fix: multilingual midel convert to tflite get wrong token (huggingfac…

…e#32079)

* fix: multilingual midel convert to tflite get wrong token

* fix: modify test_force_tokens_logits_processor the checking value as scores.dtype.min

---------

Co-authored-by: kent.sc.hung <kent.sc.hung@benq.com>
Co-authored-by: Aya <[kent831217@gmail.com]>

src/transformers/generation/tf_logits_process.py

@@ -581,7 +581,7 @@ def _force_token(generation_idx):
             batch_size = scores.shape[0]
             current_token = self.force_token_array[generation_idx]
 
-            new_scores = tf.ones_like(scores, dtype=scores.dtype) * -float("inf")
+            new_scores = tf.zeros_like(scores, dtype=scores.dtype) + tf.constant([scores.dtype.min])
             indices = tf.stack((tf.range(batch_size), tf.tile([current_token], [batch_size])), axis=1)
             updates = tf.zeros((batch_size,), dtype=scores.dtype)
             new_scores = tf.tensor_scatter_nd_update(new_scores, indices, updates)

tests/generation/test_tf_logits_process.py

@@ -406,7 +406,12 @@ def test_force_tokens_logits_processor(self, use_xla):
 
         non_forced_inds = [i for i in range(vocab_size) if i != force_token_map[cur_len]]
         self.assertTrue(
-            tf.math.reduce_all(tf.math.is_inf(tf.gather(scores, [non_forced_inds], axis=1))),
+            tf.math.reduce_all(
+                tf.experimental.numpy.isclose(
+                    tf.gather(scores, [non_forced_inds], axis=1),
+                    tf.constant(scores.dtype.min),
+                )
+            )
         )
 
         # check that if the cur_len is not contained in the force_token_map, the logits are not modified