feat: Gemma4 text generation support (CORE-30) (#13376)

* initial gemma4 support

* parity with reference implementation

outputs can 100% match transformers with same sdpa flags, checkpoint this and then optimize

* Cleanup, video fixes

* cleanup, enable fused rms norm by default

* update comment

* Cleanup

* Update sd.py

* Various fixes

* Add fp8 scaled embedding support

* small fixes

* Translate think tokens

* Fix image encoder attention mask type

So it works with basic attention

* Handle thinking tokens different only for Gemma4

* Code cleanup

* Update nodes_textgen.py

* Use embed scale class instead of buffer

Slight difference to HF, but technically more accurate and simpler code

* Default to fused rms_norm

* Update gemma4.py

comfy/rmsnorm.py

@@ -3,6 +3,7 @@ import comfy.model_management
 
 RMSNorm = torch.nn.RMSNorm
 
+# Note: torch's fused F.rms_norm is faster but produces slightly different output than manual implementations (rsqrt/reduction rounding).
 def rms_norm(x, weight=None, eps=1e-6):
     if weight is None:
         return torch.nn.functional.rms_norm(x, (x.shape[-1],), eps=eps)