Merge branch 'v0.6.2-dev-wangmin' into 'v0.6.2-dev'

Mixtral模型的gemm由TN改为NN

See merge request dcutoolkit/deeplearing/vllm!32

vllm/model_executor/model_loader/utils.py

@@ -23,7 +23,7 @@ def get_model_architecture(
         model_config: ModelConfig) -> Tuple[Type[nn.Module], str]:
     architectures = getattr(model_config.hf_config, "architectures", [])
     visions = getattr(model_config.hf_config, "visual", []) or getattr(model_config.hf_config, "vision_config", [])
-    support_nn_architectures = ['LlamaForCausalLM', 'QWenLMHeadModel', 'Qwen2ForCausalLM', 'Qwen2MoeForCausalLM', 'Qwen2VLForConditionalGeneration', 'ChatGLMModel', 'BaichuanForCausalLM', 'BloomForCausalLM', 'MedusaModel']  
+    support_nn_architectures = ['LlamaForCausalLM', 'QWenLMHeadModel', 'Qwen2ForCausalLM', 'Qwen2MoeForCausalLM', 'Qwen2VLForConditionalGeneration', 'ChatGLMModel', 'BaichuanForCausalLM', 'BloomForCausalLM', 'MedusaModel', 'MixtralForCausalLM']  
     if any(arch in architectures for arch in support_nn_architectures): 
         if os.getenv('LLAMA_NN') != '0': 
              if (architectures == ['QWenLMHeadModel'] or architectures == ['ChatGLMModel'] ) and visions != []:

vllm/model_executor/models/mixtral.py

@@ -21,6 +21,8 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 """Inference-only Mixtral model."""
+import os
+import re
 from typing import Iterable, List, Optional, Tuple
 
 import torch
@@ -46,6 +48,8 @@ from vllm.model_executor.model_loader.weight_utils import (
     default_weight_loader, maybe_remap_kv_scale_name)
 from vllm.model_executor.sampling_metadata import SamplingMetadata
 from vllm.sequence import IntermediateTensors
+from vllm import _custom_ops as ops
+from vllm.model_executor.utils import pad_weight, gemm_bank_conf
 
 from .interfaces import SupportsLoRA
 from .utils import is_pp_missing_parameter, make_layers
@@ -366,6 +370,12 @@ class MixtralForCausalLM(nn.Module, SupportsLoRA):
                                                 config.vocab_size)
         self.sampler = Sampler()
 
+        self.quant_method = None
+        if quant_config is not None:
+            self.quant_method=quant_config.get_name()
+
+        self.use_llama_nn = os.environ.get('LLAMA_NN') == '1'
+
     def forward(
         self,
         input_ids: torch.Tensor,
@@ -483,3 +493,24 @@ class MixtralForCausalLM(nn.Module, SupportsLoRA):
                     weight_loader = getattr(param, "weight_loader",
                                             default_weight_loader)
                     weight_loader(param, loaded_weight)
+
+        if self.use_llama_nn and self.quant_method is None:
+            lay_key_words = [
+                "block_sparse_moe.gate.weight",
+                "self_attn.qkv_proj.weight",
+                "self_attn.o_proj.weight",
+                "lm_head.weight",
+            ]
+            combined_words = "|".join(lay_key_words)
+            
+            for layername, weight in params_dict.items():
+                matches = re.findall(combined_words, layername)
+
+                if matches:
+                    _weight = torch.zeros_like(weight.data)
+                    ori_shape =_weight.shape
+                    
+                    ops.trans_w16_gemm(_weight, weight.data, _weight.shape[0], _weight.shape[1])
+                    weight.data.copy_(_weight)
+                    
+                    weight.data=weight.data.reshape(ori_shape[1],-1)