Cinn mid #13
Open
Cinn mid #13
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…lePaddle#67331) * Support Sharding release grad * Support Sharding release grad
…ed/communication/stream/{all_reduce,all_to_all}.py` (PaddlePaddle#67112)
* fix
* fix
* CrfDecoding op * delete config * add if condition * add return true * Update multiary_infer_sym.cc
* add cinn * add cinn * fix setvalue * fix segmentpoll * fix * fix * fix segment * fix segment * fix segment * close ctest * fix type * fix segment * fix segment * fix * fix * fix * fix codestyle * fix segment * fix * fix * fix
…pout_add,bfgs,lbfgs,math}.py` in incubate APIs (PaddlePaddle#67233) * 添加类型信息 * 追加类型信息 * 按要求进行修改 * 修改参数取值范围 * 修改默认参数 * 修改TYPE_CHECKING语句块位置
…ePaddle#67298) * udpate F.pad * udpate F.pad
…0 files (PaddlePaddle#67340) --------- Co-authored-by: Nyakku Shigure <[email protected]>
…0 files in `python/paddle/` (PaddlePaddle#67363)
…0 files in `python/paddle/` (PaddlePaddle#67376)
…ion_） (PaddlePaddle#67339) * delete -1 shape check * recommit * recover support -1 shape * optional output test * recover * modify infermeta gen
…files in `python/paddle/` (PaddlePaddle#67359)
PaddlePaddle#67115) * test_index_select_strided * testop * delete error annotation * annotation
…p.py,test_signal.py,test_stride.py`) (PaddlePaddle#67389) * fix 3 pir uts * fix camke setting
…o_box_post updata_loss_scaling_ tdm_child (PaddlePaddle#67121) * change op yolo_box yolo_box_head yolo_box_post updata_loss_scaling_ tdm_child * update multiary_infer_sym.cc multiary_infer_sym.h binary_infer_sym.cc * Update binary_infer_sym.cc * Update binary_infer_sym.cc * Update binary_infer_sym.cc * Update binary_infer_sym.cc * Update binary_infer_sym.cc
…per.py` (PaddlePaddle#67366) * fix * fix
* support fetch startup param * refine * refine
…zeAbsMax Operator for CINN compiler (PaddlePaddle#66619) * summon * add test * addTest * open * Update test_fake_quantize_op.py
…addlePaddle#67230) * fix pir ut (test_test_slice_scatter,test_multiplex_op) * update
…67477) * [PIR] Deny no operand ops in CINN (e.g. FullOp) * use FLAGS_use_cinn in with cinn only * run test in cinn mode only
* batch_fc * add constrain
…sor operator for CINN compiler (PaddlePaddle#67337) * no test * unchanged vector * moe * Update same_operands_result.cc * unchange * name * name * test * err * same same same next sym next sym next sym densetensor densetensortype densetensortype densetensortype rois_num dyn_cast ? dyn_cast * dyn_cast * new sym * no moe optest * new test * mkddim * no mkddim * no mkddim * pass * remove unused * fix * fix * fix * fix * fix * fixx * fixx * fixx * todo * todo
…ograd/ir_backward.py,python/paddle/cost_model/cost_model.py` (PaddlePaddle#66890) * ir_backward.py * cost_model.py * add optional * added type for return value * add from __future__ * add from __future__ * fixed * fixed
…max_pool3d_with_index (PaddlePaddle#67390)
* add pipeline memory estimator * adjust func location and add notes * rewrite this tool * consider no need buffer vars * remove print * init PipelineZeroBubbleVirtualPipelinePass * add _partial_programs for zb_vpp * fix bug when run estimate_program_mem_usagess * finish using `estimate_program_mem_usagess` in zb vpp * fit pipeline memory estimator for mp * finish zbv's previous 2 steps * finish zbv step `_insert_forward_jobs_before_backward_b` * finish zbv schedule and verify the vpp2 pp2 mem unlimited situation * finish _create_job_list * fix for some case * add forward number limit when fill bubble before backward_b * consider following jobs' memory when insert forward * fit for vpp_degree > 2 * fix job name in _create_job_list * fix the error about some backward_b job not insert * update * fix hang * fit for dealed_op_idx * update _can_schedule_task * fix for cannot swith to dynamic mode * remove useless notes * consider loss time * consider loss time - fix * remove chunk_id == 0 judge when insert w job * update default program_runtime * add support for using program_runtimes and memory_limit_times to do zbv * add fill_loss_stage option * use pp_group to select mem * update default value of memory_limit_times and program_runtimes * fix * update default memory_limit_times * Update constants.py * add unit test * remove print * fix unittest * update test * fix codestyle * update test time * Update pass_utils.py * Update pipeline_zero_bubble.py * Update pipeline_zero_bubble.py * fix code style
… BF16 padding 0x8000 scheme. Modify some unit tests that failed due to the upgrade. (PaddlePaddle#67460)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
PR Category
PR Types
Description