fix: gemma3_text model loading vision config (#3354)

* fix: gemma3-text mode loading vision config

* fix: improve defaults to use lora kernels

examples/gemma3/gemma-3-1b-qlora.yml

@@ -1,6 +1,7 @@
 base_model: google/gemma-3-1b-it
 
 model_type: Gemma3ForCausalLM
+cls_model_config: Gemma3TextConfig
 
 # Automatically upload checkpoint and final model to HF
 # hub_model_id: username/custom_model_name
@@ -29,7 +30,7 @@ output_dir: ./outputs/out
 adapter: qlora
 lora_r: 32
 lora_alpha: 16
-lora_dropout: 0.05
+lora_dropout: 0
 lora_target_linear: true
 
 sequence_len: 2048

examples/gemma3/gemma-3-270m-qlora.yml

@@ -1,6 +1,7 @@
 base_model: google/gemma-3-270m-it
 
 model_type: Gemma3ForCausalLM
+cls_model_config: Gemma3TextConfig
 
 # Automatically upload checkpoint and final model to HF
 # hub_model_id: username/custom_model_name
@@ -29,7 +30,7 @@ output_dir: ./outputs/out
 adapter: qlora
 lora_r: 32
 lora_alpha: 16
-lora_dropout: 0.05
+lora_dropout: 0
 lora_target_linear: true
 
 sequence_len: 2048

examples/gemma3/gemma-3-4b-qlora.yml

@@ -2,6 +2,7 @@ base_model: google/gemma-3-4b-it
 
 # Need to set else transformers tries to load vision too
 model_type: Gemma3ForCausalLM
+cls_model_config: Gemma3TextConfig
 
 load_in_4bit: true
 
@@ -32,8 +33,8 @@ sample_packing: true
 
 lora_r: 32
 lora_alpha: 16
-lora_dropout: 0.05
-lora_target_modules: 'model.language_model.layers.[\d]+.(mlp|cross_attn|self_attn).(up|down|gate|q|k|v|o)_proj'
+lora_dropout: 0
+lora_target_linear: true
 
 wandb_project:
 wandb_entity:

examples/gemma3/gemma-3-4b-vision-qlora.yml

@@ -31,7 +31,7 @@ pad_to_sequence_len: false
 
 lora_r: 32
 lora_alpha: 16
-lora_dropout: 0.05
+lora_dropout: 0
 lora_target_modules: 'model.language_model.layers.[\d]+.(mlp|cross_attn|self_attn).(up|down|gate|q|k|v|o)_proj'
 
 wandb_project: