* sequence parallel optimization
* validate sequence parallel in llama (code to be polished)
* shardformer api writing
* integrate sequence parallel in ShardFormer
* fix pp bugs and sp bugs for LlaMa model
* integrating ring-based sequence parallelism into ShardFormer
* [sequence parallelism]: Add fused megatron function
* integrating ring-based sequence parallelism into ShardFormer
---------
Co-authored-by: linsj20 <linsj20@mails.tsinghua.edu.cn>
* fix bugs when useing sp and flashattention together
* fix operation function name
* support flash attention for ulysses-style sp
* clarify sp process group
* fix compatibility bugs in moe plugin
* fix fused linear bugs
* fix linear layer test
* support gpt model all-to-all sp
* modify shard data dimension (meant to be dim=-1)
* support megtron-style sp and distributed attn for llama model
* [shardformer] add megatron sp to llama
* support llama7B 128k with distributed attention
* [shardformer] robustness enhancement
* add block attn
* sp mode 1: keep input as a complete sequence
* fix sp compatability
* finish sp mode 3 support for gpt
* using all_to_all_single when batch size is 1
* support mode 2 sp in gpt2 (#5)
* [shardformer] add megatron sp to llama
* support llama7B 128k with distributed attention
* [shardformer] robustness enhancement
* add block attn
* sp mode 1: keep input as a complete sequence
* fix sp compatability
* refactor ring implementation
* support mode 2 sp in gpt2
* polish code
* enable distributed attn mask when using sp mode 2 and 3 in llama
* automatically enable flash attn when using sp mode 2 and 3 in llama
* inplace attn mask
* add zero2 support for sequence parallel
* polish code
* fix bugs
* fix gemini checkpoint io
* loose tensor checking atol and rtol
* add comment
* fix llama layernorm grad
* fix zero grad
* fix zero grad
* fix conflict
* update split and gather auto grad func
* sequence parallel: inside text split (#6)
* polish code (part 1)
* polish code (part 2)
* polish code (part 2.5)
* polish code (part 3)
* sequence parallel: inside text split
* miscellaneous minor fixes
* polish code
* fix ulysses style ZeRO
* sequence parallel: inside text split
* miscellaneous minor fixes
* disaggregate sp group and dp group for sp
* fix llama and gpt sp
* polish code
* move ulysses grad sync to ddp (#9)
* remove zero_stage and unbind the grad sync for alltoall sp
* add 2d group creation test
* move ulysses grad sync to ddp
* add 2d group creation test
* remove useless code
* change shard config not to enable sp when enable_all_optimizations
* add sp warnings for several model
* remove useless code
---------
Co-authored-by: linsj20 <linsj20@mails.tsinghua.edu.cn>
* fix: simplify merge_batch
* fix: use return_outputs=False to eliminate extra memory consumption
* feat: add return_outputs warning
* style: remove `return_outputs=False` as it is the default value
* test: add more p2p tests
* fix: remove send_forward_recv_forward as p2p op list need to use the same group
* fix: make send and receive atomic
* feat: update P2PComm fn
* feat: add metadata cache in 1f1b
* feat: add metadata cache in interleaved pp
* feat: modify is_xx_stage fn
* revert: add _broadcast_object_list
* feat: add interleaved pp in llama policy
* feat: set NCCL_BUFFSIZE in HybridParallelPlugin
* fix 3d checkpoint load when booster boost without optimizer
fix 3d checkpoint load when booster boost without optimizer
* test ci
* revert ci
* fix
fix
* [shardformer] implement policy for all GPT-J models and test
* [shardformer] support interleaved pipeline parallel for bert finetune
* [shardformer] shardformer support falcon (#4883)
* [shardformer]: fix interleaved pipeline for bert model (#5048)
* [hotfix]: disable seq parallel for gptj and falcon, and polish code (#5093)
* Add Mistral support for Shardformer (#5103)
* [shardformer] add tests to mistral (#5105)
---------
Co-authored-by: Pengtai Xu <henryxu880@gmail.com>
Co-authored-by: ppt0011 <143150326+ppt0011@users.noreply.github.com>
Co-authored-by: flybird11111 <1829166702@qq.com>
Co-authored-by: eric8607242 <e0928021388@gmail.com>
* [feature] support no master weights for low level zero plugin
* [feature] support no master weights for low level zero plugin, remove data copy when no master weights
* remove data copy and typecasting when no master weights
* not load weights to cpu when using no master weights
* fix grad: use fp16 grad when no master weights
* only do not update working param when no master weights
* fix: only do not update working param when no master weights
* fix: passing params in dict format in hybrid plugin
* fix: remove extra params (tp_process_group) in hybrid_parallel_plugin
* add APIs
* implement save_sharded_model
* add test for hybrid checkpointio
* implement naive loading for sharded model
* implement efficient sharded model loading
* open a new file for hybrid checkpoint_io
* small fix
* fix circular importing
* fix docstring
* arrange arguments and apis
* small fix
* [shardformer/sequence parallel] Support sequence parallel for gpt2 (#4384)
* [sequence parallel] add sequence parallel linear col/row support (#4336)
* add sequence parallel linear col/row support
* add annotation
* add annotation
* add support for gpt2 fused qkv linear layer
* support sequence parallel in GPT2
* add docstring and note
* add requirments
* remove unused flash-attb
* modify flash attn test
* modify flash attn setting
* modify flash attn code
* add assert before divide, rename forward function
* [shardformer/test] fix gpt2 test with seq-parallel
* [shardformer/sequence parallel] Overlap input gather and grad computation during col backward (#4401)
* overlap gather input / grad computing during col backward
* modify test for overlap
* simplify code
* fix code and modify cuda stream synchronize
* [shardformer/sequence parallel] polish code
* add naive optimizer for 3DPlugin/refactor gpt2 shardformer test
* merge tests of PP/DP/TP combinations into one test file
* fix bug when sync grad for dp in HybridPlugin
* update supported precisions for 3DPlugin/fix bug when shifting tp_degree
* improve the passing of lazy_init
* modify lazy_init/use sync_shared_params
Zhongkai Zhao