Merge branch 'v0.5.0-dtk24.04.1'

7462218e · zhuwenwen · 6ccd3f47 · 1cec5e62 · 7462218e · 7462218e
Commit 7462218e authored Sep 05, 2024 by zhuwenwen
20 changed files
--- a/vllm/model_executor/layers/activation.py
+++ b/vllm/model_executor/layers/activation.py
--- a/vllm/model_executor/layers/fused_moe/__init__.py
+++ b/vllm/model_executor/layers/fused_moe/__init__.py
--- a/vllm/model_executor/layers/fused_moe/fused_moe.py
+++ b/vllm/model_executor/layers/fused_moe/fused_moe.py
--- a/vllm/model_executor/layers/layernorm.py
+++ b/vllm/model_executor/layers/layernorm.py
--- a/vllm/model_executor/layers/linear.py
+++ b/vllm/model_executor/layers/linear.py
--- a/vllm/model_executor/layers/ops/rand.py
+++ b/vllm/model_executor/layers/ops/rand.py
--- a/vllm/model_executor/layers/ops/sample.py
+++ b/vllm/model_executor/layers/ops/sample.py
--- a/vllm/model_executor/layers/quantization/awq.py
+++ b/vllm/model_executor/layers/quantization/awq.py
--- a/vllm/model_executor/layers/rotary_embedding.py
+++ b/vllm/model_executor/layers/rotary_embedding.py
--- a/vllm/model_executor/model_loader/loader.py
+++ b/vllm/model_executor/model_loader/loader.py
@@ -274,7 +274,7 @@ class DefaultModelLoader(BaseModelLoader):

            for _, module in model.named_modules():
                quant_method = getattr(module, "quant_method", None)
-                if quant_method is not None:
+                if quant_method is not None and quant_method!="awq" and quant_method!="gptq":
                    quant_method.process_weights_after_loading(module)
                # FIXME: Remove this after Mixtral is updated
                # to use quant_method.

--- a/vllm/model_executor/model_loader/utils.py
+++ b/vllm/model_executor/model_loader/utils.py
--- a/vllm/model_executor/models/__init__.py
+++ b/vllm/model_executor/models/__init__.py
--- a/vllm/model_executor/models/baichuan.py
+++ b/vllm/model_executor/models/baichuan.py
--- a/vllm/model_executor/models/chatglm.py
+++ b/vllm/model_executor/models/chatglm.py
--- a/vllm/model_executor/models/deepseek_v2.py
+++ b/vllm/model_executor/models/deepseek_v2.py
--- a/vllm/model_executor/models/llama.py
+++ b/vllm/model_executor/models/llama.py
--- a/vllm/model_executor/models/qwen.py
+++ b/vllm/model_executor/models/qwen.py
--- a/vllm/model_executor/models/qwen2.py
+++ b/vllm/model_executor/models/qwen2.py
--- a/vllm/model_executor/utils.py
+++ b/vllm/model_executor/utils.py
--- a/vllm/worker/model_runner.py
+++ b/vllm/worker/model_runner.py