Revert previous change, also disable flashattn in the unit tests for …

…now.
liuliu · Oct 18, 2024 · b361b30 · b361b30
1 parent a003506
commit b361b30
Show file tree

Hide file tree

Showing 5 changed files with 4 additions and 29 deletions.
diff --git a/.github/workflows/analyze.yaml b/.github/workflows/analyze.yaml
@@ -12,7 +12,7 @@ jobs:
           fetch-depth: 0
       - name: configure
         run: |
-          cd lib && ./configure --enable-sm80 --disable-openmp && cd ..
+          cd lib && ./configure --disable-openmp && cd ..
       - name: analyze
         run: |
           cd lib && scan-build -o ../_analyze --use-cc clang make -j 64

diff --git a/.github/workflows/coverage.yaml b/.github/workflows/coverage.yaml
@@ -12,7 +12,7 @@ jobs:
           fetch-depth: 0
       - name: configure & build
         run: |
-          cd lib && ./configure --enable-sm80 && cd ..
+          cd lib && ./configure && cd ..
           make -j 64 -C test COVER=1 all.tests
       - name: tests
         run: |

diff --git a/.github/workflows/cuda-int-tests.yaml b/.github/workflows/cuda-int-tests.yaml
@@ -12,7 +12,7 @@ jobs:
           fetch-depth: 0
       - name: configure & build
         run: |
-          cd lib && ./configure --enable-sm80 && cd ..
+          cd lib && ./configure && cd ..
           make -j 64 -C test/int/nnc
       - name: integration tests
         run: |

diff --git a/.github/workflows/ubsan-unit-tests.yaml b/.github/workflows/ubsan-unit-tests.yaml
@@ -12,7 +12,7 @@ jobs:
           fetch-depth: 0
       - name: configure & build
         run: |
-          cd lib && ./configure --enable-sm80 && cd ..
+          cd lib && ./configure && cd ..
           make -j 64 -C lib ubsan
           make -j 64 -C bin ubsan
           make -j 64 -C bin/nnc ubsan

diff --git a/lib/nnc/ccv_cnnp_model_gradient_checkpointing.c b/lib/nnc/ccv_cnnp_model_gradient_checkpointing.c
@@ -201,31 +201,6 @@ void ccv_cnnp_model_apply_gradient_checkpoints(ccv_cnnp_compiled_data_t* const c
 		}
 		if (input_execs->rnum <= 0 || output_execs->rnum <= 0)
 			continue;
-		ccv_nnc_graph_visit_t* const reverse_visit = ccv_nnc_graph_visit_new(graph, reversed_nodes, exec_rnum, (ccv_nnc_graph_exec_symbol_t*)ccv_array_get(output_execs, 0), output_execs->rnum, (ccv_nnc_graph_exec_symbol_t*)ccv_array_get(input_execs, 0), input_execs->rnum, 1);
-		ccv_nnc_graph_visit_for(reverse_visit, exec_info, node, idx) {
-			if (idx < exec_rnum && !CCV_NNC_GRAPH_EXEC_IS_DEAD(node->flags))
-				maskbit[idx >> 5] |= (1u << (idx & 0x1f));
-		} ccv_nnc_graph_visit_endfor
-		// Check if any of the items in input_execs is not marked, if it is not, there is no connection from it to output_execs, no need to visit.
-		for (j = 0; j < input_execs->rnum;)
-		{
-			const int idx = ((ccv_nnc_tensor_symbol_t*)ccv_array_get(input_execs, j))->d;
-			if (maskbit[idx >> 5] & (1u << (idx & 0x1f)))
-			{
-				++j;
-				continue;
-			}
-			// Not marked, remove this one.
-			if (j < input_execs->rnum - 1)
-				*(ccv_nnc_tensor_symbol_t*)ccv_array_get(input_execs, j) = *(ccv_nnc_tensor_symbol_t*)ccv_array_get(input_execs, input_execs->rnum - 1);
-			--input_execs->rnum;
-		}
-		// Reset maskbit back.
-		ccv_nnc_graph_visit_for(reverse_visit, exec_info, node, idx) {
-			if (idx < exec_rnum && !CCV_NNC_GRAPH_EXEC_IS_DEAD(node->flags))
-				maskbit[idx >> 5] &= ~(1u << (idx & 0x1f));
-		} ccv_nnc_graph_visit_endfor
-		ccv_nnc_graph_visit_free(reverse_visit);
 		// Fill in blanks (i.e. the backward ops that are not showing in above, but should be included to avoid excluding necessary ones). This is done by flowing gradients from outputs back all the way to inputs.
 		ccv_array_clear(input_gradient_execs);
 		ccv_array_clear(output_gradient_execs);