[Core] Refactor GGUF parameters packing and forwarding (vllm-project#8859)

Isotr0py · LeiWang1999 · commit 9d9ced1f7617 · 2025-03-26T10:14:44.000Z
Signed-off-by: LeiWang1999 &lt;leiwang1999@outlook.com&gt;
diff --git a/tests/models/decoder_only/language/test_gguf.py b/tests/models/decoder_only/language/test_gguf.py
@@ -19,12 +19,12 @@
 
 # FIXME: Move this to confest
 MODELS = [
-    ("TinyLlama/TinyLlama-1.1B-Chat-v1.0",
-     hf_hub_download("TheBloke/TinyLlama-1.1B-Chat-v1.0-GGUF",
-                     filename="tinyllama-1.1b-chat-v1.0.Q4_K_M.gguf")),
-    ("TinyLlama/TinyLlama-1.1B-Chat-v1.0",
-     hf_hub_download("duyntnet/TinyLlama-1.1B-Chat-v1.0-imatrix-GGUF",
-                     filename="TinyLlama-1.1B-Chat-v1.0-IQ4_XS.gguf")),
+    ("meta-llama/Llama-3.2-1B-Instruct",
+     hf_hub_download("bartowski/Llama-3.2-1B-Instruct-GGUF",
+                     filename="Llama-3.2-1B-Instruct-Q4_K_M.gguf")),
+    ("meta-llama/Llama-3.2-1B-Instruct",
+     hf_hub_download("bartowski/Llama-3.2-1B-Instruct-GGUF",
+                     filename="Llama-3.2-1B-Instruct-IQ4_XS.gguf")),
     ("Qwen/Qwen2-1.5B-Instruct",
      hf_hub_download("Qwen/Qwen2-1.5B-Instruct-GGUF",
                      filename="qwen2-1_5b-instruct-q4_k_m.gguf")),
diff --git a/vllm/model_executor/layers/linear.py b/vllm/model_executor/layers/linear.py
@@ -444,17 +444,23 @@ def weight_loader(self,
             param.shard_weight_type[loaded_shard_id] = loaded_weight.item()
             return
 
-        if is_gguf_weight and isinstance(param, UninitializedParameter):
-            from gguf.constants import GGML_QUANT_SIZES
+        if is_gguf_weight:
+            tp_size = get_tensor_model_parallel_world_size()
+            tp_rank = get_tensor_model_parallel_rank()
+
+            output_dim = getattr(param, "output_dim", None)
+            shard_size = loaded_weight.size(output_dim) // tp_size
+            start_idx = tp_rank * shard_size
+
+            loaded_weight = loaded_weight.narrow(output_dim, start_idx,
+                                                 shard_size)
 
-            ori_shape = param.tensor_shape
-            weight_types = self.qweight_type.shard_weight_type.values()
-            row_size = []
-            for weight_type in weight_types:
-                block_size, type_size = GGML_QUANT_SIZES[weight_type]
-                row_size.append(ori_shape[1] // block_size * type_size)
-            q_shape = (ori_shape[0], max(row_size))
-            param.materialize(q_shape, dtype=loaded_weight.dtype)
+            param.shard_id.append(loaded_shard_id)
+            param.shard_id_map[loaded_shard_id] = len(param.data_container)
+            param.data_container.append(loaded_weight)
+            if len(param.data_container) == 2:
+                self.qweight = param.materialize_nested()
+            return
 
         param_data = param.data
         output_dim = getattr(param, "output_dim", None)
@@ -522,18 +528,6 @@ def weight_loader(self,
                 shard_offset = loaded_weight.shape[output_dim] * \
                     loaded_shard_id
 
-            if is_gguf_weight:
-                tp_size = get_tensor_model_parallel_world_size()
-                output_dim = getattr(param, "output_dim", None)
-                shard_shape = list(loaded_weight.shape)
-                shard_shape[output_dim] = shard_shape[output_dim] // tp_size
-                param.shard_id.append(loaded_shard_id)
-                param.shard_size[loaded_shard_id] = shard_shape
-
-                input_dim = getattr(param, "input_dim", None)
-                input_size = loaded_weight.shape[input_dim]
-                param_data = param_data.narrow(input_dim, 0, input_size)
-
             param_data = param_data.narrow(output_dim, shard_offset,
                                            shard_size)
             start_idx = tp_rank * shard_size
@@ -790,17 +784,23 @@ def weight_loader(self,
             param.shard_weight_type[loaded_shard_id] = loaded_weight.item()
             return
 
-        if is_gguf_weight and isinstance(param, UninitializedParameter):
-            from gguf.constants import GGML_QUANT_SIZES
+        if is_gguf_weight:
+            tp_size = get_tensor_model_parallel_world_size()
+            tp_rank = get_tensor_model_parallel_rank()
 
-            ori_shape = param.tensor_shape
-            weight_types = self.qweight_type.shard_weight_type.values()
-            row_size = []
-            for weight_type in weight_types:
-                block_size, type_size = GGML_QUANT_SIZES[weight_type]
-                row_size.append(ori_shape[1] // block_size * type_size)
-            q_shape = (ori_shape[0], max(row_size))
-            param.materialize(q_shape, dtype=loaded_weight.dtype)
+            output_dim = getattr(param, "output_dim", None)
+            shard_size = loaded_weight.size(output_dim) // tp_size
+            start_idx = tp_rank * shard_size
+
+            loaded_weight = loaded_weight.narrow(output_dim, start_idx,
+                                                 shard_size)
+
+            param.shard_id.append(loaded_shard_id)
+            param.shard_id_map[loaded_shard_id] = len(param.data_container)
+            param.data_container.append(loaded_weight)
+            if len(param.data_container) == 3:
+                self.qweight = param.materialize_nested()
+            return
 
         param_data = param.data
         output_dim = getattr(param, "output_dim", None)
@@ -891,18 +891,6 @@ def weight_loader(self,
                 shard_size, shard_offset = adjust_bitsandbytes_4bit_shard(
                     param, orig_qkv_offsets, loaded_shard_id)
 
-            if is_gguf_weight:
-                tp_size = get_tensor_model_parallel_world_size()
-                output_dim = getattr(param, "output_dim", None)
-                shard_shape = list(loaded_weight.shape)
-                shard_shape[output_dim] = shard_shape[output_dim] // tp_size
-                param.shard_id.append(loaded_shard_id)
-                param.shard_size[loaded_shard_id] = shard_shape
-
-                input_dim = getattr(param, "input_dim", None)
-                input_size = loaded_weight.shape[input_dim]
-                param_data = param_data.narrow(input_dim, 0, input_size)
-
             param_data = param_data.narrow(output_dim, shard_offset,
                                            shard_size)
             if loaded_shard_id == "q":
diff --git a/vllm/model_executor/layers/quantization/gguf.py b/vllm/model_executor/layers/quantization/gguf.py
@@ -86,15 +86,16 @@ def create_weights(self, layer: torch.nn.Module,
         output_size_per_partition = sum(output_partition_sizes)
 
         tensor_shape = (output_size_per_partition, input_size_per_partition)
-        qweight = UninitializedParameter(requires_grad=False)
+        qweight = GGUFUninitializedParameter(requires_grad=False)
         set_weight_attrs(
             qweight, {
                 "input_dim": 1,
                 "output_dim": 0,
                 "tensor_shape": tensor_shape,
                 "is_gguf_weight": True,
-                "shard_size": {},
+                "data_container": [],
                 "shard_id": [],
+                "shard_id_map": {},
             })
         set_weight_attrs(qweight, extra_weight_attrs)
         layer.register_parameter("qweight", qweight)
@@ -116,21 +117,17 @@ def apply(self,
               layer: torch.nn.Module,
               x: torch.Tensor,
               bias: Optional[torch.Tensor] = None) -> torch.Tensor:
-        shard_size = getattr(layer.qweight, "shard_size", None)
         shard_id = getattr(layer.qweight, "shard_id", None)
 
-        if shard_id and shard_size:
-            result = []
-            offset = 0
+        if shard_id:
             # dequantize shard weights respectively
             shard_id = ["q", "k", "v"] if "q" in shard_id else shard_id
+            qweight = layer.qweight.unbind(0)
+            result = []
             for id in shard_id:
-                shard_weight = layer.qweight[
-                    offset:offset +
-                    shard_size[id][0], :shard_size[id][1]].contiguous()
+                q_idx = layer.qweight.shard_id_map[id]
                 qweight_type = layer.qweight_type.shard_weight_type[id]
-                result.append(_fuse_mul_mat(x, shard_weight, qweight_type))
-                offset += shard_size[id][0]
+                result.append(_fuse_mul_mat(x, qweight[q_idx], qweight_type))
             out = torch.cat(result, axis=1)
         else:
             qweight = layer.qweight
@@ -162,3 +159,20 @@ def embedding(self, layer: torch.nn.Module,
         dequant = ops.ggml_dequantize(quant, qweight_type, hidden_size,
                                       x_flat.shape[0])
         return dequant.view(*x.shape, hidden_size)
+
+
+class GGUFUninitializedParameter(UninitializedParameter):
+    cls_to_become = Parameter
+    data_container: List[torch.Tensor]
+
+    def materialize_nested(self) -> Parameter:
+        nested_data = torch.nested.nested_tensor(self.data_container,
+                                                 device=self.device,
+                                                 dtype=torch.uint8)
+        self.data_container.clear()
+        param = torch.Tensor._make_subclass(self.cls_to_become,
+                                            nested_data,
+                                            require_grad=False)
+        for k, v in self.__dict__.items():
+            setattr(param, k, v)
+        return param
diff --git a/vllm/model_executor/models/llama.py b/vllm/model_executor/models/llama.py
@@ -512,7 +512,7 @@ def __init__(
                 quant_config=quant_config,
             )
             if config.tie_word_embeddings:
-                self.lm_head.weight = self.model.embed_tokens.weight
+                self.lm_head = self.model.embed_tokens
 
             logit_scale = getattr(config, "logit_scale", 1.0)
             self.logits_processor = LogitsProcessor(self.unpadded_vocab_size,

Original file line number	Diff line number	Diff line change
`@@ -512,7 +512,7 @@ def __init__(`
`512`	`512`	`quant_config=quant_config,`
`513`	`513`	`)`
`514`	`514`	`if config.tie_word_embeddings:`
`515`		`- self.lm_head.weight = self.model.embed_tokens.weight`
	`515`	`+ self.lm_head = self.model.embed_tokens`
`516`	`516`
`517`	`517`	`logit_scale = getattr(config, "logit_scale", 1.0)`
`518`	`518`	`self.logits_processor = LogitsProcessor(self.unpadded_vocab_size,`