Replace deprecated batch_size with max_batch_size when using HybridCa…

…che (#35498)

* Replace deprecated batch_size with max_batch_size

- Functionality remains the same, because property getter batch_size(self) returned max_batch_size anyways.
- This change just avoids an unnecessary warning about deprecation.

* Use max_batch_size instead of deprecated batch_size with HybridCache

* Use max_batch_size instead of deprecated batch_size with HybridCache

- Change generated code to match original source

src/transformers/cache_utils.py

@@ -1640,9 +1640,9 @@ def __init__(
         )
         self.key_cache: List[torch.Tensor] = []
         self.value_cache: List[torch.Tensor] = []
-        global_cache_shape = (self.batch_size, self.num_key_value_heads, max_cache_len, self.head_dim)
+        global_cache_shape = (self.max_batch_size, self.num_key_value_heads, max_cache_len, self.head_dim)
         sliding_cache_shape = (
-            self.batch_size,
+            self.max_batch_size,
             self.num_key_value_heads,
             min(config.sliding_window, max_cache_len),
             self.head_dim,

src/transformers/models/cohere2/modeling_cohere2.py

@@ -579,7 +579,7 @@ def forward(
             batch_size, seq_len, _ = inputs_embeds.shape
             past_key_values = HybridCache(
                 self.config,
-                batch_size=batch_size,
+                max_batch_size=batch_size,
                 max_cache_len=seq_len,
                 device=self.device,
                 dtype=inputs_embeds.dtype,

src/transformers/models/cohere2/modular_cohere2.py

@@ -459,7 +459,7 @@ def forward(
             batch_size, seq_len, _ = inputs_embeds.shape
             past_key_values = HybridCache(
                 self.config,
-                batch_size=batch_size,
+                max_batch_size=batch_size,
                 max_cache_len=seq_len,
                 device=self.device,
                 dtype=inputs_embeds.dtype,

src/transformers/models/gemma2/modeling_gemma2.py

@@ -577,7 +577,7 @@ def forward(
             batch_size, seq_len, _ = inputs_embeds.shape
             past_key_values = HybridCache(
                 self.config,
-                batch_size=batch_size,
+                max_batch_size=batch_size,
                 max_cache_len=seq_len,
                 device=self.device,
                 dtype=inputs_embeds.dtype,

src/transformers/models/gemma2/modular_gemma2.py

@@ -403,7 +403,7 @@ def forward(
             batch_size, seq_len, _ = inputs_embeds.shape
             past_key_values = HybridCache(
                 self.config,
-                batch_size=batch_size,
+                max_batch_size=batch_size,
                 max_cache_len=seq_len,
                 device=self.device,
                 dtype=inputs_embeds.dtype,