[Model] Changes to MLPSpeculator to support tie_weights and input_scale (#5965)

Signed-off-by: Thomas Parnell <tpa@zurich.ibm.com>
Co-authored-by: Joshua Rosenkranz <jmrosenk@us.ibm.com>

vllm/model_executor/models/mlp_speculator.py

@@ -13,6 +13,8 @@ from vllm.model_executor.model_loader.weight_utils import default_weight_loader
 from vllm.sequence import SamplerOutput
 from vllm.transformers_utils.configs import MLPSpeculatorConfig
 
+SQRT2 = 2**0.5
+
 
 class MLPSpeculatorLayerNorm(nn.Module):
     """
@@ -26,14 +28,19 @@ class MLPSpeculatorLayerNorm(nn.Module):
         Safety term to prevent division by zero. Make sure the chosen value
          fits in the range of your encoding scheme
          (i.e. fp16 requires eps >= 6e-8).
+    elementwise_scale_and_shift : bool
+        Include a learned scaling and shift term after normalization.
     """
 
     def __init__(
         self,
         normalized_shape,
         eps=1e-06,
+        elementwise_scale_and_shift=True,
     ):
         super(MLPSpeculatorLayerNorm, self).__init__()
+        self.elementwise_scale_and_shift = elementwise_scale_and_shift
+        if self.elementwise_scale_and_shift:
             self.weight = nn.Parameter(torch.empty(normalized_shape))
             self.bias = nn.Parameter(torch.empty(normalized_shape))
         self.eps = eps
@@ -42,6 +49,7 @@ class MLPSpeculatorLayerNorm(nn.Module):
         xf = x
         xf = xf * torch.rsqrt(xf.pow(2).mean(-1, keepdim=True) + self.eps)
         x = xf.type_as(x)
+        if self.elementwise_scale_and_shift:
             x = self.weight * x
             x = x + self.bias
         return x
@@ -59,6 +67,34 @@ class MLPSpeculator(nn.Module):
 
         self.max_speculative_tokens = config.num_lookahead_tokens
 
+        self.tie_weights = config.tie_weights
+        self.scale_input = config.scale_input
+
+        if self.tie_weights:
+            assert (
+                self.n_predict >
+                1), "You cannot tie weights between stages when only 1 exists"
+            embedding = VocabParallelEmbedding(
+                config.vocab_size,
+                self.inner_dim,
+                org_num_embeddings=config.vocab_size)
+            self.emb = nn.ModuleList([embedding] * self.max_speculative_tokens)
+
+            # the initial projection from the base model may
+            # have a different size, so that stays separate.
+            proj_first = nn.Linear(self.emb_dim, self.inner_dim, bias=False)
+            proj_tied = nn.Linear(self.inner_dim, self.inner_dim, bias=False)
+            self.proj = nn.ModuleList([proj_first] + [proj_tied] *
+                                      (self.max_speculative_tokens - 1))
+
+            head = nn.Linear(self.inner_dim, self.vocab_size, bias=False)
+            self.head = nn.ModuleList([head] * self.max_speculative_tokens)
+
+            ln = MLPSpeculatorLayerNorm(self.inner_dim,
+                                        elementwise_scale_and_shift=True)
+            self.ln = nn.ModuleList([ln] * self.max_speculative_tokens)
+
+        else:
             self.emb = nn.ModuleList([
                 VocabParallelEmbedding(config.vocab_size,
                                        self.inner_dim,
@@ -69,7 +105,8 @@ class MLPSpeculator(nn.Module):
             self.proj = nn.ModuleList([
                 nn.Linear((self.emb_dim if i == 0 else self.inner_dim),
                           self.inner_dim,
-                      bias=False) for i in range(self.max_speculative_tokens)
+                          bias=False)
+                for i in range(self.max_speculative_tokens)
             ])
 
             self.head = nn.ModuleList([
@@ -77,9 +114,13 @@ class MLPSpeculator(nn.Module):
                 for _ in range(self.max_speculative_tokens)
             ])
             self.ln = nn.ModuleList([
-            MLPSpeculatorLayerNorm(self.inner_dim)
+                MLPSpeculatorLayerNorm(self.inner_dim,
+                                       elementwise_scale_and_shift=True)
                 for _ in range(self.max_speculative_tokens)
             ])
+        if self.scale_input:
+            self.ln0 = MLPSpeculatorLayerNorm(
+                self.emb_dim, elementwise_scale_and_shift=False)
 
         self.state_weight = 0.5**(0.5 / config.n_predict)
         self.emb_weight = math.sqrt(
@@ -105,6 +146,9 @@ class MLPSpeculator(nn.Module):
         # b x 1 x d
         previous_hidden_states = previous_hidden_states.unsqueeze(1)
 
+        if self.scale_input:
+            previous_hidden_states = self.ln0(previous_hidden_states) / SQRT2
+
         # b x 1
         last_tokens = input_ids.unsqueeze(1)
 

vllm/transformers_utils/configs/mlp_speculator.py

@@ -17,6 +17,8 @@ class MLPSpeculatorConfig(PretrainedConfig):
                  n_predict: int = 3,
                  top_k_tokens_per_head: Optional[List[int]] = None,
                  n_candidates: int = 5,
+                 tie_weights: bool = False,
+                 scale_input: bool = False,
                  **kwargs):
         """
         Initialize an MLPSpeculatorConfig
@@ -38,6 +40,14 @@ class MLPSpeculatorConfig(PretrainedConfig):
                 NOTE: This parameter is currently unused.
             n_candidates: int
                 number of child candidates to create per sequence
+            tie_weights: bool
+                If true, use a single set of weights for every model
+                head/stage after the first. The initial projection
+                from the base model may have a different size, so that
+                stays separate.
+            scale_input: bool
+                if True, will scale the initial hidden states from
+                the base model.
         """
         if top_k_tokens_per_head is None:
             top_k_tokens_per_head = [5, 4, 3]
@@ -49,5 +59,7 @@ class MLPSpeculatorConfig(PretrainedConfig):
         self.top_k_tokens_per_head = top_k_tokens_per_head
         self.n_candidates = n_candidates
         self.num_lookahead_tokens = n_predict
+        self.tie_weights = tie_weights
+        self.scale_input = scale_input
 
         super().__init__(**kwargs)