[Model][Quant] Fix GLM, Fix fused module mappings for quantization

vllm/model_executor/layers/quantization/base_config.py

@@ -1,6 +1,6 @@
 import inspect
 from abc import ABC, abstractmethod
-from typing import Any, Dict, List, Optional, Type
+from typing import Any, Dict, List, Mapping, Optional, Type
 
 import torch
 from torch import nn
@@ -57,6 +57,7 @@ def method_has_implemented_embedding(
 
 class QuantizationConfig(ABC):
     """Base class for quantization configs."""
+    packed_modules_mapping: Mapping[str, List[str]] = dict()
 
     @abstractmethod
     def get_name(self) -> str:

vllm/model_executor/layers/quantization/compressed_tensors/compressed_tensors.py

@@ -78,7 +78,10 @@ def get_quant_method(
 
         # Check if the layer is skipped for quantization.
         # TODO (@robertgshaw2): support module names
-        if should_ignore_layer(prefix, ignore=self.ignore):
+        if should_ignore_layer(
+                prefix,
+                ignore=self.ignore,
+                packed_modules_mapping=self.packed_modules_mapping):
             return UnquantizedLinearMethod()
         if isinstance(layer, LinearBase):
             scheme = self.get_scheme(layer=layer, layer_name=prefix)

vllm/model_executor/layers/quantization/compressed_tensors/utils.py

@@ -1,12 +1,10 @@
 import re
-from typing import Iterable, Optional
+from types import MappingProxyType
+from typing import Iterable, List, Mapping, Optional
 
 from compressed_tensors import CompressionFormat
 from torch.nn import Module
 
-from vllm.model_executor.layers.quantization.utils.quant_utils import (
-    FUSED_LAYER_NAME_MAPPING)
-
 
 def is_activation_quantization_format(format: str) -> bool:
     _ACTIVATION_QUANTIZATION_FORMATS = [
@@ -17,8 +15,11 @@ def is_activation_quantization_format(format: str) -> bool:
     return format in _ACTIVATION_QUANTIZATION_FORMATS
 
 
-def should_ignore_layer(layer_name: Optional[str],
-                        ignore: Iterable[str]) -> bool:
+def should_ignore_layer(
+    layer_name: Optional[str],
+    ignore: Iterable[str] = tuple(),
+    packed_modules_mapping: Mapping[str, List[str]] = MappingProxyType({})
+) -> bool:
     if layer_name is None:
         return False
 
@@ -30,8 +31,8 @@ def should_ignore_layer(layer_name: Optional[str],
     # in the safetensors checkpoint. So, we convert the name
     # from the fused version to unfused + check to make sure that
     # each shard of the fused layer has the same scheme.
-    if proj_name in FUSED_LAYER_NAME_MAPPING and layer_name not in ignore:
-        shard_proj_names = FUSED_LAYER_NAME_MAPPING[proj_name]
+    if proj_name in packed_modules_mapping and layer_name not in ignore:
+        shard_proj_names = packed_modules_mapping[proj_name]
 
         # Convert fused_name --> [shard_names]
         shard_names = [

vllm/model_executor/layers/quantization/quark/quark.py

@@ -16,8 +16,6 @@
     QuarkScheme, QuarkW8A8Fp8, QuarkW8A8Int8)
 from vllm.model_executor.layers.quantization.quark.utils import (
     deep_compare, should_ignore_layer)
-from vllm.model_executor.layers.quantization.utils.quant_utils import (
-    FUSED_LAYER_NAME_MAPPING)
 from vllm.platforms import current_platform
 
 __all__ = ["QuarkLinearMethod"]
@@ -56,7 +54,10 @@ def get_quant_method(self, layer: torch.nn.Module,
 
         # Check if the layer is skipped for quantization.
         exclude_layers = cast(List[str], self.quant_config.get("exclude"))
-        if should_ignore_layer(prefix, ignore=exclude_layers):
+        if should_ignore_layer(
+                prefix,
+                ignore=exclude_layers,
+                packed_modules_mapping=self.packed_modules_mapping):
             return UnquantizedLinearMethod()
         if isinstance(layer, LinearBase):
             scheme = self.get_scheme(layer=layer, layer_name=prefix)
@@ -199,8 +200,8 @@ def _find_matched_config(self, layer_name: str,
                              module: torch.nn.Module) -> Dict[str, Any]:
 
         proj_name = layer_name.split(".")[-1]
-        if proj_name in FUSED_LAYER_NAME_MAPPING:
-            shard_proj_names = FUSED_LAYER_NAME_MAPPING[proj_name]
+        if proj_name in self.packed_modules_mapping:
+            shard_proj_names = self.packed_modules_mapping[proj_name]
 
             # Convert fused_name --> [shard_names]
             shard_names = [

vllm/model_executor/layers/quantization/quark/utils.py

@@ -1,8 +1,6 @@
 import re
-from typing import Any, Iterable, Optional
-
-from vllm.model_executor.layers.quantization.utils.quant_utils import (
-    FUSED_LAYER_NAME_MAPPING)
+from types import MappingProxyType
+from typing import Any, Iterable, List, Mapping, Optional
 
 
 def deep_compare(dict1: Any, dict2: Any) -> bool:
@@ -18,8 +16,11 @@ def deep_compare(dict1: Any, dict2: Any) -> bool:
         return dict1 == dict2
 
 
-def should_ignore_layer(layer_name: Optional[str],
-                        ignore: Iterable[str]) -> bool:
+def should_ignore_layer(
+    layer_name: Optional[str],
+    ignore: Iterable[str],
+    packed_modules_mapping: Mapping[str, List[str]] = MappingProxyType({})
+) -> bool:
     if layer_name is None:
         return False
 
@@ -31,8 +32,8 @@ def should_ignore_layer(layer_name: Optional[str],
     # in the safetensors checkpoint. So, we convert the name
     # from the fused version to unfused + check to make sure that
     # each shard of the fused layer has the same scheme.
-    if proj_name in FUSED_LAYER_NAME_MAPPING:
-        shard_proj_names = FUSED_LAYER_NAME_MAPPING[proj_name]
+    if proj_name in packed_modules_mapping:
+        shard_proj_names = packed_modules_mapping[proj_name]
 
         # Convert fused_name --> [shard_names]
         shard_names = [

vllm/model_executor/layers/quantization/utils/quant_utils.py

@@ -1,5 +1,6 @@
 """This file is used for /tests and /benchmarks"""
-from typing import List, Optional
+from types import MappingProxyType
+from typing import List, Mapping, Optional
 
 import numpy
 import torch
@@ -11,14 +12,6 @@
 SUPPORTED_GPTQ_QUANT_TYPES = [scalar_types.uint4b8, scalar_types.uint8b128]
 SUPPORTED_GROUP_SIZES = [-1, 32, 64, 128]
 
-# Note: this is a hack. We should update each model to register the
-# stacked params and get it from there instead in a future PR.
-# fused_name: List[shard_name]
-FUSED_LAYER_NAME_MAPPING = {
-    "qkv_proj": ["q_proj", "k_proj", "v_proj"],
-    "gate_up_proj": ["gate_proj", "up_proj"]
-}
-
 
 def pack_quantized_values_into_int32(w_q: torch.Tensor,
                                      wtype: ScalarType,
@@ -63,14 +56,18 @@ def unpack_quantized_values_into_int32(w_q: torch.Tensor,
     return res.permute(inv_perm)
 
 
-def is_layer_skipped(prefix: str, ignored_layers: List[str]) -> bool:
+def is_layer_skipped(
+    prefix: str,
+    ignored_layers: List[str],
+    packed_modules_mapping: Mapping[str, List[str]] = MappingProxyType({})
+) -> bool:
     # prefix: model.layers.0.self_attn.q_proj
     # proj_name: q_proj
     proj_name = prefix.split(".")[-1]
-    if proj_name in FUSED_LAYER_NAME_MAPPING:
+    if proj_name in packed_modules_mapping:
         shard_prefixes = [
             prefix.replace(proj_name, shard_proj_name)
-            for shard_proj_name in FUSED_LAYER_NAME_MAPPING[proj_name]
+            for shard_proj_name in packed_modules_mapping[proj_name]
         ]
 
         is_skipped = None

vllm/model_executor/model_loader/loader.py

@@ -110,6 +110,11 @@ def _initialize_model(
     model_config = vllm_config.model_config
     model_class, _ = get_model_architecture(model_config)
 
+    # pass packed_modules_mapping by reference to quant_config
+    packed_mapping = getattr(model_class, "packed_modules_mapping", None)
+    if packed_mapping is not None and vllm_config.quant_config is not None:
+        vllm_config.quant_config.packed_modules_mapping = packed_mapping
+
     signatures = inspect.signature(model_class.__init__)
     all_params = [param.name for param in signatures.parameters.values()]
     if "vllm_config" in all_params and "prefix" in all_params:

vllm/model_executor/models/chatglm.py

@@ -263,12 +263,14 @@ def __init__(
             self.total_num_kv_heads,
             bias=config.add_bias_linear or config.add_qkv_bias,
             quant_config=quant_config,
+            prefix=f"{prefix}.query_key_value",
         )
         self.dense = RowParallelLinear(
             self.total_num_heads * self.head_dim,
             config.hidden_size,
             bias=config.add_bias_linear,
             quant_config=quant_config,
+            prefix=f"{prefix}.dense",
         )
 
         # https://huggingface.co/THUDM/chatglm3-6b-32k/blob/e210410255278dd9d74463cf396ba559c0ef801c/modeling_chatglm.py#L141
@@ -325,6 +327,7 @@ def __init__(
         self,
         config: ChatGLMConfig,
         quant_config: Optional[QuantizationConfig] = None,
+        prefix: str = "",
     ):
         super().__init__()
 
@@ -336,6 +339,7 @@ def __init__(
             [config.ffn_hidden_size] * 2,
             bias=config.add_bias_linear,
             quant_config=quant_config,
+            prefix=f"{prefix}.dense_h_to_4h",
         )
 
         self.activation_func = SiluAndMul()
@@ -346,6 +350,7 @@ def __init__(
             config.hidden_size,
             bias=config.add_bias_linear,
             quant_config=quant_config,
+            prefix=f"{prefix}.dense_4h_to_h",
         )
 
     def forward(self, hidden_states):
@@ -394,7 +399,7 @@ def __init__(
             config.hidden_size, eps=config.layernorm_epsilon)
 
         # MLP
-        self.mlp = GLMMLP(config, quant_config)
+        self.mlp = GLMMLP(config, quant_config, prefix=f"{prefix}.mlp")
 
     def forward(
         self,
@@ -505,7 +510,8 @@ def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
 
         self.embedding = VocabParallelEmbedding(config.padded_vocab_size,
                                                 config.hidden_size,
-                                                quant_config=quant_config)
+                                                quant_config=quant_config,
+                                                prefix=f"{prefix}.embedding")
 
         self.num_layers = config.num_layers
         self.multi_query_group_num = config.multi_query_group_num
@@ -764,6 +770,7 @@ class ChatGLMForCausalLM(ChatGLMBaseModel, SupportsLoRA, SupportsPP,
                          SupportsMultiModal):
     # Ensure that the LoRA support check passes when the class is not
     # initialized, but set all these attributes to empty.
+    # These will be updated when an instance class is selected
     packed_modules_mapping = {}
     supported_lora_modules = []
     embedding_modules = {}
@@ -775,9 +782,16 @@ def __new__(
         prefix: str = "",
     ) -> None:
         config = vllm_config.model_config.hf_config
+
         # Initialize VL
-        if hasattr(config, "vision_config"):
-            return ChatGLMV(vllm_config=vllm_config, prefix=prefix)
+        if hasattr(config, "vision_config"):  # noqa: SIM108
+            instance_cls = ChatGLMV
         # Initialize LLM
         else:
-            return ChatGLM(vllm_config=vllm_config, prefix=prefix)
+            instance_cls = ChatGLM
+
+        cls.packed_modules_mapping.update(instance_cls.packed_modules_mapping)
+        cls.supported_lora_modules += instance_cls.supported_lora_modules
+        cls.embedding_modules.update(instance_cls.embedding_modules)
+        cls.embedding_padding_modules += instance_cls.embedding_padding_modules
+        return instance_cls(vllm_config=vllm_config, prefix=prefix)

vllm/model_executor/models/glm4_vision_encoder.py

@@ -72,11 +72,13 @@ def __init__(
             self.head_dim,
             config.num_heads,
             quant_config=quant_config,
+            prefix=f"{prefix}.query_key_value",
         )
         self.dense = RowParallelLinear(
             config.hidden_size,
             config.hidden_size,
             quant_config=quant_config,
+            prefix=f"{prefix}.dense",
         )
 
         self.attn = MultiHeadAttention(self.num_heads_per_rank, self.head_dim,
@@ -99,6 +101,7 @@ def __init__(
         self,
         config,
         quant_config: Optional[QuantizationConfig] = None,
+        prefix: str = '',
     ):
         super().__init__()
         self.config = config
@@ -107,11 +110,13 @@ def __init__(
             config.hidden_size,
             config.intermediate_size,
             quant_config=quant_config,
+            prefix=f"{prefix}.fc1",
         )
         self.fc2 = RowParallelLinear(
             config.intermediate_size,
             config.hidden_size,
             quant_config=quant_config,
+            prefix=f"{prefix}.fc2",
         )
 
     def forward(self, x: torch.Tensor) -> torch.Tensor:
@@ -135,7 +140,9 @@ def __init__(
         self.attention = Attention(config,
                                    quant_config=quant_config,
                                    prefix=f"{prefix}.attention")
-        self.mlp = MLP(config, quant_config=quant_config)
+        self.mlp = MLP(config,
+                       quant_config=quant_config,
+                       prefix=f"{prefix}.mlp")
         self.post_attention_layernorm = LayerNorm(config.hidden_size,
                                                   eps=config.layer_norm_eps)
 
@@ -162,7 +169,7 @@ def __init__(
         self.layers = nn.ModuleList([
             TransformerLayer(config,
                              quant_config=quant_config,
-                             prefix=f"{prefix}.layer.{layer_idx}")
+                             prefix=f"{prefix}.layers.{layer_idx}")
             for layer_idx in range(config.num_hidden_layers)
         ])
 
@@ -179,6 +186,7 @@ def __init__(
         config,
         in_features,
         quant_config: Optional[QuantizationConfig] = None,
+        prefix: str = '',
     ):
         """
         The original implementation is the same as:
@@ -220,20 +228,24 @@ def __init__(
         self.linear_proj = ReplicatedLinear(in_features,
                                             config.hidden_size,
                                             bias=False,
-                                            quant_config=quant_config)
+                                            quant_config=quant_config,
+                                            prefix=f"{prefix}.linear_proj")
         self.norm1 = nn.LayerNorm(config.hidden_size)
         self.act1 = nn.GELU()
         self.act2 = SiluAndMul()
 
         self.merged_proj = MergedColumnParallelLinear(
             config.hidden_size, [config.ffn_hidden_size] * 2,
             bias=False,
-            quant_config=quant_config)
+            quant_config=quant_config,
+            prefix=f"{prefix}.merged_proj")
 
-        self.dense_4h_to_h = RowParallelLinear(config.ffn_hidden_size,
-                                               config.hidden_size,
-                                               bias=False,
-                                               quant_config=quant_config)
+        self.dense_4h_to_h = RowParallelLinear(
+            config.ffn_hidden_size,
+            config.hidden_size,
+            bias=False,
+            quant_config=quant_config,
+            prefix=f"{prefix}.dense_4h_to_h")
 
     def forward(self, x):
         x, _ = self.linear_proj(x)
@@ -260,7 +272,8 @@ def __init__(
                                        prefix=f"{prefix}.transformer")
         self.linear_proj = GLU(config,
                                in_features=config.hidden_size,
-                               quant_config=quant_config)
+                               quant_config=quant_config,
+                               prefix=f"{prefix}.linear_proj")
         self.conv = nn.Conv2d(in_channels=vision_config.hidden_size,
                               out_channels=config.hidden_size,
                               kernel_size=2,