Skip to content

Commit

Permalink
Remove cases where the deprecated --xla_gpu_simplify_all_fp_conversio…
Browse files Browse the repository at this point in the history
…ns is set to its default value.

PiperOrigin-RevId: 633645462
  • Loading branch information
dimitar-asenov authored and maxtext authors committed May 14, 2024
1 parent 87c6430 commit 618450b
Show file tree
Hide file tree
Showing 6 changed files with 6 additions and 6 deletions.
2 changes: 1 addition & 1 deletion MaxText/configs/a3/llama_2_7b/16vm.sh
Original file line number Diff line number Diff line change
Expand Up @@ -26,7 +26,7 @@ export NVTE_FUSED_ATTN=1
export NCCL_DEBUG=VERSION
export XLA_FLAGS="--xla_dump_to=$OUTPUT_PATH/$RUN_NAME/HLO_dumps/
--xla_gpu_enable_latency_hiding_scheduler=true --xla_gpu_enable_async_all_gather=true --xla_gpu_enable_async_reduce_scatter=true
--xla_gpu_enable_triton_gemm=false --xla_gpu_simplify_all_fp_conversions --xla_gpu_graph_level=0
--xla_gpu_enable_triton_gemm=false --xla_gpu_graph_level=0
--xla_gpu_enable_async_all_reduce=true --xla_gpu_enable_highest_priority_async_stream=true
--xla_gpu_all_reduce_combine_threshold_bytes=1073741824 --xla_gpu_all_gather_combine_threshold_bytes=134217728
--xla_gpu_reduce_scatter_combine_threshold_bytes=134217728 --xla_gpu_enable_pipelined_all_gather=true
Expand Down
2 changes: 1 addition & 1 deletion MaxText/configs/a3/llama_2_7b/1vm.sh
Original file line number Diff line number Diff line change
Expand Up @@ -26,7 +26,7 @@ export NVTE_FUSED_ATTN=1
export NCCL_DEBUG=VERSION
export XLA_FLAGS="--xla_dump_to=$OUTPUT_PATH/$RUN_NAME/HLO_dumps/
--xla_gpu_enable_latency_hiding_scheduler=true --xla_gpu_enable_async_all_gather=true
--xla_gpu_enable_async_reduce_scatter=true --xla_gpu_enable_triton_gemm=false --xla_gpu_simplify_all_fp_conversions
--xla_gpu_enable_async_reduce_scatter=true --xla_gpu_enable_triton_gemm=false
--xla_gpu_graph_level=0 --xla_gpu_enable_async_all_reduce=true --xla_gpu_enable_highest_priority_async_stream=true
--xla_gpu_all_reduce_combine_threshold_bytes=134217728 --xla_gpu_all_gather_combine_threshold_bytes=134217728
--xla_gpu_reduce_scatter_combine_threshold_bytes=67108864 --xla_gpu_enable_pipelined_all_gather=true
Expand Down
2 changes: 1 addition & 1 deletion MaxText/configs/a3/llama_2_7b/2vm.sh
Original file line number Diff line number Diff line change
Expand Up @@ -26,7 +26,7 @@ export NVTE_FUSED_ATTN=1
export NCCL_DEBUG=VERSION
export XLA_FLAGS="--xla_dump_to=$OUTPUT_PATH/$RUN_NAME/HLO_dumps/
--xla_gpu_enable_latency_hiding_scheduler=true --xla_gpu_enable_async_all_gather=true --xla_gpu_enable_async_reduce_scatter=true
--xla_gpu_enable_triton_gemm=false --xla_gpu_simplify_all_fp_conversions --xla_gpu_graph_level=0
--xla_gpu_enable_triton_gemm=false --xla_gpu_graph_level=0
--xla_gpu_enable_async_all_reduce=true --xla_gpu_enable_highest_priority_async_stream=true
--xla_gpu_all_reduce_combine_threshold_bytes=67108864 --xla_gpu_all_gather_combine_threshold_bytes=134217728
--xla_gpu_reduce_scatter_combine_threshold_bytes=67108864 --xla_gpu_enable_pipelined_all_gather=true
Expand Down
2 changes: 1 addition & 1 deletion MaxText/configs/a3/llama_2_7b/4vm.sh
Original file line number Diff line number Diff line change
Expand Up @@ -26,7 +26,7 @@ export NVTE_FUSED_ATTN=1
export NCCL_DEBUG=VERSION
export XLA_FLAGS="--xla_dump_to=$OUTPUT_PATH/$RUN_NAME/HLO_dumps/
--xla_gpu_enable_latency_hiding_scheduler=true --xla_gpu_enable_async_all_gather=true
--xla_gpu_enable_async_reduce_scatter=true --xla_gpu_enable_triton_gemm=false --xla_gpu_simplify_all_fp_conversions
--xla_gpu_enable_async_reduce_scatter=true --xla_gpu_enable_triton_gemm=false
--xla_gpu_graph_level=0 --xla_gpu_enable_async_all_reduce=true --xla_gpu_enable_highest_priority_async_stream=true
--xla_gpu_all_reduce_combine_threshold_bytes=536870912 --xla_gpu_all_gather_combine_threshold_bytes=134217728
--xla_gpu_reduce_scatter_combine_threshold_bytes=67108864 --xla_gpu_enable_pipelined_all_gather=true
Expand Down
2 changes: 1 addition & 1 deletion MaxText/configs/a3/llama_2_7b/8vm.sh
Original file line number Diff line number Diff line change
Expand Up @@ -26,7 +26,7 @@ export NVTE_FUSED_ATTN=1
export NCCL_DEBUG=VERSION
export XLA_FLAGS="--xla_dump_to=$OUTPUT_PATH/$RUN_NAME/HLO_dumps/
--xla_gpu_enable_latency_hiding_scheduler=true --xla_gpu_enable_async_all_gather=true --xla_gpu_enable_async_reduce_scatter=true
--xla_gpu_enable_triton_gemm=false --xla_gpu_simplify_all_fp_conversions --xla_gpu_graph_level=0
--xla_gpu_enable_triton_gemm=false --xla_gpu_graph_level=0
--xla_gpu_enable_async_all_reduce=true --xla_gpu_enable_highest_priority_async_stream=true
--xla_gpu_all_reduce_combine_threshold_bytes=1073741824 --xla_gpu_all_gather_combine_threshold_bytes=134217728
--xla_gpu_reduce_scatter_combine_threshold_bytes=67108864 --xla_gpu_enable_pipelined_all_gather=true
Expand Down
2 changes: 1 addition & 1 deletion end_to_end/gpu/a3/test_llama2_7b.sh
Original file line number Diff line number Diff line change
Expand Up @@ -54,7 +54,7 @@ export NCCL_DEBUG=VERSION

export XLA_FLAGS="--xla_dump_to=$BASE_OUTPUT_PATH/$RUN_NAME/HLO_dumps/
--xla_gpu_enable_latency_hiding_scheduler=true --xla_gpu_enable_async_all_gather=true
--xla_gpu_enable_async_reduce_scatter=true --xla_gpu_enable_triton_gemm=false --xla_gpu_simplify_all_fp_conversions
--xla_gpu_enable_async_reduce_scatter=true --xla_gpu_enable_triton_gemm=false
--xla_gpu_graph_level=0 --xla_gpu_enable_async_all_reduce=true --xla_gpu_enable_highest_priority_async_stream=true
--xla_gpu_all_reduce_combine_threshold_bytes=134217728 --xla_gpu_all_gather_combine_threshold_bytes=134217728
--xla_gpu_reduce_scatter_combine_threshold_bytes=67108864 --xla_gpu_enable_pipelined_all_gather=true
Expand Down

0 comments on commit 618450b

Please sign in to comment.