optimum-rbln 0.9.2a5__py3-none-any.whl → 0.9.2a7__py3-none-any.whl

optimum/rbln/__version__.py

@@ -28,7 +28,7 @@ version_tuple: VERSION_TUPLE
 commit_id: COMMIT_ID
 __commit_id__: COMMIT_ID
 
-__version__ = version = '0.9.2a5'
-__version_tuple__ = version_tuple = (0, 9, 2, 'a5')
+__version__ = version = '0.9.2a7'
+__version_tuple__ = version_tuple = (0, 9, 2, 'a7')
 
 __commit_id__ = commit_id = None

optimum/rbln/modeling.py

@@ -249,37 +249,11 @@ class RBLNModel(RBLNBaseModel):
         trust_remote_code: bool = False,
         # Some rbln-config should be applied before loading torch module (i.e. quantized llm)
         rbln_config: Optional[RBLNModelConfig] = None,
-        dtype: Optional[Union[str, torch.dtype, dict]] = None,
         **kwargs,
     ) -> "PreTrainedModel":
         kwargs = cls.update_kwargs(kwargs)
 
-        hf_class = cls.get_hf_class()
-
-        if dtype is not None:
-            config = hf_class.config_class.from_pretrained(
-                model_id,
-                subfolder=subfolder,
-                revision=revision,
-                cache_dir=cache_dir,
-                use_auth_token=use_auth_token,
-                local_files_only=local_files_only,
-                force_download=force_download,
-                trust_remote_code=trust_remote_code,
-            )
-
-            config, processed_dtype, dtype_orig = _get_dtype(
-                cls=hf_class,
-                dtype=dtype,
-                config=config,
-            )
-
-            kwargs["torch_dtype"] = processed_dtype
-
-            if dtype_orig is not None:
-                hf_class._set_default_dtype(dtype_orig)
-
-        return hf_class.from_pretrained(
+        return cls.get_hf_class().from_pretrained(
             model_id,
             subfolder=subfolder,
             revision=revision,

optimum/rbln/transformers/models/decoderonly/modeling_decoderonly.py

@@ -322,22 +322,30 @@ class RBLNDecoderOnlyModel(RBLNModel, RBLNDecoderOnlyFlashAttentionMixin):
         *args,
         rbln_config: Optional[RBLNDecoderOnlyModelConfig] = None,
         num_hidden_layers: Optional[int] = None,
+        trust_remote_code: Optional[bool] = None,
+        torch_dtype: Optional[torch.dtype] = None,
+        dtype: Optional[torch.dtype] = None,
         **kwargs,
     ) -> PreTrainedModel:
         if rbln_config and rbln_config.quantization:
             model = cls.get_quantized_model(model_id, *args, rbln_config=rbln_config, **kwargs)
         else:
+            # TODO : resolve how to control PreTrainedConfig for hf_kwargs
             if num_hidden_layers is not None:
-                trust_remote_code = kwargs.get("trust_remote_code", None)
                 config, kwargs = AutoConfig.from_pretrained(
-                    model_id, return_unused_kwargs=True, num_hidden_layers=num_hidden_layers, **kwargs
+                    model_id,
+                    return_unused_kwargs=True,
+                    trust_remote_code=trust_remote_code,
+                    num_hidden_layers=num_hidden_layers,
+                    **kwargs,
                 )
                 if hasattr(config, "layer_types"):
                     config.layer_types = config.layer_types[:num_hidden_layers]
                 kwargs["config"] = config
-                kwargs["trust_remote_code"] = trust_remote_code
 
-            model = super().get_pytorch_model(model_id, *args, **kwargs)
+            model = super().get_pytorch_model(
+                model_id, *args, trust_remote_code=trust_remote_code, torch_dtype=torch_dtype, dtype=dtype, **kwargs
+            )
 
         return model
 

{optimum_rbln-0.9.2a5.dist-info → optimum_rbln-0.9.2a7.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: optimum-rbln
-Version: 0.9.2a5
+Version: 0.9.2a7
 Summary: Optimum RBLN is the interface between the HuggingFace Transformers and Diffusers libraries and RBLN accelerators. It provides a set of tools enabling easy model loading and inference on single and multiple rbln device settings for different downstream tasks.
 Project-URL: Homepage, https://rebellions.ai
 Project-URL: Documentation, https://docs.rbln.ai

{optimum_rbln-0.9.2a5.dist-info → optimum_rbln-0.9.2a7.dist-info}/RECORD

@@ -1,7 +1,7 @@
 optimum/rbln/__init__.py,sha256=AZ-7X3ZCjMNcz4mkC_98y-HWRw38ijh8gETJet9tiyM,18828
-optimum/rbln/__version__.py,sha256=SknjPrWnnK8RN1n8X1WB74WmhEFBtAnbp5Zb4GrhDlA,712
+optimum/rbln/__version__.py,sha256=tBr7Pz7fkShuqUHVnIf1VZ9ey-Bfa4EMQ2i12M-O1gY,712
 optimum/rbln/configuration_utils.py,sha256=KFibQ8IYcapw3M1GpgNLgQ1ZolRIip0_bOlwfU0OYac,38193
-optimum/rbln/modeling.py,sha256=jfYPOl0FmYq6B3et3rmoCbiWV4939LlVgs-PrZ7IWIk,16928
+optimum/rbln/modeling.py,sha256=h-Iiku3l9KWF1fBpg3loG74VWU13_n7VjVdry5OC06A,16082
 optimum/rbln/modeling_base.py,sha256=poXfHZCAlFd28MY9dvMi7tC2RytLx77Lee2XGS_KeZg,27684
 optimum/rbln/diffusers/__init__.py,sha256=1tgU_xWA42BmInqu9bBz_5R_E9TGhhK3mI06YlaiTLg,7232
 optimum/rbln/diffusers/modeling_diffusers.py,sha256=egx137ECmNA0bK2JPRUOpxCl1Wb3qOpE-xSaPaJOs-g,20549
@@ -110,7 +110,7 @@ optimum/rbln/transformers/models/decoderonly/decoderonly_architecture.py,sha256=
 optimum/rbln/transformers/models/decoderonly/decoderonly_runtime_utils.py,sha256=2yGU1JfxCepG37DbyY8oWq3gMIKhMnPyBkLH8VniKfA,21530
 optimum/rbln/transformers/models/decoderonly/generation_decoderonly.py,sha256=zabSgQd2VzHhkpbhUFW5Z-CjYB1JvSJOb5yXKjXCQV0,4326
 optimum/rbln/transformers/models/decoderonly/lora_architecture.py,sha256=jo-jYy95JhdvOsX1UTCXeYTNer37wBbtY578C0QQpZo,8306
-optimum/rbln/transformers/models/decoderonly/modeling_decoderonly.py,sha256=myXdyYLivoPVJVGa2vmeaGLFxmjw-vUyYYsNTAX41yk,35907
+optimum/rbln/transformers/models/decoderonly/modeling_decoderonly.py,sha256=lAPVrH-EoW48g3TGHgLs0Wvx9UGpyQh3eOO90qvQAlI,36213
 optimum/rbln/transformers/models/depth_anything/__init__.py,sha256=xvPSIriMJWyNeVYoVB1Z7YqB4kkHOIkaHq7loNps-dk,756
 optimum/rbln/transformers/models/depth_anything/configuration_depth_anything.py,sha256=JujBVEUa_zZDXNPr1y-B_PhK5SgFFcY8Ib4EoGjjtmE,989
 optimum/rbln/transformers/models/depth_anything/modeling_depth_anything.py,sha256=tTmsVaW9Wb2WD3nKRLwp7swn3hbMvgwUEJwwVIfNYEc,1008
@@ -245,7 +245,7 @@ optimum/rbln/utils/model_utils.py,sha256=4k5879Kh75m3x_vS4-qOGfqsOiAvc2kdNFFfvsF
 optimum/rbln/utils/runtime_utils.py,sha256=R6uXDbeJP03-FWdd4vthNe2D4aCra5n12E3WB1ifiGM,7933
 optimum/rbln/utils/save_utils.py,sha256=hG5uOtYmecSXZuGTvCXsTM-SiyZpr5q3InUGCCq_jzQ,3619
 optimum/rbln/utils/submodule.py,sha256=SKLnM3KsX8_rv3HauO4oB2-JSjzuadQjRwo_BhMUzLI,6362
-optimum_rbln-0.9.2a5.dist-info/METADATA,sha256=8gm204ZrCL7yjPNG5vEGsSE5RhIDtPk2H-eWGFxX8yw,5350
-optimum_rbln-0.9.2a5.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
-optimum_rbln-0.9.2a5.dist-info/licenses/LICENSE,sha256=QwcOLU5TJoTeUhuIXzhdCEEDDvorGiC6-3YTOl4TecE,11356
-optimum_rbln-0.9.2a5.dist-info/RECORD,,
+optimum_rbln-0.9.2a7.dist-info/METADATA,sha256=JjNz9cm2rrGv-28JFHeo8X1MbyTE3YCTURDRoQlA79g,5350
+optimum_rbln-0.9.2a7.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+optimum_rbln-0.9.2a7.dist-info/licenses/LICENSE,sha256=QwcOLU5TJoTeUhuIXzhdCEEDDvorGiC6-3YTOl4TecE,11356
+optimum_rbln-0.9.2a7.dist-info/RECORD,,