Wan training: Set default dropout to 0.0 in Wan configs

Co-authored-by: martinarroyo <martinarroyo@google.com>

Author: ninatu

src/maxdiffusion/configs/base_wan_14b.yml

@@ -72,7 +72,7 @@ mask_padding_tokens: True
 # 2. attention_sharding_uniform = False : Heads are sharded uniformly across devices for self attention while sequence is sharded
 #    in cross attention q.
 attention_sharding_uniform: True 
-dropout: 0.1
+dropout: 0.0
 
 flash_block_sizes: {
   "block_q" : 512,

src/maxdiffusion/configs/base_wan_1_3b.yml

@@ -72,7 +72,7 @@ mask_padding_tokens: True
 # 2. attention_sharding_uniform = False : Heads are sharded uniformly across devices for self attention while sequence is sharded
 #    in cross attention q.
 attention_sharding_uniform: True
-dropout: 0.1
+dropout: 0.0
 
 flash_block_sizes: {
   "block_q" : 512,

src/maxdiffusion/configs/base_wan_27b.yml

@@ -71,7 +71,7 @@ mask_padding_tokens: True
 # 2. attention_sharding_uniform = False : Heads are sharded uniformly across devices for self attention while sequence is sharded
 #    in cross attention q.
 attention_sharding_uniform: True 
-dropout: 0.1
+dropout: 0.0
 
 flash_block_sizes: {
   "block_q" : 512,

src/maxdiffusion/configs/base_wan_i2v_14b.yml

@@ -62,7 +62,7 @@ from_pt: True
 split_head_dim: True
 attention: 'flash' # Supported attention: dot_product, flash, cudnn_flash_te, ring
 flash_min_seq_length: 4096
-dropout: 0.1
+dropout: 0.0
 
 # If mask_padding_tokens is True, we pass in segment ids to splash attention to avoid attending to padding tokens.
 # Else we do not pass in segment ids and on vpu bound hardware like trillium this is faster.

src/maxdiffusion/configs/base_wan_i2v_27b.yml

@@ -62,7 +62,7 @@ from_pt: True
 split_head_dim: True
 attention: 'flash' # Supported attention: dot_product, flash, cudnn_flash_te, ring
 flash_min_seq_length: 4096
-dropout: 0.1
+dropout: 0.0
 
 # If mask_padding_tokens is True, we pass in segment ids to splash attention to avoid attending to padding tokens.
 # Else we do not pass in segment ids and on vpu bound hardware like trillium this is faster.