[Model] Align nemotron config with final HF state and fix lm-eval-small (#7611)

.buildkite/lm-eval-harness/configs/Minitron-4B-Base-FP8.yaml

@@ -1,11 +1,11 @@
-# bash .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh -m nvidia/Minitron-4B-Base -b auto -l 1000 -f 5 -t 1
+# bash .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh -m mgoin/Minitron-4B-Base-FP8 -b auto -l 1000 -f 5 -t 1
-model_name: "nvidia/Minitron-4B-Base"
+model_name: "mgoin/Minitron-4B-Base-FP8"
 tasks:
 - name: "gsm8k"
   metrics:
   - name: "exact_match,strict-match"
-    value: 0.252
+    value: 0.233
   - name: "exact_match,flexible-extract"
-    value: 0.252
+    value: 0.236
 limit: 1000
 num_fewshot: 5

.buildkite/lm-eval-harness/configs/models-small.txt

@@ -4,7 +4,7 @@ Meta-Llama-3-8B-Instruct-FP8-compressed-tensors.yaml
 Meta-Llama-3-8B-Instruct-INT8-compressed-tensors.yaml
 Meta-Llama-3-8B-Instruct-nonuniform-compressed-tensors.yaml
 Meta-Llama-3-8B-Instruct-Channelwise-compressed-tensors.yaml
-Minitron-4B-Base.yaml
+Minitron-4B-Base-FP8.yaml
 Qwen2-1.5B-Instruct-INT8-compressed-tensors.yaml
 Qwen2-1.5B-Instruct-FP8W8.yaml
 Meta-Llama-3-8B-QQQ.yaml

vllm/model_executor/layers/rotary_embedding.py

@@ -774,7 +774,7 @@ def get_rope(
     is_neox_style: bool = True,
     rope_scaling: Optional[Dict[str, Any]] = None,
     dtype: Optional[torch.dtype] = None,
-    rotary_percent: float = 1.0,
+    partial_rotary_factor: float = 1.0,
 ) -> RotaryEmbedding:
     if dtype is None:
         dtype = torch.get_default_dtype()
@@ -787,8 +787,8 @@ def get_rope(
         rope_scaling_args = tuple(rope_scaling_tuple.items())
     else:
         rope_scaling_args = None
-    if rotary_percent < 1.0:
+    if partial_rotary_factor < 1.0:
-        rotary_dim = int(rotary_dim * rotary_percent)
+        rotary_dim = int(rotary_dim * partial_rotary_factor)
     key = (head_size, rotary_dim, max_position, base, is_neox_style,
            rope_scaling_args, dtype)
     if key in _ROPE_DICT:

vllm/model_executor/models/nemotron.py

@@ -53,7 +53,7 @@ from .utils import PPMissingLayer, is_pp_missing_parameter, make_layers
 # - There is no gate_proj, just up_proj
 # - Normal LayerNorm (with a +1 to the weights) instead of RMSNorm
 # - Squared ReLU instead of SwiGLU
-# - Adds a rotary_percent to RoPE
+# - Adds a partial_rotary_factor to RoPE
 
 
 def _cast_if_autocast_enabled(*args):
@@ -161,7 +161,7 @@ class NemotronAttention(nn.Module):
         self.kv_size = self.num_kv_heads * self.head_dim
         self.scaling = self.head_dim**-0.5
         self.rope_theta = rope_theta
-        self.rotary_percent = config.rope_percent
+        self.partial_rotary_factor = config.partial_rotary_factor
         self.max_position_embeddings = max_position_embeddings
 
         self.qkv_proj = QKVParallelLinear(
@@ -187,7 +187,7 @@ class NemotronAttention(nn.Module):
             max_position=max_position_embeddings,
             base=rope_theta,
             rope_scaling=rope_scaling,
-            rotary_percent=self.rotary_percent,
+            partial_rotary_factor=self.partial_rotary_factor,
         )
         self.attn = Attention(self.num_heads,
                               self.head_dim,

vllm/transformers_utils/configs/nemotron.py

@@ -35,20 +35,20 @@ class NemotronConfig(PretrainedConfig):
 
 
     Args:
-        vocab_size (`int`, *optional*, defaults to 32000):
+        vocab_size (`int`, *optional*, defaults to 256000):
             Vocabulary size of the Nemotron model. Defines the number of
             different tokens that can be represented by the
             `inputs_ids` passed when calling [`NemotronModel`]
-        hidden_size (`int`, *optional*, defaults to 4096):
+        hidden_size (`int`, *optional*, defaults to 6144):
             Dimension of the hidden representations.
-        intermediate_size (`int`, *optional*, defaults to 11008):
+        intermediate_size (`int`, *optional*, defaults to 24576):
             Dimension of the MLP representations.
         num_hidden_layers (`int`, *optional*, defaults to 32):
             Number of hidden layers in the Transformer decoder.
-        num_attention_heads (`int`, *optional*, defaults to 32):
+        num_attention_heads (`int`, *optional*, defaults to 48):
             Number of attention heads for each attention layer in the
             Transformer decoder.
-        head_dim (`int`, *optional*, defaults to None):
+        head_dim (`int`, *optional*):
             Projection weights dimension in multi-head attention. Set to
             hidden_size // num_attention_heads if None
         num_key_value_heads (`int`, *optional*):
@@ -63,16 +63,16 @@ class NemotronConfig(PretrainedConfig):
             heads within that group. For more details checkout 
             [this paper](https://arxiv.org/pdf/2305.13245.pdf). If it
             is not specified, will default to `num_attention_heads`.
-        hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
+        hidden_act (`str` or `function`, *optional*, defaults to `"relu2"`):
             The non-linear activation function (function or string) in the
             decoder.
-        max_position_embeddings (`int`, *optional*, defaults to 2048):
+        max_position_embeddings (`int`, *optional*, defaults to 4096):
             The maximum sequence length that this model might ever be used
             with.
-        initializer_range (`float`, *optional*, defaults to 0.02):
+        initializer_range (`float`, *optional*, defaults to 0.0134):
             The standard deviation of the truncated_normal_initializer for
             initializing all weight matrices.
-        norm_eps (`float`, *optional*, defaults to 1e-06):
+        norm_eps (`float`, *optional*, defaults to 1e-05):
             The epsilon used by the normalization layers.
         use_cache (`bool`, *optional*, defaults to `True`):
             Whether or not the model should return the last key/values
@@ -80,21 +80,16 @@ class NemotronConfig(PretrainedConfig):
             `config.is_decoder=True`.
         pad_token_id (`int`, *optional*):
             Padding token id.
-        bos_token_id (`int`, *optional*, defaults to 1):
+        bos_token_id (`int`, *optional*, defaults to 2):
             Beginning of stream token id.
-        eos_token_id (`int`, *optional*, defaults to 2):
+        eos_token_id (`int`, *optional*, defaults to 3):
             End of stream token id.
         tie_word_embeddings (`bool`, *optional*, defaults to `False`):
             Whether to tie weight embeddings
         rope_theta (`float`, *optional*, defaults to 10000.0):
             The base period of the RoPE embeddings.
-        rope_scaling (`Dict`, *optional*):
+        partial_rotary_factor (`float`, *optional*, defaults to 0.5):
-            Dictionary containing the scaling configuration for the RoPE
+            Percentage of the query and keys which will have rotary embedding.
-            embeddings. Currently supports two scaling strategies: linear
-            and dynamic. Their scaling factor must be a float greater than 1.
-            The expected format is `{"type": strategy name,
-            "factor": scaling factor}`. When using this flag, don't update
-            `max_position_embeddings` to the expected new maximum.
         attention_bias (`bool`, *optional*, defaults to `False`):
             Whether to use a bias in the query, key, value and output
             projection layers during self-attention.
@@ -106,13 +101,10 @@ class NemotronConfig(PretrainedConfig):
 
     ```python
     >>> from transformers import NemotronModel, NemotronConfig
-
     >>> # Initializing a Nemotron nemotron-15b style configuration
     >>> configuration = NemotronConfig()
-
     >>> # Initializing a model from the nemotron-15b style configuration
     >>> model = NemotronModel(configuration)
-
     >>> # Accessing the model configuration
     >>> configuration = model.config
     ```"""
@@ -140,7 +132,7 @@ class NemotronConfig(PretrainedConfig):
         tie_word_embeddings=False,
         rope_theta=10000.0,
         rope_scaling=None,
-        rope_percent=0.5,
+        partial_rotary_factor=0.5,
         attention_bias=False,
         attention_dropout=0.0,
         mlp_bias=False,
@@ -167,8 +159,10 @@ class NemotronConfig(PretrainedConfig):
         self.use_cache = use_cache
         self.rope_theta = rope_theta
         self.rope_scaling = rope_scaling
-        rope_percent = rope_percent or kwargs.get("rope_percentage", None)
+        # for backward compatibility
-        self.rope_percent = rope_percent
+        partial_rotary_factor = kwargs.get("rope_percent", None) or kwargs.get(
+            "rope_percentage", None) or partial_rotary_factor
+        self.partial_rotary_factor = partial_rotary_factor
         self._rope_scaling_validation()
         self.attention_bias = attention_bias
         self.attention_dropout = attention_dropout