review comments and kv cache compression disable in fp

Author: eaidova

optimum/exporters/openvino/convert.py

@@ -29,7 +29,6 @@
 from optimum.exporters import TasksManager
 from optimum.exporters.utils import (
     DECODER_NAME,
-    DECODER_WITH_PAST_NAME,
     ENCODER_NAME,
     _get_submodels_for_export_encoder_decoder,
     get_diffusion_models_for_export,
@@ -48,7 +47,6 @@
     _transformers_version,
     compare_versions,
     is_diffusers_version,
-    is_openvino_version,
     is_openvino_tokenizers_version,
     is_tokenizers_version,
     is_transformers_version,
@@ -110,10 +108,13 @@ def _set_runtime_options(
             "diffusers" in library_name
             or "text-generation" in task
             or ("image-text-to-text" in task and model_name == "language_model")
+            or getattr(sub_export_config, "stateful", False)
         ):
             sub_export_config.runtime_options["ACTIVATIONS_SCALE_FACTOR"] = "8.0"
         if not quantized_model and (
-            "text-generation" in task or ("image-text-to-text" in task and model_name == "language_model")
+            "text-generation" in task
+            or ("image-text-to-text" in task and model_name == "language_model")
+            or getattr(sub_export_config, "stateful", False)
         ):
             sub_export_config.runtime_options["KV_CACHE_PRECISION"] = "f16"
 
@@ -643,7 +644,7 @@ def export_from_model(
     is_encoder_decoder = getattr(getattr(model, "config", {}), "is_encoder_decoder", False)
     model_type = getattr(getattr(model, "config", {}), "model_type", "")
     stateful = stateful and (
-        ensure_export_task_support_stateful(task, is_encoder_decoder) or ensure_model_type_support_stateful(model_type)
+        ensure_export_task_support_stateful(task) or ensure_model_type_support_stateful(model_type)
     )
 
     if stateful and is_encoder_decoder and not getattr(model, "_supports_cache_class", False):
@@ -1251,17 +1252,16 @@ def _get_encoder_decoder_stateful_models_for_export(
     all_variants = "\n".join([f"    - {name}: {description}" for name, description in export_config.VARIANTS.items()])
     logger.info(f"Using the export variant {export_config.variant}. Available variants are:\n{all_variants}")
 
-    models_for_export = _get_submodels_for_export_encoder_decoder(model, use_past=True)
+    models_for_export = _get_submodels_for_export_encoder_decoder(model, use_past=False)
 
     encoder_export_config = export_config.with_behavior("encoder")
     models_for_export[ENCODER_NAME] = (models_for_export[ENCODER_NAME], encoder_export_config)
 
     decoder_export_config_with_past = export_config.with_behavior("decoder", use_past=True, use_past_in_inputs=True)
 
     decoder_export_config_with_past.stateful = True
-    decoder_with_past_model = models_for_export.pop(DECODER_WITH_PAST_NAME)
     models_for_export[DECODER_NAME] = (
-        decoder_with_past_model,
+        models_for_export[DECODER_NAME],
         decoder_export_config_with_past,
     )
     return None, models_for_export

optimum/exporters/openvino/model_patcher.py

@@ -3700,4 +3700,4 @@ def patched_forward(*args, **kwargs):
 
         model.forward = patched_forward
 
-        super().__init__(config, model, model_kwargs)
+        super().__init__(config, model, model_kwargs)

optimum/intel/openvino/modeling_seq2seq.py

@@ -374,9 +374,9 @@ def forward(
         # Decode
         if past_key_values is None or self.decoder_with_past is None:
             decoder_outputs = self.decoder(
-                input_ids=decoder_input_ids[:, -1:]
-                if past_key_values is not None and self.use_cache
-                else decoder_input_ids,
+                input_ids=(
+                    decoder_input_ids[:, -1:] if past_key_values is not None and self.use_cache else decoder_input_ids
+                ),
                 past_key_values=past_key_values,
                 encoder_hidden_states=encoder_outputs.last_hidden_state,
                 encoder_attention_mask=attention_mask,