optimum-rbln 0.7.5rc3__py3-none-any.whl → 0.8.1a0__py3-none-any.whl

optimum/rbln/__version__.py

@@ -17,5 +17,5 @@ __version__: str
 __version_tuple__: VERSION_TUPLE
 version_tuple: VERSION_TUPLE
 
-__version__ = version = '0.7.5rc3'
-__version_tuple__ = version_tuple = (0, 7, 5, 'rc3')
+__version__ = version = '0.8.1a0'
+__version_tuple__ = version_tuple = (0, 8, 1, 'a0')

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

@@ -709,7 +709,9 @@ class RBLNDecoderOnlyModelForCausalLM(RBLNModel):
                 compiled_models[f"decoder_batch_{batch_size}"].get_alloc_per_node_by_key().items()
             ):
                 alloc_memory_by_key[key] += sum(memory_per_node)
-        alloc_memory_by_key.pop("PortRecur")  # kv-cache
+
+        alloc_memory_by_key.pop("PortRecur", None)  # Old compiler's kv-cache Key
+        alloc_memory_by_key.pop("DramTensor", None)  # kv-cache
         kernel_size = alloc_memory_by_key.pop("Kernel")  # model weight
 
         # Get the maximum number of blocks that can be allocated

optimum/rbln/transformers/models/qwen2_5_vl/modeling_qwen2_5_vl.py

@@ -595,7 +595,7 @@ class RBLNQwen2_5_VLForConditionalGeneration(RBLNDecoderOnlyModelForCausalLM):
                 )
                 logits.append(output.logits)
             logits = torch.cat(logits, dim=0)
-            # Decoder
+        # Decoder
         else:
             inputs_embeds, position_embed = self._preprocess_decoder(input_ids, cache_position)
             output = self.decoder(

{optimum_rbln-0.7.5rc3.dist-info → optimum_rbln-0.8.1a0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: optimum-rbln
-Version: 0.7.5rc3
+Version: 0.8.1a0
 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-0.7.5rc3.dist-info → optimum_rbln-0.8.1a0.dist-info}/RECORD

@@ -1,5 +1,5 @@
 optimum/rbln/__init__.py,sha256=oAnsJSMrPYwBGEttUt3CMXTIESVNe15ftTWRTShwhZI,14386
-optimum/rbln/__version__.py,sha256=IAHCNgmlAYNwyLUwYX8yN274g--7rbrAmdvwkfDYBjI,521
+optimum/rbln/__version__.py,sha256=8KrQ-gTLuRdO8nDDly9pCF_MZDqgHg6xW9QXyIjIHug,519
 optimum/rbln/configuration_utils.py,sha256=gvAjRFEGw5rnSoH0IoyuLrE4fkxtk3DN1pikqrN_Rpk,31277
 optimum/rbln/modeling.py,sha256=BpydF-bLBF60NnRMbtZwn5odOUjU4Awu9azqGeSufTI,11462
 optimum/rbln/modeling_base.py,sha256=HQgscr5jpUEtuXU1ACJHSLIntX-kq6Ef0SQ_W2-rp5A,25341
@@ -87,7 +87,7 @@ optimum/rbln/transformers/models/clip/modeling_clip.py,sha256=UslcDN6otyQ_psou7F
 optimum/rbln/transformers/models/decoderonly/__init__.py,sha256=vQYZDDdoddwA7yKc5zzrq2Zs9sax-0p8rNF_aYfF4bk,1006
 optimum/rbln/transformers/models/decoderonly/configuration_decoderonly.py,sha256=WQlHUjcMuD_3BPLo1Ytzz-xAI-8J9BYfr7dEzmpLVks,6389
 optimum/rbln/transformers/models/decoderonly/decoderonly_architecture.py,sha256=m_wTGXSK8cBG3NjLDNLTxSWyErl7MB4PPU9fQ6zijEI,45991
-optimum/rbln/transformers/models/decoderonly/modeling_decoderonly.py,sha256=bYhw5co6aeueDNR2VHc9vwzhn9JL_fHTJS7E2bE6500,49471
+optimum/rbln/transformers/models/decoderonly/modeling_decoderonly.py,sha256=hta8L3LB4nOMstSOFrPrbr2fIqzU5FMLev7UPB7eSJk,49561
 optimum/rbln/transformers/models/dpt/__init__.py,sha256=Nzep9mlzKyL1kV726IBqY8DnLp1DkH9JzFeknWSRhok,714
 optimum/rbln/transformers/models/dpt/configuration_dpt.py,sha256=4fW6bzVhaAxym4wGV3F785rvUOoWPyw_gdEMqB08Leg,755
 optimum/rbln/transformers/models/dpt/modeling_dpt.py,sha256=oKLX7MQZvfk1QB8wOtcdi7AmZH2fOIVbypa9A3RA9MI,733
@@ -139,7 +139,7 @@ optimum/rbln/transformers/models/qwen2/modeling_qwen2.py,sha256=9-aFDvjMzPNUyGOz
 optimum/rbln/transformers/models/qwen2/qwen2_architecture.py,sha256=XlNAMYAcDLohnSAhIFGKOPuCB5XLgzYs5ABWdeQSaZs,720
 optimum/rbln/transformers/models/qwen2_5_vl/__init__.py,sha256=rAW3DKQUzGL6EMwa5r1iLu94yhpiZpk6zfoD7TtYXrc,865
 optimum/rbln/transformers/models/qwen2_5_vl/configuration_qwen2_5_vl.py,sha256=dPcGNaLwJf61PIvVbyt-lvBflp_dvK0hubhNoA3len0,3123
-optimum/rbln/transformers/models/qwen2_5_vl/modeling_qwen2_5_vl.py,sha256=fD8_EdETHfu0oOqic5WtoETQ0RHD3jMUEBYl1aUQz94,25121
+optimum/rbln/transformers/models/qwen2_5_vl/modeling_qwen2_5_vl.py,sha256=pb4OC5rRG-Y-a8WMBGqrV0owgjeGH83Lqx2fXztnlJg,25117
 optimum/rbln/transformers/models/qwen2_5_vl/qwen2_5_vl_architecture.py,sha256=OR-tTu8uzeFryenohoqRwgIVqw5zM7dcsZHwKL_jD0A,7232
 optimum/rbln/transformers/models/seq2seq/__init__.py,sha256=6WKstWiS1kW0oFDn_jyrKMW5QEJAWkmsSRAaadNedDM,715
 optimum/rbln/transformers/models/seq2seq/configuration_seq2seq2.py,sha256=vSNP1eILfL32cbiLOAD58Ocz6lk3hYFnhIRLDVqlSoI,2624
@@ -178,7 +178,7 @@ optimum/rbln/utils/model_utils.py,sha256=V2kFpUe2aqVzLwbpztD8JOVFQqRHncvIWwJbgnU
 optimum/rbln/utils/runtime_utils.py,sha256=LoKNK3AQNV_BSScstIZWjICkJf265MnUgy360BOocVI,5454
 optimum/rbln/utils/save_utils.py,sha256=hG5uOtYmecSXZuGTvCXsTM-SiyZpr5q3InUGCCq_jzQ,3619
 optimum/rbln/utils/submodule.py,sha256=ZfI7e3YzbjbbBW4Yjfucj8NygEsukfIkaJi3PtwHrhc,5105
-optimum_rbln-0.7.5rc3.dist-info/METADATA,sha256=QkWbttk1MCudGA7BsBLrgVztVRt9nKIRPT8EQP5etGI,5300
-optimum_rbln-0.7.5rc3.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
-optimum_rbln-0.7.5rc3.dist-info/licenses/LICENSE,sha256=QwcOLU5TJoTeUhuIXzhdCEEDDvorGiC6-3YTOl4TecE,11356
-optimum_rbln-0.7.5rc3.dist-info/RECORD,,
+optimum_rbln-0.8.1a0.dist-info/METADATA,sha256=oggHE1bTirnlPGJnS1TWULpx6etS6lUuS3-dB6vJxXQ,5299
+optimum_rbln-0.8.1a0.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+optimum_rbln-0.8.1a0.dist-info/licenses/LICENSE,sha256=QwcOLU5TJoTeUhuIXzhdCEEDDvorGiC6-3YTOl4TecE,11356
+optimum_rbln-0.8.1a0.dist-info/RECORD,,