♻️ use get_sampler in all models

Signed-off-by: Joe Runde <Joseph.Runde@ibm.com>

vllm/model_executor/models/arctic.py

@@ -23,7 +23,7 @@
 from vllm.model_executor.layers.quantization.deepspeedfp import (
     DeepSpeedFPConfig, DeepSpeedFPParameter)
 from vllm.model_executor.layers.rotary_embedding import get_rope
-from vllm.model_executor.layers.sampler import Sampler, SamplerOutput
+from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     ParallelLMHead, VocabParallelEmbedding)
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
@@ -436,7 +436,7 @@ def __init__(self,
         self.unpadded_vocab_size = config.vocab_size
         self.logits_processor = LogitsProcessor(self.unpadded_vocab_size,
                                                 config.vocab_size)
-        self.sampler = Sampler()
+        self.sampler = get_sampler()
         self.make_empty_intermediate_tensors = (
             self.model.make_empty_intermediate_tensors)
 

vllm/model_executor/models/baichuan.py

@@ -38,7 +38,7 @@
 from vllm.model_executor.layers.logits_processor import LogitsProcessor
 from vllm.model_executor.layers.quantization import QuantizationConfig
 from vllm.model_executor.layers.rotary_embedding import get_rope
-from vllm.model_executor.layers.sampler import Sampler, SamplerOutput
+from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     ParallelLMHead, VocabParallelEmbedding)
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
@@ -353,7 +353,7 @@ def __init__(
         if self.config.tie_word_embeddings:
             self.lm_head.weight = self.model.embed_tokens.weight
         self.logits_processor = LogitsProcessor(config.vocab_size)
-        self.sampler = Sampler()
+        self.sampler = get_sampler()
         self.make_empty_intermediate_tensors = (
             self.model.make_empty_intermediate_tensors)
 

vllm/model_executor/models/bart.py

@@ -34,7 +34,7 @@
 from vllm.model_executor.layers.logits_processor import LogitsProcessor
 from vllm.model_executor.layers.quantization.base_config import (
     QuantizationConfig)
-from vllm.model_executor.layers.sampler import Sampler, SamplerOutput
+from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     ParallelLMHead, VocabParallelEmbedding)
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
@@ -842,7 +842,7 @@ def __init__(self,
 
         self.logits_processor = LogitsProcessor(self.unpadded_vocab_size,
                                                 config.vocab_size)
-        self.sampler = Sampler()
+        self.sampler = get_sampler()
 
     def forward(
         self,

vllm/model_executor/models/blip2.py

@@ -13,7 +13,7 @@
                          token_inputs)
 from vllm.model_executor.layers.activation import get_act_fn
 from vllm.model_executor.layers.quantization import QuantizationConfig
-from vllm.model_executor.layers.sampler import Sampler, SamplerOutput
+from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.sampling_metadata import SamplingMetadata
 from vllm.multimodal import MULTIMODAL_REGISTRY
 from vllm.sequence import IntermediateTensors, SequenceData
@@ -520,7 +520,7 @@ def sampler(self):
         if hasattr(self.language_model, "sampler"):
             return self.language_model.sampler
 
-        return Sampler()
+        return get_sampler()
 
     def _validate_pixel_values(self, data: torch.Tensor) -> torch.Tensor:
         h = w = self.config.vision_config.image_size

vllm/model_executor/models/bloom.py

@@ -34,7 +34,7 @@
                                                RowParallelLinear)
 from vllm.model_executor.layers.logits_processor import LogitsProcessor
 from vllm.model_executor.layers.quantization import QuantizationConfig
-from vllm.model_executor.layers.sampler import Sampler, SamplerOutput
+from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     ParallelLMHead, VocabParallelEmbedding)
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
@@ -299,7 +299,7 @@ def __init__(
                                           self.config.hidden_size)
 
         self.logits_processor = LogitsProcessor(config.vocab_size)
-        self.sampler = Sampler()
+        self.sampler = get_sampler()
         self.make_empty_intermediate_tensors = (
             self.transformer.make_empty_intermediate_tensors)
 

vllm/model_executor/models/chameleon.py

@@ -21,7 +21,7 @@
 from vllm.model_executor.layers.logits_processor import LogitsProcessor
 from vllm.model_executor.layers.quantization import QuantizationConfig
 from vllm.model_executor.layers.rotary_embedding import get_rope
-from vllm.model_executor.layers.sampler import Sampler, SamplerOutput
+from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     ParallelLMHead, VocabParallelEmbedding)
 from vllm.model_executor.model_loader.weight_utils import (
@@ -936,7 +936,7 @@ def __init__(
         logit_scale = getattr(config, "logit_scale", 1.0)
         self.logits_processor = LogitsProcessor(self.unpadded_vocab_size,
                                                 config.vocab_size, logit_scale)
-        self.sampler = Sampler()
+        self.sampler = get_sampler()
         self.make_empty_intermediate_tensors = (
             self.model.make_empty_intermediate_tensors)
 

vllm/model_executor/models/chatglm.py

@@ -25,7 +25,7 @@
 from vllm.model_executor.layers.logits_processor import LogitsProcessor
 from vllm.model_executor.layers.quantization import QuantizationConfig
 from vllm.model_executor.layers.rotary_embedding import get_rope
-from vllm.model_executor.layers.sampler import Sampler, SamplerOutput
+from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     ParallelLMHead, VocabParallelEmbedding)
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
@@ -618,7 +618,7 @@ def __init__(
                 self.transformer.embedding.weight)
         self.lm_head = self.transformer.output_layer
         self.logits_processor = LogitsProcessor(config.padded_vocab_size)
-        self.sampler = Sampler()
+        self.sampler = get_sampler()
 
     def forward(self,
                 input_ids: torch.Tensor,

vllm/model_executor/models/commandr.py

@@ -38,7 +38,7 @@
 from vllm.model_executor.layers.logits_processor import LogitsProcessor
 from vllm.model_executor.layers.quantization import QuantizationConfig
 from vllm.model_executor.layers.rotary_embedding import get_rope
-from vllm.model_executor.layers.sampler import Sampler, SamplerOutput
+from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     VocabParallelEmbedding)
 from vllm.model_executor.model_loader.weight_utils import (
@@ -356,7 +356,7 @@ def __init__(
                                  cache_config,
                                  quant_config,
                                  lora_config=lora_config)
-        self.sampler = Sampler()
+        self.sampler = get_sampler()
         self.make_empty_intermediate_tensors = (
             self.model.make_empty_intermediate_tensors)
 

vllm/model_executor/models/dbrx.py

@@ -15,7 +15,7 @@
 from vllm.model_executor.layers.logits_processor import LogitsProcessor
 from vllm.model_executor.layers.quantization import QuantizationConfig
 from vllm.model_executor.layers.rotary_embedding import get_rope
-from vllm.model_executor.layers.sampler import Sampler, SamplerOutput
+from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     DEFAULT_VOCAB_PADDING_SIZE, ParallelLMHead, VocabParallelEmbedding)
 from vllm.model_executor.model_loader.weight_utils import (
@@ -374,7 +374,7 @@ def __init__(
         )
         self.logits_processor = LogitsProcessor(self.unpadded_vocab_size,
                                                 config.vocab_size)
-        self.sampler = Sampler()
+        self.sampler = get_sampler()
         self.make_empty_intermediate_tensors = (
             self.transformer.make_empty_intermediate_tensors)
 

vllm/model_executor/models/deepseek.py

@@ -42,7 +42,7 @@
 from vllm.model_executor.layers.logits_processor import LogitsProcessor
 from vllm.model_executor.layers.quantization import QuantizationConfig
 from vllm.model_executor.layers.rotary_embedding import get_rope
-from vllm.model_executor.layers.sampler import Sampler, SamplerOutput
+from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     ParallelLMHead, VocabParallelEmbedding)
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
@@ -400,7 +400,7 @@ def __init__(
         if self.config.tie_word_embeddings:
             self.lm_head.weight = self.model.embed_tokens.weight
         self.logits_processor = LogitsProcessor(config.vocab_size)
-        self.sampler = Sampler()
+        self.sampler = get_sampler()
         self.make_empty_intermediate_tensors = (
             self.model.make_empty_intermediate_tensors)
 

vllm/model_executor/models/deepseek_v2.py

@@ -43,7 +43,7 @@
 from vllm.model_executor.layers.logits_processor import LogitsProcessor
 from vllm.model_executor.layers.quantization import QuantizationConfig
 from vllm.model_executor.layers.rotary_embedding import get_rope
-from vllm.model_executor.layers.sampler import Sampler, SamplerOutput
+from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     ParallelLMHead, VocabParallelEmbedding)
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
@@ -497,7 +497,7 @@ def __init__(
                                       config.hidden_size,
                                       quant_config=quant_config)
         self.logits_processor = LogitsProcessor(config.vocab_size)
-        self.sampler = Sampler()
+        self.sampler = get_sampler()
         self.make_empty_intermediate_tensors = (
             self.model.make_empty_intermediate_tensors)
 

vllm/model_executor/models/exaone.py

@@ -43,7 +43,7 @@
 from vllm.model_executor.layers.quantization.compressed_tensors.utils import (
     get_compressed_tensors_cache_scale)
 from vllm.model_executor.layers.rotary_embedding import get_rope
-from vllm.model_executor.layers.sampler import Sampler, SamplerOutput
+from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     DEFAULT_VOCAB_PADDING_SIZE, ParallelLMHead, VocabParallelEmbedding)
 from vllm.model_executor.model_loader.weight_utils import (
@@ -479,7 +479,7 @@ def __init__(
             self.logits_processor = LogitsProcessor(self.unpadded_vocab_size,
                                                     config.vocab_size,
                                                     logit_scale)
-            self.sampler = Sampler()
+            self.sampler = get_sampler()
         else:
             self.lm_head = PPMissingLayer()
 

vllm/model_executor/models/falcon.py

@@ -38,7 +38,7 @@
 from vllm.model_executor.layers.logits_processor import LogitsProcessor
 from vllm.model_executor.layers.quantization import QuantizationConfig
 from vllm.model_executor.layers.rotary_embedding import get_rope
-from vllm.model_executor.layers.sampler import Sampler, SamplerOutput
+from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     ParallelLMHead, VocabParallelEmbedding)
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
@@ -427,7 +427,7 @@ def __init__(
                 quant_config=quant_config,
             )
         self.logits_processor = LogitsProcessor(config.vocab_size)
-        self.sampler = Sampler()
+        self.sampler = get_sampler()
         self.make_empty_intermediate_tensors = (
             self.transformer.make_empty_intermediate_tensors)
 

vllm/model_executor/models/florence2.py

@@ -10,7 +10,7 @@
 from vllm.model_executor.layers.logits_processor import LogitsProcessor
 from vllm.model_executor.layers.quantization.base_config import (
     QuantizationConfig)
-from vllm.model_executor.layers.sampler import Sampler, SamplerOutput
+from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
 from vllm.model_executor.models.bart import (BartDecoder, BartEncoder,
                                              BartParallelLMHead,
@@ -112,7 +112,7 @@ def __init__(self,
 
         self.logits_processor = LogitsProcessor(self.vocab_size,
                                                 config.vocab_size)
-        self.sampler = Sampler()
+        self.sampler = get_sampler()
 
     def forward(
         self,

vllm/model_executor/models/gemma.py

@@ -34,7 +34,7 @@
 from vllm.model_executor.layers.logits_processor import LogitsProcessor
 from vllm.model_executor.layers.quantization import QuantizationConfig
 from vllm.model_executor.layers.rotary_embedding import get_rope
-from vllm.model_executor.layers.sampler import Sampler, SamplerOutput
+from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     VocabParallelEmbedding)
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
@@ -397,7 +397,7 @@ def __init__(
                                 quant_config,
                                 prefix=maybe_prefix(prefix, "model"))
         self.logits_processor = LogitsProcessor(config.vocab_size)
-        self.sampler = Sampler()
+        self.sampler = get_sampler()
         self.make_empty_intermediate_tensors = (
             self.model.make_empty_intermediate_tensors)
 

vllm/model_executor/models/gemma2.py

@@ -34,7 +34,7 @@
 from vllm.model_executor.layers.pooler import Pooler, PoolingType
 from vllm.model_executor.layers.quantization import QuantizationConfig
 from vllm.model_executor.layers.rotary_embedding import get_rope
-from vllm.model_executor.layers.sampler import Sampler, SamplerOutput
+from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     VocabParallelEmbedding)
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
@@ -417,7 +417,7 @@ def __init__(
         self.model = Gemma2Model(config, cache_config, quant_config)
         self.logits_processor = LogitsProcessor(
             config.vocab_size, soft_cap=config.final_logit_softcapping)
-        self.sampler = Sampler()
+        self.sampler = get_sampler()
         self.make_empty_intermediate_tensors = (
             self.model.make_empty_intermediate_tensors)
 

vllm/model_executor/models/gpt_bigcode.py

@@ -34,7 +34,7 @@
                                                RowParallelLinear)
 from vllm.model_executor.layers.logits_processor import LogitsProcessor
 from vllm.model_executor.layers.quantization import QuantizationConfig
-from vllm.model_executor.layers.sampler import Sampler, SamplerOutput
+from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     ParallelLMHead, VocabParallelEmbedding)
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
@@ -287,7 +287,7 @@ def __init__(
             self.unpadded_vocab_size += lora_config.lora_extra_vocab_size
         self.logits_processor = LogitsProcessor(self.unpadded_vocab_size,
                                                 config.vocab_size)
-        self.sampler = Sampler()
+        self.sampler = get_sampler()
         self.make_empty_intermediate_tensors = (
             self.transformer.make_empty_intermediate_tensors)
 

vllm/model_executor/models/gpt_j.py

@@ -33,7 +33,7 @@
 from vllm.model_executor.layers.logits_processor import LogitsProcessor
 from vllm.model_executor.layers.quantization import QuantizationConfig
 from vllm.model_executor.layers.rotary_embedding import get_rope
-from vllm.model_executor.layers.sampler import Sampler, SamplerOutput
+from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     ParallelLMHead, VocabParallelEmbedding)
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
@@ -248,7 +248,7 @@ def __init__(
             quant_config=quant_config,
         )
         self.logits_processor = LogitsProcessor(config.vocab_size)
-        self.sampler = Sampler()
+        self.sampler = get_sampler()
         self.make_empty_intermediate_tensors = (
             self.transformer.make_empty_intermediate_tensors)
 

vllm/model_executor/models/gpt_neox.py

@@ -33,7 +33,7 @@
 from vllm.model_executor.layers.logits_processor import LogitsProcessor
 from vllm.model_executor.layers.quantization import QuantizationConfig
 from vllm.model_executor.layers.rotary_embedding import get_rope
-from vllm.model_executor.layers.sampler import Sampler, SamplerOutput
+from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     ParallelLMHead, VocabParallelEmbedding)
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
@@ -262,7 +262,7 @@ def __init__(
         if self.config.tie_word_embeddings:
             self.embed_out.weight = self.gpt_neox.embed_in.weight
         self.logits_processor = LogitsProcessor(config.vocab_size)
-        self.sampler = Sampler()
+        self.sampler = get_sampler()
         self.make_empty_intermediate_tensors = (
             self.gpt_neox.make_empty_intermediate_tensors)
 

vllm/model_executor/models/granite.py

@@ -43,7 +43,7 @@
 from vllm.model_executor.layers.quantization.compressed_tensors.utils import (
     get_compressed_tensors_cache_scale)
 from vllm.model_executor.layers.rotary_embedding import get_rope
-from vllm.model_executor.layers.sampler import Sampler, SamplerOutput
+from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     DEFAULT_VOCAB_PADDING_SIZE, ParallelLMHead, VocabParallelEmbedding)
 from vllm.model_executor.model_loader.weight_utils import (
@@ -412,7 +412,7 @@ def __init__(
             self.logits_processor = LogitsProcessor(self.unpadded_vocab_size,
                                                     config.vocab_size,
                                                     scale=logit_scale)
-            self.sampler = Sampler()
+            self.sampler = get_sampler()
         else:
             self.lm_head = PPMissingLayer()
 

vllm/model_executor/models/granitemoe.py

@@ -40,7 +40,7 @@
 from vllm.model_executor.layers.quantization.base_config import (
     QuantizationConfig)
 from vllm.model_executor.layers.rotary_embedding import get_rope
-from vllm.model_executor.layers.sampler import Sampler, SamplerOutput
+from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     DEFAULT_VOCAB_PADDING_SIZE, ParallelLMHead, VocabParallelEmbedding)
 from vllm.model_executor.sampling_metadata import SamplingMetadata
@@ -372,7 +372,7 @@ def __init__(
                                                 scale=1 /
                                                 self.config.logits_scaling)
 
-        self.sampler = Sampler()
+        self.sampler = get_sampler()
 
     def forward(
         self,

vllm/model_executor/models/internlm2.py

@@ -21,7 +21,7 @@
 from vllm.model_executor.layers.logits_processor import LogitsProcessor
 from vllm.model_executor.layers.quantization import QuantizationConfig
 from vllm.model_executor.layers.rotary_embedding import get_rope
-from vllm.model_executor.layers.sampler import Sampler, SamplerOutput
+from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     ParallelLMHead, VocabParallelEmbedding)
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
@@ -339,7 +339,7 @@ def __init__(
         if self.config.tie_word_embeddings:
             self.output.weight = self.model.tok_embeddings.weight
         self.logits_processor = LogitsProcessor(config.vocab_size)
-        self.sampler = Sampler()
+        self.sampler = get_sampler()
         self.make_empty_intermediate_tensors = (
             self.model.make_empty_intermediate_tensors)