optimum-rbln 0.7.5a1__py3-none-any.whl → 0.7.5rc0__py3-none-any.whl

optimum/rbln/transformers/models/gpt2/gpt2_architecture.py

@@ -38,7 +38,10 @@ class GPT2Wrapper(DecoderOnlyWrapper):
         new_layers = []
         for layer in causal_lm.transformer.h:
             new_self_attn = GPT2Attention(
-                layer.attn, self.use_attention_mask, kvcache_block_size=self.kvcache_block_size
+                layer.attn,
+                self.use_attention_mask,
+                kvcache_block_size=self.kvcache_block_size,
+                use_position_ids=self.use_position_ids,
             )
             new_layer = GPT2Layer(layer, new_self_attn)
             new_layers.append(new_layer)

optimum/rbln/transformers/models/midm/midm_architecture.py

@@ -61,7 +61,10 @@ class MidmLMHeadModelWrapper(DecoderOnlyWrapper):
         new_layers = []
         for layer in causal_lm.transformer.h:
             new_self_attn = MidmAttention(
-                layer.attn, self.use_attention_mask, kvcache_block_size=self.kvcache_block_size
+                layer.attn,
+                self.use_attention_mask,
+                kvcache_block_size=self.kvcache_block_size,
+                use_position_ids=self.use_position_ids,
             )
             new_layer = MidmLayer(layer, new_self_attn)
             new_layers.append(new_layer)

optimum/rbln/transformers/models/opt/modeling_opt.py

@@ -70,6 +70,8 @@ class RBLNOPTForCausalLM(RBLNDecoderOnlyModelForCausalLM):
             "kvcache_block_size": rbln_config.kvcache_block_size,
             "use_rotary_emb": cls._use_rotary_emb,
             "use_attention_mask": rbln_config.use_attention_mask,
+            "use_position_ids": rbln_config.use_position_ids,
+            "use_inputs_embeds": rbln_config.use_inputs_embeds,
         }
 
         for i in range(len(model.model.decoder.layers)):

optimum/rbln/transformers/models/opt/opt_architecture.py

@@ -38,7 +38,10 @@ class OPTWrapper(DecoderOnlyWrapper):
 
         for layer in causal_lm.model.decoder.layers:
             new_self_attn = OPTAttention(
-                layer.self_attn, self.use_attention_mask, kvcache_block_size=self.kvcache_block_size
+                layer.self_attn,
+                self.use_attention_mask,
+                kvcache_block_size=self.kvcache_block_size,
+                use_position_ids=self.use_position_ids,
             )
             new_layer = OPTDecoderLayer(layer, new_self_attn)
             new_layers.append(new_layer)

optimum/rbln/transformers/models/phi/phi_architecture.py

@@ -37,7 +37,10 @@ class PhiWrapper(DecoderOnlyWrapper):
         for layer in causal_lm.model.layers:
             if self.attn_impl == "eager":
                 new_self_attn = PhiAttention(
-                    layer.self_attn, self.use_attention_mask, kvcache_block_size=self.kvcache_block_size
+                    layer.self_attn,
+                    self.use_attention_mask,
+                    kvcache_block_size=self.kvcache_block_size,
+                    use_position_ids=self.use_position_ids,
                 )
             elif self.attn_impl == "flash_attn":
                 raise NotImplementedError(f"flash attn for {self.__class__} is not implemented yet.")

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

@@ -394,7 +394,7 @@ class RBLNQwen2_5_VLForConditionalGeneration(RBLNDecoderOnlyModelForCausalLM):
             hidden_size,
             head_dim,
         )
-        pos_idx = 4 if query_length > 1 else 5
+        pos_idx = 3
         input_info.insert(pos_idx, ("position_emb", [2, batch_size, 1, query_length, head_dim], "float32"))
 
         return input_info
@@ -595,8 +595,9 @@ class RBLNQwen2_5_VLForConditionalGeneration(RBLNDecoderOnlyModelForCausalLM):
                 )
                 logits.append(output.logits)
             logits = torch.cat(logits, dim=0)
-        # Decoder
+            # Decoder
         else:
+            print(input_ids[0], cache_position[0])
             inputs_embeds, position_embed = self._preprocess_decoder(input_ids, cache_position)
             output = self.decoder(
                 inputs_embeds=inputs_embeds,

{optimum_rbln-0.7.5a1.dist-info → optimum_rbln-0.7.5rc0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: optimum-rbln
-Version: 0.7.5a1
+Version: 0.7.5rc0
 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-0.7.5a1.dist-info → optimum_rbln-0.7.5rc0.dist-info}/RECORD

@@ -1,5 +1,5 @@
-optimum/rbln/__init__.py,sha256=MXUvNNwbZhrPpxjpDSIV5fCQIT_rcNaLPstQgUYj4gE,13970
-optimum/rbln/__version__.py,sha256=mPJDzmeS5WyCHJnuvAR_w6Lpxp2T9tocDUBPoe9Zwoo,519
+optimum/rbln/__init__.py,sha256=oAnsJSMrPYwBGEttUt3CMXTIESVNe15ftTWRTShwhZI,14386
+optimum/rbln/__version__.py,sha256=34rdkaF19JfWW5k_S6Q9ZJaeOOAsCoPR3_vN57l-b28,521
 optimum/rbln/configuration_utils.py,sha256=gvAjRFEGw5rnSoH0IoyuLrE4fkxtk3DN1pikqrN_Rpk,31277
 optimum/rbln/modeling.py,sha256=4Xwi3ovWDHOOqxUDH_ZgsgTuea8Kyg25D9s81zVYpr0,9669
 optimum/rbln/modeling_base.py,sha256=iQKw2IORu1cN6sOK0xeBVrhatt-ZPeinT_v6l2FnGRw,24173
@@ -61,16 +61,16 @@ optimum/rbln/ops/flash_attn.py,sha256=z39DJZSk94630ueoOCkiybxR5gzvNR-SRADHs0F6pz
 optimum/rbln/ops/kv_cache_update.py,sha256=HjnHBR-oFrJQibsVnkYb0P5_-wEma8jl0mkjkylwakU,1270
 optimum/rbln/ops/linear.py,sha256=1_7Hg-9wXxhu97fqPobotLQx17k7VPeSSL91_9Z7EDg,1018
 optimum/rbln/ops/sliding_window_attn.py,sha256=EQrV_yRGc5z6kvwEsAcLP028bJWkQg2UPI3xubt9skU,3487
-optimum/rbln/transformers/__init__.py,sha256=M9hEzEWIwLdkS-T0-sMof8BtYeFH-NAj4iX73mB3Rj8,8767
+optimum/rbln/transformers/__init__.py,sha256=LW6abfb0W0jHziE8dIEDBeyb4Cj-aq8dUldOYKNR6_E,9183
 optimum/rbln/transformers/configuration_alias.py,sha256=qFVfg6ohsR7a6b-CBgxjBUPDrk9MyiJwtO8AQah_RTU,1505
 optimum/rbln/transformers/configuration_generic.py,sha256=XIiZ1-5p1CMHhG7Sr2qR4SLYKcYw9aph7eGlga3Opx0,5056
 optimum/rbln/transformers/modeling_alias.py,sha256=yx7FnZQWAnrWzivaO5hI7T6i-fyLzt2tMIXG2oDNbPo,1657
 optimum/rbln/transformers/modeling_generic.py,sha256=nT_lytAILkYtwBVJKxXg0dxmh0UpjGYO6zOdLoMs1uU,12891
 optimum/rbln/transformers/modeling_rope_utils.py,sha256=3zwkhYUyTZhxCJUSmwCc88iiY1TppRWEY9ShwUqNB2k,14293
-optimum/rbln/transformers/models/__init__.py,sha256=rofipeYNupozWru5PfSV9BudskognIUuNRIo7z-TEKE,8039
-optimum/rbln/transformers/models/auto/__init__.py,sha256=GvGbb3ZpMv-h6euXeZ42jSizoOfrL2O1uvpAnfKxYEo,1034
+optimum/rbln/transformers/models/__init__.py,sha256=qNh_d7bBKxhxBbUImXJ66n0Vo0NW1m7tMIU5M2ZxGmw,8510
+optimum/rbln/transformers/models/auto/__init__.py,sha256=34Xghf1ogG4u-jhBMlj134nHdgnR3JEHSeZTPuy3MpY,1071
 optimum/rbln/transformers/models/auto/auto_factory.py,sha256=Uf5rCUoxec2qhIAwbAeZNZN4NIMFaLurSB1EdI79lwA,7044
-optimum/rbln/transformers/models/auto/modeling_auto.py,sha256=Un9qoqdy3dO8JBza_bTJF_6_fRVNM9QisihSgTRFI-o,3933
+optimum/rbln/transformers/models/auto/modeling_auto.py,sha256=Ex2ARRRIt3LtKhazr0UWy67R6WFAwSKRoi5n6B8TvoI,4213
 optimum/rbln/transformers/models/bart/__init__.py,sha256=fVo-gZEmJ0yxkIxEX6ciuRAGgXNyuvaXE2s88bhbjAE,830
 optimum/rbln/transformers/models/bart/bart_architecture.py,sha256=Oo-Cdne7igKEex8wwP-gztKJHgs5GLHQjK1oc3IZIDE,5801
 optimum/rbln/transformers/models/bart/configuration_bart.py,sha256=iIuotDRU0IZteTNUY97jnsecIEYokaJ0BF5Sv_MJk00,909
@@ -86,22 +86,26 @@ optimum/rbln/transformers/models/clip/configuration_clip.py,sha256=wgfZeVvcVdSzr
 optimum/rbln/transformers/models/clip/modeling_clip.py,sha256=UslcDN6otyQ_psou7F_YcdK5vCImEtgIdcbwmexSfOM,7256
 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=89I4A0WEdY0bOX_iGeG5qS7Cd2RPbwvdgYjfZIhcUy4,42035
-optimum/rbln/transformers/models/decoderonly/modeling_decoderonly.py,sha256=iaB1lgElTaCfq5-HY5mbOcbQ8CICyTONckJf25RPjpU,48565
+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/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
 optimum/rbln/transformers/models/exaone/__init__.py,sha256=eUL0mq3yGVzCQfjLlOtVF2MecIN3DQWm07EmXubGSTs,921
 optimum/rbln/transformers/models/exaone/configuration_exaone.py,sha256=97r1wkO9cgwErePk2UX6n7WeVkOmWUeGgpK6qAGOUi4,774
-optimum/rbln/transformers/models/exaone/exaone_architecture.py,sha256=ZM5vvz8KBipOiMVi8vqfvejkDSknW69xh4GrvJix-g0,3350
+optimum/rbln/transformers/models/exaone/exaone_architecture.py,sha256=UTpMOLFOt3r9eziNRd8qWU6ekVmM3CxTxg5x8mA12MI,3511
 optimum/rbln/transformers/models/exaone/modeling_exaone.py,sha256=WjyH8PmsMljSea7kJn_Cq1FJ96OXwXAoU7hv2Q8zUnI,1747
 optimum/rbln/transformers/models/gemma/__init__.py,sha256=VqPIlokw3kjn_ZoLXINCLXw3vaysQFo5oPGGy6bnt4Q,708
 optimum/rbln/transformers/models/gemma/configuration_gemma.py,sha256=ek7PiMWx4LDAJ0mCedJ55o_TKTiOCSGD5V-1o1Fy4f8,773
-optimum/rbln/transformers/models/gemma/gemma_architecture.py,sha256=bmCx405FVcffhgrQ53qMMZDbSlPxWOjucMHbvq19Gnw,2286
+optimum/rbln/transformers/models/gemma/gemma_architecture.py,sha256=_Ip8J4IKEkXwEiynKvoj7NEYDR0eJ4IEWXqgjXibBy4,2447
 optimum/rbln/transformers/models/gemma/modeling_gemma.py,sha256=-U3w3cEOv3ps1S8aL7uOq6Kq2siCPZz7Z8MXhDQgQqo,1530
+optimum/rbln/transformers/models/gemma3/__init__.py,sha256=6rugk3615SEt4lh7gduo_J9VyGiSReuEIvL0Uno0eaI,790
+optimum/rbln/transformers/models/gemma3/configuration_gemma3.py,sha256=nndcYVwDYkOige_qO4td-YwLNtUz6aLiSQDIfPdGG9A,2840
+optimum/rbln/transformers/models/gemma3/gemma3_architecture.py,sha256=Uer27wG06hgV1WNf92x1ZeUpl4Q0zskfCqzlLhtgtNU,17348
+optimum/rbln/transformers/models/gemma3/modeling_gemma3.py,sha256=OUGsML-H6FOKldld7KRW9l0nRoT4DojWBDl8ZHpfXVA,44982
 optimum/rbln/transformers/models/gpt2/__init__.py,sha256=socBMIBZSiLbrVN12rQ4nL9gFeT0axMgz6SWaCaD4Ac,704
 optimum/rbln/transformers/models/gpt2/configuration_gpt2.py,sha256=vKvJD8P9Li4W9wdVoQcqMEr1MwEXojPBnF2NE85VXAo,772
-optimum/rbln/transformers/models/gpt2/gpt2_architecture.py,sha256=1IxqHmB-GlH2Dv2Yk4z0rMxL9CpxMGHhSu_x8_4cxvs,3008
+optimum/rbln/transformers/models/gpt2/gpt2_architecture.py,sha256=kf5jk7Djv9XXX3Q83oTosiMpt9g44TF_gCT-vMiWDJk,3097
 optimum/rbln/transformers/models/gpt2/modeling_gpt2.py,sha256=qBDanUk_O-HtOIVCA4IE3FYyCsnL9xIDK00vft-0caw,1490
 optimum/rbln/transformers/models/idefics3/__init__.py,sha256=ulxE7HEfXsNJhd25J9Fvi6vggo9aZH9sLKJjWB6LlzQ,814
 optimum/rbln/transformers/models/idefics3/configuration_idefics3.py,sha256=sM0pXsvkxcpDXagoKlqwKdBAcNdayB9KlWdYC9xlyDU,1889
@@ -115,7 +119,7 @@ optimum/rbln/transformers/models/llava_next/configuration_llava_next.py,sha256=Q
 optimum/rbln/transformers/models/llava_next/modeling_llava_next.py,sha256=Igq_VCcOgSxltzm7eV9F5MiHS-vyDAp4DIswH80q76c,26022
 optimum/rbln/transformers/models/midm/__init__.py,sha256=IC3FETwgYinbp3wDj7tp4zIHJhbqM-c6GfTRdYcMNj8,913
 optimum/rbln/transformers/models/midm/configuration_midm.py,sha256=Kv5g5dIsBrhGcZ2_pFUOPNB80np4Xiw0wPH1IZm1PHI,772
-optimum/rbln/transformers/models/midm/midm_architecture.py,sha256=357iviqQkzI0s_lU_teH1sVOChNRDUABe3GA0HuhZZY,5444
+optimum/rbln/transformers/models/midm/midm_architecture.py,sha256=I2oEE6IjV0Khx5X9t2NpOzEF59zacw6XJgEFxKL26cA,5533
 optimum/rbln/transformers/models/midm/modeling_midm.py,sha256=GG25BozEZriAL-OPFGpzOjyDtSFB-NfeiLJTDAqxe20,1734
 optimum/rbln/transformers/models/mistral/__init__.py,sha256=9FE64bCYfSIyrBkRcwlqF8QyacSJFWvwEufHFi1ZIrM,716
 optimum/rbln/transformers/models/mistral/configuration_mistral.py,sha256=zCHZnmjTXs7UKYIIRR2IYjccROm5ODXfonWploITNxg,775
@@ -123,19 +127,19 @@ optimum/rbln/transformers/models/mistral/mistral_architecture.py,sha256=_aU8TE_t
 optimum/rbln/transformers/models/mistral/modeling_mistral.py,sha256=7nrddoBIHf8S12LZWBUpotnvG3gND11vMQda9yYXJ-s,1560
 optimum/rbln/transformers/models/opt/__init__.py,sha256=w0v8GzbzlR5_4yL851njGDSJgX89TrYxrHnpNfMHZEI,700
 optimum/rbln/transformers/models/opt/configuration_opt.py,sha256=1BIxum4m6An5dsYpw1Kmf-QIHSGQK8uDjMd2hUkRvzE,771
-optimum/rbln/transformers/models/opt/modeling_opt.py,sha256=A40krCtNLOWDex4tHx894wBTcWhyrsQBtGgZZDDc3fg,3005
-optimum/rbln/transformers/models/opt/opt_architecture.py,sha256=7ANZ1aIK0HZ8HPrAl22V1Hc0B-8gBigsa4VNMh9HVk8,2592
+optimum/rbln/transformers/models/opt/modeling_opt.py,sha256=NteN2EIW0le-QT_gWyGUb7yC-eAqHrI1x1UflCldu6Q,3131
+optimum/rbln/transformers/models/opt/opt_architecture.py,sha256=zrgOrZ2C_mdHcX-beOu6TB52rjNhe3PLqS5BpC-_UpU,2681
 optimum/rbln/transformers/models/phi/__init__.py,sha256=uqQb-sO1HXuaju2hfo7qJHk_IWhnptY-qFjNjK_uOc0,700
 optimum/rbln/transformers/models/phi/configuration_phi.py,sha256=9Mk06ZNymGnC2P0eiU02oguH1KFKxL-LYtMJSNqpmmo,771
 optimum/rbln/transformers/models/phi/modeling_phi.py,sha256=j-6Pqd5rR2JE8I1pnKFlCi4nW5Dv3wZjoPWxohissoo,1516
-optimum/rbln/transformers/models/phi/phi_architecture.py,sha256=TueyqmjPXWmOPOxBm4dIFyd0X3iV1jgw0U6c26iCAPk,4090
+optimum/rbln/transformers/models/phi/phi_architecture.py,sha256=MioLLM4uxZvJqaWQn25WvDo85pI_rXKTJqlwaptqZ80,4191
 optimum/rbln/transformers/models/qwen2/__init__.py,sha256=Tu4_AXy3ktTvxGwxED3kewiv62S75HgDWD6-TeC1DfA,708
 optimum/rbln/transformers/models/qwen2/configuration_qwen2.py,sha256=sQBu4UjM8Ctiy696FLjKeck1t49MR31zWTCN_bMvBl4,773
 optimum/rbln/transformers/models/qwen2/modeling_qwen2.py,sha256=9-aFDvjMzPNUyGOz0qo33RE18bUFGYZ3Wt_68zb5uJY,1530
 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=etE0y5UlNpNcRZO_68TAYK1Wg2TsqZUiM0L2qeTZacs,25144
+optimum/rbln/transformers/models/qwen2_5_vl/modeling_qwen2_5_vl.py,sha256=DUlTQc24IJwWMTIi3ou3F7eraYJy9WERjH7yAZo80u4,25172
 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
@@ -174,7 +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-0.7.5a1.dist-info/METADATA,sha256=r1huaKWrPkBuoc7Zm743oYmueJxP7g5hlz9Ywtxf-r8,5299
-optimum_rbln-0.7.5a1.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
-optimum_rbln-0.7.5a1.dist-info/licenses/LICENSE,sha256=QwcOLU5TJoTeUhuIXzhdCEEDDvorGiC6-3YTOl4TecE,11356
-optimum_rbln-0.7.5a1.dist-info/RECORD,,
+optimum_rbln-0.7.5rc0.dist-info/METADATA,sha256=aXeccsNinGR5xXxBOKIMhxfeHyF-wQE5DxWtOrt2WyI,5300
+optimum_rbln-0.7.5rc0.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+optimum_rbln-0.7.5rc0.dist-info/licenses/LICENSE,sha256=QwcOLU5TJoTeUhuIXzhdCEEDDvorGiC6-3YTOl4TecE,11356
+optimum_rbln-0.7.5rc0.dist-info/RECORD,,