Isolate src code and testing for DeepSpeed-FastGen (#4610)

Co-authored-by: Jeff Rasley <[email protected]>
Co-authored-by: Michael Wyatt <[email protected]>
Co-authored-by: Ammar Ahmad Awan <[email protected]>
Co-authored-by: Masahiro Tanaka <[email protected]>
Co-authored-by: Logan Adams <[email protected]>

.github/workflows/nv-a6000.yml

@@ -2,9 +2,9 @@ name: nv-a6000
 
 on:
   pull_request:
-    paths-ignore:
-      - 'docs/**'
-      - 'blogs/**'
+    paths:
+      - "deepspeed/inference/v2/**"
+      - "tests/unit/inference/v2/**"
   workflow_dispatch:
 
 concurrency:

.github/workflows/nv-accelerate-v100.yml

@@ -5,6 +5,7 @@ on:
     paths-ignore:
       - 'docs/**'
       - 'blogs/**'
+      - 'deepspeed/inference/v2/**'
   merge_group:
     branches: [ master ]
   schedule:

.github/workflows/nv-inference.yml

@@ -5,6 +5,7 @@ on:
     paths-ignore:
       - 'docs/**'
       - 'blogs/**'
+      - 'deepspeed/inference/v2/**'
   merge_group:
     branches: [ master ]
   schedule:

.github/workflows/nv-lightning-v100.yml

@@ -5,6 +5,7 @@ on:
     paths-ignore:
       - 'docs/**'
       - 'blogs/**'
+      - 'deepspeed/inference/v2/**'
   merge_group:
     branches: [ master ]
   schedule:

.github/workflows/nv-megatron.yml

@@ -5,6 +5,7 @@ on:
     paths-ignore:
       - 'docs/**'
       - 'blogs/**'
+      - 'deepspeed/inference/v2/**'
   merge_group:
     branches: [ master ]
   schedule:

.github/workflows/nv-pre-compile-ops.yml

@@ -7,6 +7,7 @@ on:
     paths-ignore:
       - 'docs/**'
       - 'blogs/**'
+      - 'deepspeed/inference/v2/**'
   merge_group:
     branches: [ master ]
   schedule:

.github/workflows/nv-torch-latest-cpu.yml

@@ -5,6 +5,7 @@ on:
     paths-ignore:
       - 'docs/**'
       - 'blogs/**'
+      - 'deepspeed/inference/v2/**'
   merge_group:
     branches: [ master ]
   schedule:

.github/workflows/nv-torch-latest-v100.yml

@@ -5,6 +5,7 @@ on:
     paths-ignore:
       - 'docs/**'
       - 'blogs/**'
+      - 'deepspeed/inference/v2/**'
   merge_group:
     branches: [ master ]
   schedule:

.github/workflows/nv-transformers-v100.yml

@@ -5,6 +5,7 @@ on:
     paths-ignore:
       - 'docs/**'
       - 'blogs/**'
+      - 'deepspeed/inference/v2/**'
   merge_group:
     branches: [ master ]
   schedule:

deepspeed/inference/v2/kernels/includes/activation_type.h

@@ -0,0 +1,17 @@
+// Copyright (c) Microsoft Corporation.
+// SPDX-License-Identifier: Apache-2.0
+
+// DeepSpeed Team
+
+#pragma once
+
+enum ActivationType {
+    GELU = 0,
+    RELU = 1,
+    SILU = 2,
+    GEGLU = 3,
+    ReGLU = 4,
+    SiGLU = 5,
+    IDENTITY = 6,
+    InvalidType = -1
+};