Deprecate #36741 and map Causal to Conditional (#36917)

* deprecate the prev fix

* reword warning and update docs

* reword warning

* tests

* dont bloat `get_text_config()`

docs/source/en/model_doc/gemma3.md

@@ -204,7 +204,7 @@ visualizer("<img>What is shown in this image?")
     +   do_pan_and_scan=True,
         ).to("cuda")
     ```
-- For text-only inputs, use [`AutoModelForCausalLM`] instead to skip loading the vision components and save resources.
+- For Gemma-3 1B checkpoint trained in text-only mode, use [`AutoModelForCausalLM`] instead.
 
     ```py
     import torch