feat: avoid validating return_multivector
Browse files
modeling_jina_embeddings_v4.py
CHANGED
|
@@ -357,7 +357,6 @@ class JinaEmbeddingsV4Model(Qwen2_5_VLForConditionalGeneration):
|
|
| 357 |
|
| 358 |
def _validate_encoding_params(
|
| 359 |
self,
|
| 360 |
-
return_multivector: Optional[bool] = None,
|
| 361 |
truncate_dim: Optional[int] = None,
|
| 362 |
prompt_name: Optional[str] = None,
|
| 363 |
) -> Dict[str, Any]:
|
|
@@ -374,9 +373,6 @@ class JinaEmbeddingsV4Model(Qwen2_5_VLForConditionalGeneration):
|
|
| 374 |
else PREFIX_DICT["query"]
|
| 375 |
)
|
| 376 |
|
| 377 |
-
return_multivector = return_multivector or False
|
| 378 |
-
encode_kwargs["return_multivector"] = return_multivector
|
| 379 |
-
|
| 380 |
truncate_dim = truncate_dim or self.config.truncate_dim
|
| 381 |
if truncate_dim is not None and truncate_dim not in self.config.matryoshka_dims:
|
| 382 |
raise ValueError(
|
|
@@ -429,9 +425,7 @@ class JinaEmbeddingsV4Model(Qwen2_5_VLForConditionalGeneration):
|
|
| 429 |
List of text embeddings as tensors or numpy arrays when encoding multiple texts, or single text embedding as tensor when encoding a single text
|
| 430 |
"""
|
| 431 |
prompt_name = prompt_name or "query"
|
| 432 |
-
encode_kwargs = self._validate_encoding_params(
|
| 433 |
-
return_multivector=return_multivector, truncate_dim=truncate_dim, prompt_name=prompt_name
|
| 434 |
-
)
|
| 435 |
|
| 436 |
task = self._validate_task(task)
|
| 437 |
|
|
@@ -449,6 +443,7 @@ class JinaEmbeddingsV4Model(Qwen2_5_VLForConditionalGeneration):
|
|
| 449 |
processor_fn=processor_fn,
|
| 450 |
desc="Encoding texts...",
|
| 451 |
task_label=task,
|
|
|
|
| 452 |
return_numpy=return_numpy,
|
| 453 |
batch_size=batch_size,
|
| 454 |
**encode_kwargs,
|
|
@@ -499,7 +494,7 @@ class JinaEmbeddingsV4Model(Qwen2_5_VLForConditionalGeneration):
|
|
| 499 |
self.processor.image_processor.max_pixels = (
|
| 500 |
max_pixels # change during encoding
|
| 501 |
)
|
| 502 |
-
encode_kwargs = self._validate_encoding_params(
|
| 503 |
task = self._validate_task(task)
|
| 504 |
|
| 505 |
# Convert single image to list
|
|
@@ -513,6 +508,7 @@ class JinaEmbeddingsV4Model(Qwen2_5_VLForConditionalGeneration):
|
|
| 513 |
desc="Encoding images...",
|
| 514 |
task_label=task,
|
| 515 |
batch_size=batch_size,
|
|
|
|
| 516 |
return_numpy=return_numpy,
|
| 517 |
**encode_kwargs,
|
| 518 |
)
|
|
|
|
| 357 |
|
| 358 |
def _validate_encoding_params(
|
| 359 |
self,
|
|
|
|
| 360 |
truncate_dim: Optional[int] = None,
|
| 361 |
prompt_name: Optional[str] = None,
|
| 362 |
) -> Dict[str, Any]:
|
|
|
|
| 373 |
else PREFIX_DICT["query"]
|
| 374 |
)
|
| 375 |
|
|
|
|
|
|
|
|
|
|
| 376 |
truncate_dim = truncate_dim or self.config.truncate_dim
|
| 377 |
if truncate_dim is not None and truncate_dim not in self.config.matryoshka_dims:
|
| 378 |
raise ValueError(
|
|
|
|
| 425 |
List of text embeddings as tensors or numpy arrays when encoding multiple texts, or single text embedding as tensor when encoding a single text
|
| 426 |
"""
|
| 427 |
prompt_name = prompt_name or "query"
|
| 428 |
+
encode_kwargs = self._validate_encoding_params(truncate_dim=truncate_dim, prompt_name=prompt_name)
|
|
|
|
|
|
|
| 429 |
|
| 430 |
task = self._validate_task(task)
|
| 431 |
|
|
|
|
| 443 |
processor_fn=processor_fn,
|
| 444 |
desc="Encoding texts...",
|
| 445 |
task_label=task,
|
| 446 |
+
return_multivector=return_multivector,
|
| 447 |
return_numpy=return_numpy,
|
| 448 |
batch_size=batch_size,
|
| 449 |
**encode_kwargs,
|
|
|
|
| 494 |
self.processor.image_processor.max_pixels = (
|
| 495 |
max_pixels # change during encoding
|
| 496 |
)
|
| 497 |
+
encode_kwargs = self._validate_encoding_params(truncate_dim=truncate_dim)
|
| 498 |
task = self._validate_task(task)
|
| 499 |
|
| 500 |
# Convert single image to list
|
|
|
|
| 508 |
desc="Encoding images...",
|
| 509 |
task_label=task,
|
| 510 |
batch_size=batch_size,
|
| 511 |
+
return_multivector=return_multivector,
|
| 512 |
return_numpy=return_numpy,
|
| 513 |
**encode_kwargs,
|
| 514 |
)
|