffusion\clip_g.safetensors (next to flux1-dev.safetensors)
Vision encoder:
models\clip_vision\CLIP-ViT-bigG-14-laion2B-39B-b160k.safetensors
Restarted Forge + Settings → Unload all models.
UI: Checkpoint = flux1-dev.safetensors, WaI/Text Encoder = t5xxl_fp16.safetensors.
Log tail:
File "backend/loader.py", line 63, in load_huggingface_component
raise AssertionError('You do not have CLIP state dict!')
AssertionError: You do not have CLIP state dict!
Env:
Fresh Forge (latest commit)
RTX 3060 12GB (CUDA native)
Torch/xformers from Forge setup
Windows 11
Questions:
Exact path/name Forge expects for OpenCLIP ViT-G/14 when using FLUX?
Any specific clip_g.safetensors build required (or fallback clip_l.safetensors)?
Any UI flags/args needed to force FLUX to pick up clip_g?
Thanks a lot! Happy to share full logs or a short screen recording.

[triasbrata]

it because some flux model dont include clip and VAE. to use it you need to use the download the clip and VAE separately. here example https://civitai.com/models/1799857?modelVersionId=2287992

hope this help mate

[Haoming02]

Flux uses T5XXL and Clip-L, not Clip-G