optimum-rbln 0.8.1rc1__py3-none-any.whl → 0.8.2a0__py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -12,54 +12,16 @@
12
12
  # See the License for the specific language governing permissions and
13
13
  # limitations under the License.
14
14
 
15
- from typing import TYPE_CHECKING
16
15
 
17
16
  from ...models.decoderonly.decoderonly_architecture import (
18
- DecoderOnlyAttention,
19
- DecoderOnlyFlashAttention,
20
- DecoderOnlyForCausalLM,
21
- DecoderOnlyLayer,
22
17
  DecoderOnlyModel,
23
18
  DecoderOnlyWrapper,
24
19
  )
25
20
 
26
21
 
27
- if TYPE_CHECKING:
28
- from transformers import GemmaForCausalLM
29
-
30
-
31
22
  class GemmaWrapper(DecoderOnlyWrapper):
32
- def convert_to_rbln_causal_lm(self, causal_lm: "GemmaForCausalLM", max_seq_len: int):
33
- new_layers = []
34
- for layer in causal_lm.model.layers:
35
- if self.attn_impl == "eager":
36
- new_self_attn = DecoderOnlyAttention(
37
- layer.self_attn,
38
- self.use_attention_mask,
39
- kvcache_block_size=self.kvcache_block_size,
40
- use_position_ids=self.use_position_ids,
41
- )
42
- elif self.attn_impl == "flash_attn":
43
- new_self_attn = DecoderOnlyFlashAttention(
44
- layer.self_attn,
45
- kvcache_partition_len=self.kvcache_partition_len,
46
- use_attention_mask=self.use_attention_mask,
47
- kvcache_block_size=self.kvcache_block_size,
48
- use_position_ids=self.use_position_ids,
49
- )
50
- else:
51
- raise NotImplementedError(f"Unknwon attn : {self.attn_impl}")
52
- new_layer = DecoderOnlyLayer(layer, new_self_attn)
53
- new_layers.append(new_layer)
54
- new_model = GemmaModel(
55
- causal_lm.model,
56
- new_layers,
57
- partition_len=self.kvcache_partition_len,
58
- max_seq_len=max_seq_len,
59
- sliding_window_layers=self.sliding_window_layers,
60
- )
61
- new_causal_lm = DecoderOnlyForCausalLM(causal_lm, new_model)
62
- return new_causal_lm
23
+ def get_rbln_model_class(self):
24
+ return GemmaModel
63
25
 
64
26
 
65
27
  class GemmaModel(DecoderOnlyModel):
@@ -13,15 +13,13 @@
13
13
  # limitations under the License.
14
14
 
15
15
  import copy
16
- from typing import TYPE_CHECKING, Optional, Tuple, Union
16
+ from typing import Optional, Tuple, Union
17
17
 
18
18
  import torch
19
19
  from transformers.models.gemma3.modeling_gemma3 import Gemma3RMSNorm
20
20
 
21
21
  from ..decoderonly.decoderonly_architecture import (
22
22
  DecoderOnlyAttention,
23
- DecoderOnlyFlashAttention,
24
- DecoderOnlyForCausalLM,
25
23
  DecoderOnlyLayer,
26
24
  DecoderOnlyModel,
27
25
  DecoderOnlyWrapper,
@@ -30,10 +28,6 @@ from ..decoderonly.decoderonly_architecture import (
30
28
  )
31
29
 
32
30
 
33
- if TYPE_CHECKING:
34
- from transformers import Gemma3ForCausalLM
35
-
36
-
37
31
  class Gemma3ForCausalLMWrapper(DecoderOnlyWrapper):
38
32
  def get_rotary_emb(self, max_seq_len):
39
33
  rotary_emb_global = RotaryEmbedding(config=self.config, max_seq_len_cached=max_seq_len)
@@ -45,49 +39,14 @@ class Gemma3ForCausalLMWrapper(DecoderOnlyWrapper):
45
39
 
46
40
  return (rotary_emb_global, rotary_emb_local)
47
41
 
48
- def convert_to_rbln_causal_lm(self, causal_lm: "Gemma3ForCausalLM", max_seq_len: int):
49
- new_layers = []
50
- for layer_idx, layer in enumerate(causal_lm.model.layers):
51
- if layer_idx in self.sliding_window_layers:
52
- new_self_attn = Gemma3Attention(
53
- layer.self_attn,
54
- use_attention_mask=None, # FIXME: no use in SWA
55
- use_position_ids=self.use_position_ids,
56
- kvcache_block_size=self.config.sliding_window,
57
- is_sliding=True,
58
- )
59
- else:
60
- if self.attn_impl == "eager":
61
- new_self_attn = Gemma3Attention(
62
- layer.self_attn,
63
- use_attention_mask=self.use_attention_mask,
64
- use_position_ids=self.use_position_ids,
65
- kvcache_block_size=self.kvcache_block_size,
66
- is_sliding=False,
67
- )
68
- elif self.attn_impl == "flash_attn":
69
- new_self_attn = Gemma3FlashAttention(
70
- layer.self_attn,
71
- kvcache_partition_len=self.kvcache_partition_len,
72
- use_attention_mask=self.use_attention_mask,
73
- kvcache_block_size=self.kvcache_block_size,
74
- use_position_ids=self.use_position_ids,
75
- )
76
- else:
77
- raise NotImplementedError(f"Unknwon attn : {self.attn_impl}")
78
-
79
- new_layer = Gemma3DecoderLayer(layer, new_self_attn)
80
- new_layers.append(new_layer)
81
-
82
- new_model = Gemma3TextModel(
83
- causal_lm.model,
84
- new_layers,
85
- partition_len=self.kvcache_partition_len,
86
- max_seq_len=max_seq_len,
87
- sliding_window_layers=self.sliding_window_layers,
88
- )
89
- new_causal_lm = DecoderOnlyForCausalLM(causal_lm, new_model)
90
- return new_causal_lm
42
+ def get_rbln_attn_class(self):
43
+ return Gemma3Attention
44
+
45
+ def get_rbln_layer_class(self):
46
+ return Gemma3DecoderLayer
47
+
48
+ def get_rbln_model_class(self):
49
+ return Gemma3TextModel
91
50
 
92
51
 
93
52
  class Gemma3TextModel(DecoderOnlyModel):
@@ -199,16 +158,3 @@ class Gemma3Attention(DecoderOnlyAttention):
199
158
 
200
159
  def get_attn_scale(self):
201
160
  return self._original_mod.config.query_pre_attn_scalar**-0.5
202
-
203
-
204
- class Gemma3FlashAttention(DecoderOnlyFlashAttention):
205
- def __post_init__(self):
206
- self.q_proj = self._original_mod.q_proj
207
- self.k_proj = self._original_mod.k_proj
208
- self.v_proj = self._original_mod.v_proj
209
- self.o_proj = self._original_mod.o_proj
210
- self.q_norm = self._original_mod.q_norm
211
- self.k_norm = self._original_mod.k_norm
212
-
213
- def get_attn_scale(self):
214
- return self._original_mod.config.query_pre_attn_scalar**-0.5
@@ -17,10 +17,10 @@ from typing import TYPE_CHECKING, Tuple
17
17
 
18
18
  import torch
19
19
  import torch.nn as nn
20
+ from transformers import PreTrainedModel
20
21
 
21
22
  from ..decoderonly.decoderonly_architecture import (
22
23
  DecoderOnlyAttention,
23
- DecoderOnlyForCausalLM,
24
24
  DecoderOnlyLayer,
25
25
  DecoderOnlyModel,
26
26
  DecoderOnlyWrapper,
@@ -32,27 +32,23 @@ if TYPE_CHECKING:
32
32
 
33
33
 
34
34
  class GPT2Wrapper(DecoderOnlyWrapper):
35
- def convert_to_rbln_causal_lm(self, causal_lm: "GPT2LMHeadModel", max_seq_len: int):
36
- if self.attn_impl != "eager":
37
- raise NotImplementedError(f"flash attention ({self.attn_impl}) is not implemented for {self.__class__}")
38
- new_layers = []
39
- for layer in causal_lm.transformer.h:
40
- new_self_attn = GPT2Attention(
41
- layer.attn,
42
- self.use_attention_mask,
43
- kvcache_block_size=self.kvcache_block_size,
44
- use_position_ids=self.use_position_ids,
45
- )
46
- new_layer = GPT2Layer(layer, new_self_attn)
47
- new_layers.append(new_layer)
48
- new_model = GPT2Model(
49
- causal_lm.transformer,
50
- new_layers,
51
- max_seq_len=max_seq_len,
52
- sliding_window_layers=self.sliding_window_layers,
53
- )
54
- new_causal_lm = DecoderOnlyForCausalLM(causal_lm, new_model)
55
- return new_causal_lm
35
+ def get_rbln_attn_class(self):
36
+ return GPT2Attention
37
+
38
+ def get_rbln_layer_class(self):
39
+ return GPT2Layer
40
+
41
+ def get_rbln_model_class(self):
42
+ return GPT2Model
43
+
44
+ def get_attn_layer(self, layer: nn.Module):
45
+ return layer.attn
46
+
47
+ def get_model_layer(self, causal_lm: "GPT2LMHeadModel"):
48
+ return causal_lm.transformer
49
+
50
+ def get_decoder_layers(self, causal_lm: PreTrainedModel):
51
+ return causal_lm.transformer.h
56
52
 
57
53
 
58
54
  class GPT2Model(DecoderOnlyModel):
@@ -20,7 +20,6 @@ import torch.nn as nn
20
20
 
21
21
  from ..decoderonly.decoderonly_architecture import (
22
22
  DecoderOnlyAttention,
23
- DecoderOnlyForCausalLM,
24
23
  DecoderOnlyLayer,
25
24
  DecoderOnlyModel,
26
25
  DecoderOnlyWrapper,
@@ -55,27 +54,20 @@ class MidmLMHeadModelWrapper(DecoderOnlyWrapper):
55
54
  self.config.partial_rotary_factor = self.config.rotary_percentage
56
55
  return super().get_rotary_emb(max_seq_len=max_seq_len)
57
56
 
58
- def convert_to_rbln_causal_lm(self, causal_lm: "MidmLMHeadModel", max_seq_len: int):
59
- if self.attn_impl != "eager":
60
- raise NotImplementedError(f"flash attention ({self.attn_impl}) is not implemented for {self.__class__}")
61
- new_layers = []
62
- for layer in causal_lm.transformer.h:
63
- new_self_attn = MidmAttention(
64
- layer.attn,
65
- self.use_attention_mask,
66
- kvcache_block_size=self.kvcache_block_size,
67
- use_position_ids=self.use_position_ids,
68
- )
69
- new_layer = MidmLayer(layer, new_self_attn)
70
- new_layers.append(new_layer)
71
- new_model = MidmModel(
72
- causal_lm.transformer,
73
- new_layers,
74
- max_seq_len=max_seq_len,
75
- sliding_window_layers=self.sliding_window_layers,
76
- )
77
- new_causal_lm = DecoderOnlyForCausalLM(causal_lm, new_model)
78
- return new_causal_lm
57
+ def get_rbln_attn_class(self):
58
+ return MidmAttention
59
+
60
+ def get_rbln_layer_class(self):
61
+ return MidmLayer
62
+
63
+ def get_rbln_model_class(self):
64
+ return MidmModel
65
+
66
+ def get_model_layer(self, causal_lm: "MidmLMHeadModel"):
67
+ return causal_lm.transformer
68
+
69
+ def get_decoder_layers(self, causal_lm: "MidmLMHeadModel"):
70
+ return causal_lm.transformer.h
79
71
 
80
72
 
81
73
  class MidmModel(DecoderOnlyModel):
@@ -18,7 +18,6 @@ import torch.nn as nn
18
18
 
19
19
  from ...models.decoderonly.decoderonly_architecture import (
20
20
  DecoderOnlyAttention,
21
- DecoderOnlyForCausalLM,
22
21
  DecoderOnlyLayer,
23
22
  DecoderOnlyModel,
24
23
  DecoderOnlyWrapper,
@@ -30,30 +29,22 @@ if TYPE_CHECKING:
30
29
 
31
30
 
32
31
  class OPTWrapper(DecoderOnlyWrapper):
33
- def convert_to_rbln_causal_lm(self, causal_lm: "OPTForCausalLM", max_seq_len: int):
34
- if self.attn_impl != "eager":
35
- raise NotImplementedError(f"flash attention ({self.attn_impl}) is not implemented for {self.__class__}")
36
-
37
- new_layers = []
38
-
39
- for layer in causal_lm.model.decoder.layers:
40
- new_self_attn = OPTAttention(
41
- layer.self_attn,
42
- self.use_attention_mask,
43
- kvcache_block_size=self.kvcache_block_size,
44
- use_position_ids=self.use_position_ids,
45
- )
46
- new_layer = OPTDecoderLayer(layer, new_self_attn)
47
- new_layers.append(new_layer)
48
- new_model = OPTModel(
49
- causal_lm.model.decoder,
50
- new_layers,
51
- max_seq_len=max_seq_len,
52
- use_learned_pos_emb=True,
53
- sliding_window_layers=self.sliding_window_layers,
54
- )
55
- new_causal_lm = DecoderOnlyForCausalLM(causal_lm, new_model)
56
- return new_causal_lm
32
+ _use_learned_pos_emb = True
33
+
34
+ def get_rbln_attn_class(self):
35
+ return OPTAttention
36
+
37
+ def get_rbln_layer_class(self):
38
+ return OPTDecoderLayer
39
+
40
+ def get_rbln_model_class(self):
41
+ return OPTModel
42
+
43
+ def get_model_layer(self, causal_lm: "OPTForCausalLM"):
44
+ return causal_lm.model.decoder
45
+
46
+ def get_decoder_layers(self, causal_lm: "OPTForCausalLM"):
47
+ return causal_lm.model.decoder.layers
57
48
 
58
49
 
59
50
  class OPTAttention(DecoderOnlyAttention):
@@ -19,7 +19,6 @@ from transformers import PhiForCausalLM
19
19
 
20
20
  from ..decoderonly.decoderonly_architecture import (
21
21
  DecoderOnlyAttention,
22
- DecoderOnlyForCausalLM,
23
22
  DecoderOnlyLayer,
24
23
  DecoderOnlyModel,
25
24
  DecoderOnlyWrapper,
@@ -32,25 +31,20 @@ if TYPE_CHECKING:
32
31
 
33
32
 
34
33
  class PhiWrapper(DecoderOnlyWrapper):
35
- def convert_to_rbln_causal_lm(self, causal_lm: "PhiForCausalLM", max_seq_len: int):
36
- new_layers = []
37
- for layer in causal_lm.model.layers:
38
- if self.attn_impl == "eager":
39
- new_self_attn = PhiAttention(
40
- layer.self_attn,
41
- self.use_attention_mask,
42
- kvcache_block_size=self.kvcache_block_size,
43
- use_position_ids=self.use_position_ids,
44
- )
45
- elif self.attn_impl == "flash_attn":
46
- raise NotImplementedError(f"flash attn for {self.__class__} is not implemented yet.")
47
- else:
48
- raise NotImplementedError(f"Unknwon attn : {self.attn_impl}")
49
- new_layer = PhiLayer(layer, new_self_attn)
50
- new_layers.append(new_layer)
51
- new_model = PhiModel(causal_lm.model, new_layers, sliding_window_layers=self.sliding_window_layers)
52
- new_causal_lm = DecoderOnlyForCausalLM(causal_lm, new_model)
53
- return new_causal_lm
34
+ def get_rbln_attn_class(self):
35
+ return PhiAttention
36
+
37
+ def get_rbln_layer_class(self):
38
+ return PhiLayer
39
+
40
+ def get_rbln_model_class(self):
41
+ return PhiModel
42
+
43
+ def get_model_layer(self, causal_lm: "PhiForCausalLM"):
44
+ return causal_lm.model
45
+
46
+ def get_decoder_layers(self, causal_lm: "PhiForCausalLM"):
47
+ return causal_lm.model.layers
54
48
 
55
49
 
56
50
  class PhiAttention(DecoderOnlyAttention):
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: optimum-rbln
3
- Version: 0.8.1rc1
3
+ Version: 0.8.2a0
4
4
  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.
5
5
  Project-URL: Homepage, https://rebellions.ai
6
6
  Project-URL: Documentation, https://docs.rbln.ai
@@ -1,5 +1,5 @@
1
1
  optimum/rbln/__init__.py,sha256=MZCYmY4Y_Zfk0TGo3xK52osHDLZHz4cSdduXZt6RfSI,15316
2
- optimum/rbln/__version__.py,sha256=_HBfPPqmrksIsmr93RnGS_EXQHyrC_e0NtnXkbI_leI,521
2
+ optimum/rbln/__version__.py,sha256=kKne35dFUj-l3bjR0tLZka8O-dDdB-rFDsjhN13A2r4,519
3
3
  optimum/rbln/configuration_utils.py,sha256=o5oer7fBdE-MHLGNXoP35FjmuQbMmjEIDv0QE_k3kpo,32336
4
4
  optimum/rbln/modeling.py,sha256=bsvK6GQtoH9vx72Ea59kvv61jguOk9XDTzVjsY1ugkk,14248
5
5
  optimum/rbln/modeling_base.py,sha256=QpNkU_Do__JKmnHjaPzv47OhQwgGfVohisip1jqXa7A,23871
@@ -8,16 +8,16 @@ optimum/rbln/diffusers/modeling_diffusers.py,sha256=RjZNcYMU5daUIj-PAxyAwVoo2a9h
8
8
  optimum/rbln/diffusers/configurations/__init__.py,sha256=vMRnPY4s-Uju43xP038D2EA18X_mhy2YfsZVpSU-VoA,1322
9
9
  optimum/rbln/diffusers/configurations/models/__init__.py,sha256=7q95gtgDzCeIBogGw8SLQoHT4Wch7vpLJVF2UQovuoo,567
10
10
  optimum/rbln/diffusers/configurations/models/configuration_autoencoder_kl.py,sha256=ySetuNq6koleFIZ542zZLTzEEyl_CTul9l12ufWlQ_Y,3218
11
- optimum/rbln/diffusers/configurations/models/configuration_autoencoder_kl_cosmos.py,sha256=0BPZ4Tcz72j0a1KPdfpvk5G511wWNvS4MDJKbHAPzhA,4145
11
+ optimum/rbln/diffusers/configurations/models/configuration_autoencoder_kl_cosmos.py,sha256=SJfgbUz1LlRVuKQ_sHwPS262oOHF2TliKqM2z13wjEw,4172
12
12
  optimum/rbln/diffusers/configurations/models/configuration_controlnet.py,sha256=VDO_YFS_QhcHhuRIXQL53JZXEO27yoKHtecq5hd2la8,2637
13
13
  optimum/rbln/diffusers/configurations/models/configuration_prior_transformer.py,sha256=vE8RsXc27Z4-9k0KEM_vP7AWd5UUYvDgfX1g6nUrPp4,2224
14
- optimum/rbln/diffusers/configurations/models/configuration_transformer_cosmos.py,sha256=WOsC3sufsbwEEtsx0q8KqXUQE2VAFFCKRC7f2H4fmhg,3100
14
+ optimum/rbln/diffusers/configurations/models/configuration_transformer_cosmos.py,sha256=tqzBWzkl5PX60v8REGHuUC1WdJuIQv_2BGUOne5UYL8,3127
15
15
  optimum/rbln/diffusers/configurations/models/configuration_transformer_sd3.py,sha256=TAwHUyVy_9HSEZdXIuFCtrBfNIuYIedklJaCut5wEys,2412
16
16
  optimum/rbln/diffusers/configurations/models/configuration_unet_2d_condition.py,sha256=mxcbrOqLMnPpP-jnjSeRWPj2zwPMsgeQSq6LzhG2btA,3630
17
17
  optimum/rbln/diffusers/configurations/models/configuration_vq_model.py,sha256=dslGcfCZL_hNeVyjV-4FnCT1POmXuiaLbr6NcQSKgHg,3259
18
18
  optimum/rbln/diffusers/configurations/pipelines/__init__.py,sha256=RfJXQiYvgGc3Rp7JYk5s0AQd0XB5JCAb37_riGWQAYg,1268
19
19
  optimum/rbln/diffusers/configurations/pipelines/configuration_controlnet.py,sha256=nTtr2vqyr3zNSJXI0kiTAhOnVNhA-cVyaSnKOwBBZIo,14215
20
- optimum/rbln/diffusers/configurations/pipelines/configuration_cosmos.py,sha256=8GK9pAGzhtmQN2DQjgQZJo9S7y6dDko-Ii-3CagFxo4,4588
20
+ optimum/rbln/diffusers/configurations/pipelines/configuration_cosmos.py,sha256=tncXVraSYfrezqL9cT4kg5nuoifzYVfP0qHbgg0QUjA,4615
21
21
  optimum/rbln/diffusers/configurations/pipelines/configuration_kandinsky2_2.py,sha256=1ve6o4OEpjPzTXWHXy_T5MAI0V-F08PMv2W6nBFfeKU,16386
22
22
  optimum/rbln/diffusers/configurations/pipelines/configuration_stable_diffusion.py,sha256=kR8dV_RsmoDxhK5bAfv3PbtS5LpN5g-O-snAX1sP6Fo,6591
23
23
  optimum/rbln/diffusers/configurations/pipelines/configuration_stable_diffusion_3.py,sha256=f2VOwvCd-9kDnUpwhb0LaMWgfwdmBzUKMpmCdhUv2sc,7923
@@ -44,7 +44,7 @@ optimum/rbln/diffusers/pipelines/controlnet/pipeline_controlnet_sd_xl.py,sha256=
44
44
  optimum/rbln/diffusers/pipelines/controlnet/pipeline_controlnet_sd_xl_img2img.py,sha256=HX56itORMqXLjZcwv25C-_z3JyZn3v6BpfIjsrDO3mE,46640
45
45
  optimum/rbln/diffusers/pipelines/cosmos/__init__.py,sha256=h2j6S8IJPVHeNU8qmW9vyXMgHBw0d7kQcuMAA5YoHPU,795
46
46
  optimum/rbln/diffusers/pipelines/cosmos/configuration_cosmos_guardrail.py,sha256=kDVnUBBGdumpDj2DaOpo5MSsFvlFIGY6BU1LZaFVqao,3327
47
- optimum/rbln/diffusers/pipelines/cosmos/cosmos_guardrail.py,sha256=lzT6j0RifZfIktETC5gd8GAKPHwekeb3JH56aPChOtI,18088
47
+ optimum/rbln/diffusers/pipelines/cosmos/cosmos_guardrail.py,sha256=EAt2UICPRTaUz4SNsQYOa9aoW0USj2qamqdhlf2ajrA,18261
48
48
  optimum/rbln/diffusers/pipelines/cosmos/pipeline_cosmos_text2world.py,sha256=TfhgAWVHUHvxsagBGLAVYKBoSMvuH7rg_xP5ZZ0rVU0,3910
49
49
  optimum/rbln/diffusers/pipelines/cosmos/pipeline_cosmos_video2world.py,sha256=-dl8AMwSuorIOxRNfyu1XhkJfmNVbSo3_Wkb2gAmUpo,3917
50
50
  optimum/rbln/diffusers/pipelines/kandinsky2_2/__init__.py,sha256=I4YQq2HfA3xONbWsdJ870IEJPyLWeCDDG-UCJsu9YO8,1035
@@ -90,18 +90,18 @@ optimum/rbln/transformers/models/bert/__init__.py,sha256=86FuGRBLw315_Roa9D5OUx6
90
90
  optimum/rbln/transformers/models/bert/configuration_bert.py,sha256=nEZnX6LXpLKWaoPEd4pWSysw9h-PLb2ld0ibC3dcJ7w,1611
91
91
  optimum/rbln/transformers/models/bert/modeling_bert.py,sha256=zR0US2laTT0yUkL6yyvrR5STQNJcYqtG98ez4SUYQAY,2040
92
92
  optimum/rbln/transformers/models/blip_2/__init__.py,sha256=L01gPXcUCa8Vg-bcng20vZvBIN_jlqCzwUSFuq0QOag,855
93
- optimum/rbln/transformers/models/blip_2/configuration_blip_2.py,sha256=Xl24Ao6xLblSR6u1TK--OHDsdKBVG6sDZz_hcEeoScY,3183
93
+ optimum/rbln/transformers/models/blip_2/configuration_blip_2.py,sha256=ke75GqPU139dNOY1nm6QE661LepbD_0V9Bx1QbtHhKA,3210
94
94
  optimum/rbln/transformers/models/blip_2/modeling_blip_2.py,sha256=2sIVGrIn1B2nUZ8hw1sgW3VbJ2vxrlBRN37GgDiw0GU,16191
95
95
  optimum/rbln/transformers/models/clip/__init__.py,sha256=TLeXDqcFK6M6v9x7Xr64kBbqGu3hFHM7p754dQ8UVQc,938
96
96
  optimum/rbln/transformers/models/clip/configuration_clip.py,sha256=D7CIWpbMhXUrGv-CnhxRtSS3vAYb427-w7zSkfuJHEU,3455
97
97
  optimum/rbln/transformers/models/clip/modeling_clip.py,sha256=QbYrt7pUWNal-p93fxmuKrHa2CPlCaq8F16qOfMAst0,8090
98
98
  optimum/rbln/transformers/models/colpali/__init__.py,sha256=n3rueXT_oC0N8myoZiic0YkVK24CW5hZBPa-0L8so6Y,119
99
99
  optimum/rbln/transformers/models/colpali/colpali_architecture.py,sha256=bWG7TehWRZkTh2y6mGkpd85_onWAyiyKdaQC9TFsy3E,8065
100
- optimum/rbln/transformers/models/colpali/configuration_colpali.py,sha256=yPzLYON6qRJlBkzxFfIBzBWd2KjYWvdClO4iAqd_V7E,2609
100
+ optimum/rbln/transformers/models/colpali/configuration_colpali.py,sha256=ieY-tuyDPObFUIJ5sfpcfuCsJ_HTAizN7ZGqirqeFRU,2636
101
101
  optimum/rbln/transformers/models/colpali/modeling_colpali.py,sha256=jzvJCBrrCXSpjfmJ3O-VvPNFGWGaNbpOV09JwLPAZWs,15757
102
102
  optimum/rbln/transformers/models/decoderonly/__init__.py,sha256=vQYZDDdoddwA7yKc5zzrq2Zs9sax-0p8rNF_aYfF4bk,1006
103
103
  optimum/rbln/transformers/models/decoderonly/configuration_decoderonly.py,sha256=cakn8RGo8gS3nmXdEqOfC2xUBOMGInROgLEbCOoLFR0,13398
104
- optimum/rbln/transformers/models/decoderonly/decoderonly_architecture.py,sha256=YAn8J_lIq4IS-HM_gbi5Qov8_osxhWtBr5z_28QRbGM,49667
104
+ optimum/rbln/transformers/models/decoderonly/decoderonly_architecture.py,sha256=8ovJ5_q_asqVTuVnAuK1m6genW0OSJ30Cd7HS9JXJgc,46363
105
105
  optimum/rbln/transformers/models/decoderonly/modeling_decoderonly.py,sha256=NmWdodIcXXChI61-Ej7StTe52iQvalMYRUDuNtcQVEI,53342
106
106
  optimum/rbln/transformers/models/distilbert/__init__.py,sha256=zXL78SOEORTnUN_wrdoaDaYpntG8lcFHvPobM6jC0CI,841
107
107
  optimum/rbln/transformers/models/distilbert/configuration_distilbert.py,sha256=O3BW9JjyYk9PLyiofvOKEgTdMZ_jpIuPfot281pSsyg,984
@@ -111,19 +111,19 @@ optimum/rbln/transformers/models/dpt/configuration_dpt.py,sha256=3Bb_K0sKI6TKeoH
111
111
  optimum/rbln/transformers/models/dpt/modeling_dpt.py,sha256=uIwdHAhGgSyj_ljwJsRv6i5nUr9lTzB2Ss0iz0HplfY,978
112
112
  optimum/rbln/transformers/models/exaone/__init__.py,sha256=eUL0mq3yGVzCQfjLlOtVF2MecIN3DQWm07EmXubGSTs,921
113
113
  optimum/rbln/transformers/models/exaone/configuration_exaone.py,sha256=S4s4kJemPbmn-otYv-XNHE40DJaEYY6cmzaWV6MTGsY,1388
114
- optimum/rbln/transformers/models/exaone/exaone_architecture.py,sha256=-gCUDIJ1SJqNgIALJejJ75XPtAkd83oyomBke9xGRsc,3610
114
+ optimum/rbln/transformers/models/exaone/exaone_architecture.py,sha256=lY4FwH2EZn_OY6sBIHlwxbfaEOEJ1eueUQJGB6Js62M,2306
115
115
  optimum/rbln/transformers/models/exaone/modeling_exaone.py,sha256=sr_ICK-rw_fYmLY5r0IOc-vDtSZEcSwFIQp3Gn92zqE,3929
116
116
  optimum/rbln/transformers/models/gemma/__init__.py,sha256=VqPIlokw3kjn_ZoLXINCLXw3vaysQFo5oPGGy6bnt4Q,708
117
117
  optimum/rbln/transformers/models/gemma/configuration_gemma.py,sha256=3hAxl7LL9vFpCHrs-g3BwVDdVjnnJ-fzSO88wdfyGDQ,1361
118
- optimum/rbln/transformers/models/gemma/gemma_architecture.py,sha256=I9EyRIEtMw9u7HhSa8PqUco4uNe3gl6_lx0r1uDT8hA,2546
118
+ optimum/rbln/transformers/models/gemma/gemma_architecture.py,sha256=2Ivay8NTSHmQAqXFh9JvG6Ja5rMThcRAjYPzyipcRI8,956
119
119
  optimum/rbln/transformers/models/gemma/modeling_gemma.py,sha256=Ojvum34EhDHWfMB4D6S1BrwoTNwuBSZuBzwdnAgvq38,3095
120
120
  optimum/rbln/transformers/models/gemma3/__init__.py,sha256=6rugk3615SEt4lh7gduo_J9VyGiSReuEIvL0Uno0eaI,790
121
121
  optimum/rbln/transformers/models/gemma3/configuration_gemma3.py,sha256=eupMGTHJGJNNrAZ3GE6M6GQBAQzBb7KFJvalyDmbM-A,3063
122
- optimum/rbln/transformers/models/gemma3/gemma3_architecture.py,sha256=sgFQQbvEr15tb2Sxk_tgcgQFcjhKGbNSW6fm2u7-Vck,8609
122
+ optimum/rbln/transformers/models/gemma3/gemma3_architecture.py,sha256=fpLDAXCe5paWVsfc0tL59JkRQMRF-WNgIzOIb_QpSLU,6191
123
123
  optimum/rbln/transformers/models/gemma3/modeling_gemma3.py,sha256=-cpU0ot46VFUZ6PtfwN9VJ-E44n6mP1E3dKwB99MtBM,38389
124
124
  optimum/rbln/transformers/models/gpt2/__init__.py,sha256=socBMIBZSiLbrVN12rQ4nL9gFeT0axMgz6SWaCaD4Ac,704
125
125
  optimum/rbln/transformers/models/gpt2/configuration_gpt2.py,sha256=9sS6-EGapmow3rG9ViejK9qwrqy_X86VBxQ7u9x0Yqk,923
126
- optimum/rbln/transformers/models/gpt2/gpt2_architecture.py,sha256=pnGgixjgjW7HULbs5211cC2guw_4e4-MlS69vdCRMMg,3206
126
+ optimum/rbln/transformers/models/gpt2/gpt2_architecture.py,sha256=O7hBiaFJrpLSswGwW83cX9S9Q2wKRBDrpAqOgOS7zQg,2733
127
127
  optimum/rbln/transformers/models/gpt2/modeling_gpt2.py,sha256=qBDanUk_O-HtOIVCA4IE3FYyCsnL9xIDK00vft-0caw,1490
128
128
  optimum/rbln/transformers/models/idefics3/__init__.py,sha256=ulxE7HEfXsNJhd25J9Fvi6vggo9aZH9sLKJjWB6LlzQ,814
129
129
  optimum/rbln/transformers/models/idefics3/configuration_idefics3.py,sha256=wKroy3m65zS41G80QXssbndHoHU8wtHTteGU2Q6qbws,2390
@@ -137,7 +137,7 @@ optimum/rbln/transformers/models/llava_next/configuration_llava_next.py,sha256=b
137
137
  optimum/rbln/transformers/models/llava_next/modeling_llava_next.py,sha256=paYtCk58--FSZp8xjVrfZAxkJxO02X-jxaVPqL-l7ZU,27421
138
138
  optimum/rbln/transformers/models/midm/__init__.py,sha256=IC3FETwgYinbp3wDj7tp4zIHJhbqM-c6GfTRdYcMNj8,913
139
139
  optimum/rbln/transformers/models/midm/configuration_midm.py,sha256=DxhcSJlApxfi00XxYmSkKZ6bY9vfLXT0zh-oMKkZot0,1365
140
- optimum/rbln/transformers/models/midm/midm_architecture.py,sha256=XXY_uDGkXeVQnKpmSWrgljgxtSdTgLLFLfMqtZdRJdM,5642
140
+ optimum/rbln/transformers/models/midm/midm_architecture.py,sha256=RlkmNhaWE5h_awt9aTtR8VZfshNTah0IoUfD2Z9vfxI,5055
141
141
  optimum/rbln/transformers/models/midm/modeling_midm.py,sha256=zbziYZ3f_dX_MOLwORTfJn22psZ1g3FFeQffM_TIh7A,3876
142
142
  optimum/rbln/transformers/models/mistral/__init__.py,sha256=9FE64bCYfSIyrBkRcwlqF8QyacSJFWvwEufHFi1ZIrM,716
143
143
  optimum/rbln/transformers/models/mistral/configuration_mistral.py,sha256=pMYJSwqmtx0uD2uExHx4S-JXal9rqQ5A2ulT2IoglTg,1383
@@ -146,11 +146,11 @@ optimum/rbln/transformers/models/mistral/modeling_mistral.py,sha256=SGzmn9EJeM27
146
146
  optimum/rbln/transformers/models/opt/__init__.py,sha256=w0v8GzbzlR5_4yL851njGDSJgX89TrYxrHnpNfMHZEI,700
147
147
  optimum/rbln/transformers/models/opt/configuration_opt.py,sha256=HgNCxnuoyZZwPoDMU41nvXG5DU9UHHSG8gvUSsm-r34,920
148
148
  optimum/rbln/transformers/models/opt/modeling_opt.py,sha256=aDijHHFOWBAjCJ_YrI7dcmuVuY69S1QD0115MQO9YFU,3667
149
- optimum/rbln/transformers/models/opt/opt_architecture.py,sha256=xRN0nNoZB4ZxKOmliFkI0xFQ1jy0hs42dv-hMyGOZ_Q,2802
149
+ optimum/rbln/transformers/models/opt/opt_architecture.py,sha256=El2l0n7YUWTakzZvqWyu58KNEbCc6zoHQhkqSLSsVm0,2202
150
150
  optimum/rbln/transformers/models/phi/__init__.py,sha256=uqQb-sO1HXuaju2hfo7qJHk_IWhnptY-qFjNjK_uOc0,700
151
151
  optimum/rbln/transformers/models/phi/configuration_phi.py,sha256=58jv3bIo_BcPcS9wU6NVgh67mGpHafdoQzStLKmfuU4,1349
152
152
  optimum/rbln/transformers/models/phi/modeling_phi.py,sha256=sd8XYKJkpZM7pWqN0DE7B-dJuTpF9b2_ebZgJK1AuJ8,3061
153
- optimum/rbln/transformers/models/phi/phi_architecture.py,sha256=nv3jx0zWeExYDSr9xHzgG8pssxC5qe6QGy6HmZKXYas,4241
153
+ optimum/rbln/transformers/models/phi/phi_architecture.py,sha256=yShRckC62i2nky3MvM_qGrhTXOo3FswwtNxd5fogduM,3574
154
154
  optimum/rbln/transformers/models/qwen2/__init__.py,sha256=Tu4_AXy3ktTvxGwxED3kewiv62S75HgDWD6-TeC1DfA,708
155
155
  optimum/rbln/transformers/models/qwen2/configuration_qwen2.py,sha256=Jc7qTFQgB9tbhJ-aPDN_lfyz9u0omNL84HWYBQ5fvcs,1359
156
156
  optimum/rbln/transformers/models/qwen2/modeling_qwen2.py,sha256=OKd7SXQLLtzPVolr26P1TvCV7Gf0XG7k6BjzjuvrL4s,3885
@@ -205,7 +205,7 @@ optimum/rbln/utils/model_utils.py,sha256=4k5879Kh75m3x_vS4-qOGfqsOiAvc2kdNFFfvsF
205
205
  optimum/rbln/utils/runtime_utils.py,sha256=D9PS8hfH1NBf8yH8cAu-XfdC9fxKzPbt4LFBVpADbbs,7180
206
206
  optimum/rbln/utils/save_utils.py,sha256=hG5uOtYmecSXZuGTvCXsTM-SiyZpr5q3InUGCCq_jzQ,3619
207
207
  optimum/rbln/utils/submodule.py,sha256=w5mgPgncI740gVKMu3S-69DGNdUSI0bTZxegQGcZ98Y,5011
208
- optimum_rbln-0.8.1rc1.dist-info/METADATA,sha256=52EcXRgXYVTSOWiLL_fQhxzsbZFrESQLFpluOvePehw,5300
209
- optimum_rbln-0.8.1rc1.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
210
- optimum_rbln-0.8.1rc1.dist-info/licenses/LICENSE,sha256=QwcOLU5TJoTeUhuIXzhdCEEDDvorGiC6-3YTOl4TecE,11356
211
- optimum_rbln-0.8.1rc1.dist-info/RECORD,,
208
+ optimum_rbln-0.8.2a0.dist-info/METADATA,sha256=dHMIEdFF_IuTWww99Iypz6HQKVDDk___EVJ8cK77eG0,5299
209
+ optimum_rbln-0.8.2a0.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
210
+ optimum_rbln-0.8.2a0.dist-info/licenses/LICENSE,sha256=QwcOLU5TJoTeUhuIXzhdCEEDDvorGiC6-3YTOl4TecE,11356
211
+ optimum_rbln-0.8.2a0.dist-info/RECORD,,