fix the head file

flexflow · Feb 9, 2024 · a7ef500 · a7ef500
1 parent 0cdfb1a
commit a7ef500
Show file tree

Hide file tree

Showing 5 changed files with 17 additions and 14 deletions.
diff --git a/deps/legion b/deps/legion
diff --git a/deps/nccl b/deps/nccl
diff --git a/include/flexflow/ops/fused.h b/include/flexflow/ops/fused.h
@@ -6,6 +6,18 @@
 
 namespace FlexFlow {
 
+// declare Legion names
+using Legion::Context;
+using Legion::coord_t;
+using Legion::Domain;
+using Legion::Future;
+using Legion::LogicalPartition;
+using Legion::LogicalRegion;
+using Legion::Memory;
+using Legion::PhysicalRegion;
+using Legion::Runtime;
+using Legion::Task;
+
 class FusedOp;
 class FusedOpMeta {
 public:

diff --git a/llama.sh b/llama.sh
@@ -0,0 +1 @@
+./build/inference/incr_decoding/incr_decoding --ll:util 8 -ll:cpu 4 -ll:fsize 20000 -ll:zsize 30000 -llm-model meta-llama/Llama-2-7b-hf -cache-folder /home/xiaoxias/weights --fusion -ll:gpu 4 -prompt /home/xiaoxias/cudagraph/inference/prompt/chatgpt.json  -tensor-parallelism-degree 4 > llama2.log 2>&1
diff --git a/src/ops/fused.cu b/src/ops/fused.cu
@@ -47,17 +47,6 @@
 #include "flexflow/ffconst_utils.h"
 
 namespace FlexFlow {
-// declare Legion names
-using Legion::Context;
-using Legion::coord_t;
-using Legion::Domain;
-using Legion::Future;
-using Legion::LogicalPartition;
-using Legion::LogicalRegion;
-using Legion::Memory;
-using Legion::PhysicalRegion;
-using Legion::Runtime;
-using Legion::Task;
 
 OpMeta *FusedOp::init_task(Task const *task,
                            std::vector<PhysicalRegion> const &regions,
@@ -1127,7 +1116,7 @@ __host__ void FusedOp::capture_graph(Task const *task,
         output_accessors_to_save.push_back(output_accessor[i + ooff]);
       }
       assert(task->index_point.get_dim() == 1);
-      int shard_id = task->index_point.point_data[0];
+      shard_id = task->index_point.point_data[0];
       FusedOp::save_inference_tensors_to_file(metas->meta[op],
                                               shard_id,
                                               bc,
@@ -1176,6 +1165,7 @@ __host__ void
                       bc->num_active_tokens(),
                       bc->num_generation_tokens > 0);
   int scenario = 0;
+  int shard_id = task->index_point.point_data[0];
   auto it = metas->graph_collections.find(graph_params);
   if(it != metas->graph_collections.end()) {
     instance = it->second;
Original file line number	Diff line number	Diff line change
		@@ -0,0 +1 @@
		./build/inference/incr_decoding/incr_decoding --ll:util 8 -ll:cpu 4 -ll:fsize 20000 -ll:zsize 30000 -llm-model meta-llama/Llama-2-7b-hf -cache-folder /home/xiaoxias/weights --fusion -ll:gpu 4 -prompt /home/xiaoxias/cudagraph/inference/prompt/chatgpt.json -tensor-parallelism-degree 4 > llama2.log 2>&1