[moe] align train settings and losses by oahzxl · Pull Request #4655 · hpcaitech/ColossalAI

oahzxl · 2023-09-07T08:37:04Z

add expert router loss
add crossentropy + z loss
use adafactor as optimizer

…into moe_train

github-actions · 2023-09-07T11:29:16Z

The code coverage for the changed files is 5%.

Click me to view the complete report

Name                                                      Stmts   Miss  Cover
-----------------------------------------------------------------------------
colossalai/amp/naive_amp/mixed_precision_optimizer.py        98     98     0%
colossalai/booster/booster.py                                66     66     0%
colossalai/booster/plugin/__init__.py                        11     11     0%
colossalai/booster/plugin/hybrid_parallel_plugin.py         152    152     0%
colossalai/booster/plugin/pp_plugin_base.py                   9      9     0%
colossalai/cluster/__init__.py                                5      0   100%
colossalai/cluster/process_group_mesh.py                     72     46    36%
colossalai/context/moe_context.py                            60     30    50%
colossalai/engine/gradient_handler/__init__.py                6      0   100%
colossalai/interface/optimizer.py                            45     20    56%
colossalai/kernel/cuda_native/__init__.py                     5      0   100%
colossalai/lazy/lazy_init.py                                315    246    22%
colossalai/nn/layer/moe/__init__.py                           6      0   100%
colossalai/nn/layer/moe/_operation.py                       143     98    31%
colossalai/nn/layer/moe/checkpoint.py                        56     37    34%
colossalai/nn/layer/moe/experts.py                           84     65    23%
colossalai/nn/layer/moe/layers.py                            81     60    26%
colossalai/nn/layer/moe/routers.py                          143    119    17%
colossalai/nn/layer/moe/utils.py                             58     38    34%
colossalai/pipeline/p2p.py                                  102    102     0%
colossalai/pipeline/schedule/__init__.py                      3      3     0%
colossalai/pipeline/schedule/_utils.py                       50     50     0%
colossalai/pipeline/schedule/base.py                         10     10     0%
colossalai/pipeline/schedule/one_f_one_b.py                 116    116     0%
colossalai/pipeline/stage_manager.py                         68     68     0%
colossalai/shardformer/_utils.py                             54     54     0%
colossalai/shardformer/layer/__init__.py                      8      8     0%
colossalai/shardformer/layer/embedding.py                   130    130     0%
colossalai/shardformer/layer/linear.py                      181    181     0%
colossalai/shardformer/layer/normalization.py                51     51     0%
colossalai/shardformer/layer/qkv_fused_linear.py            292    292     0%
colossalai/shardformer/layer/utils.py                        84     84     0%
colossalai/shardformer/modeling/bert.py                     431    431     0%
colossalai/shardformer/modeling/blip2.py                     53     53     0%
colossalai/shardformer/modeling/bloom.py                    387    387     0%
colossalai/shardformer/modeling/chatglm.py                  149    149     0%
colossalai/shardformer/modeling/gpt2.py                     293    293     0%
colossalai/shardformer/modeling/jit.py                       19     19     0%
colossalai/shardformer/modeling/llama.py                    204    204     0%
colossalai/shardformer/modeling/opt.py                      285    285     0%
colossalai/shardformer/modeling/sam.py                       94     94     0%
colossalai/shardformer/modeling/t5.py                       297    297     0%
colossalai/shardformer/modeling/vit.py                      149    149     0%
colossalai/shardformer/modeling/whisper.py                   95     95     0%
colossalai/shardformer/policies/auto_policy.py               27     27     0%
colossalai/shardformer/policies/base_policy.py               87     87     0%
colossalai/shardformer/policies/bert.py                     257    257     0%
colossalai/shardformer/policies/blip2.py                     54     54     0%
colossalai/shardformer/policies/bloom.py                    151    151     0%
colossalai/shardformer/policies/chatglm.py                  100    100     0%
colossalai/shardformer/policies/gpt2.py                     181    181     0%
colossalai/shardformer/policies/llama.py                    114    114     0%
colossalai/shardformer/policies/opt.py                      140    140     0%
colossalai/shardformer/policies/sam.py                       32     32     0%
colossalai/shardformer/policies/t5.py                       182    182     0%
colossalai/shardformer/policies/vit.py                      108    108     0%
colossalai/shardformer/policies/whisper.py                   61     61     0%
colossalai/shardformer/shard/shard_config.py                 28     28     0%
colossalai/shardformer/shard/sharder.py                      95     95     0%
colossalai/shardformer/shard/shardformer.py                  15     15     0%
colossalai/shardformer/shard/utils.py                        11     11     0%
colossalai/tensor/d_tensor/api.py                           149    113    24%
colossalai/tensor/moe_tensor/__init__.py                      0      0   100%
colossalai/tensor/moe_tensor/api.py                          20      8    60%
colossalai/tensor/moe_tensor/moe_info.py                     10      7    30%
colossalai/zero/low_level/low_level_optim.py                330    288    13%
tests/kit/model_zoo/transformers/__init__.py                 12     12     0%
tests/kit/model_zoo/transformers/bert.py                     50     50     0%
tests/kit/model_zoo/transformers/blip2.py                    21     21     0%
tests/kit/model_zoo/transformers/bloom.py                    36     36     0%
tests/kit/model_zoo/transformers/chatglm.py                  20     20     0%
tests/kit/model_zoo/transformers/gpt.py                      39     39     0%
tests/kit/model_zoo/transformers/opt.py                      32     32     0%
tests/kit/model_zoo/transformers/sam.py                      14     14     0%
tests/kit/model_zoo/transformers/t5.py                       25     25     0%
tests/kit/model_zoo/transformers/vit.py                      24     24     0%
tests/kit/model_zoo/transformers/whisper.py                  23     23     0%
tests/test_shardformer/test_model/_utils.py                 142    142     0%
tests/test_shardformer/test_model/test_shard_bert.py         62     62     0%
tests/test_shardformer/test_model/test_shard_blip2.py        40     40     0%
tests/test_shardformer/test_model/test_shard_bloom.py        59     59     0%
tests/test_shardformer/test_model/test_shard_chatglm.py      60     60     0%
tests/test_shardformer/test_model/test_shard_gpt2.py         65     65     0%
tests/test_shardformer/test_model/test_shard_llama.py        62     62     0%
tests/test_shardformer/test_model/test_shard_opt.py          62     62     0%
tests/test_shardformer/test_model/test_shard_sam.py          39     39     0%
tests/test_shardformer/test_model/test_shard_t5.py           59     59     0%
tests/test_shardformer/test_model/test_shard_vit.py          61     61     0%
tests/test_shardformer/test_model/test_shard_whisper.py      46     46     0%
tests/test_shardformer/test_shard_utils.py                   21     21     0%
tests/test_shardformer/test_with_torch_ddp.py                52     52     0%
-----------------------------------------------------------------------------
TOTAL                                                      8298   7885     5%

github-actions · 2023-09-08T04:30:04Z

The code coverage for the changed files is 82%.

Click me to view the complete report

Name                                                                    Stmts   Miss  Cover
-------------------------------------------------------------------------------------------
colossalai/amp/naive_amp/mixed_precision_optimizer.py                      98     20    80%
colossalai/booster/booster.py                                              66      9    86%
colossalai/booster/plugin/__init__.py                                      11      0   100%
colossalai/booster/plugin/hybrid_parallel_plugin.py                       152     17    89%
colossalai/booster/plugin/pp_plugin_base.py                                 9      1    89%
colossalai/cluster/__init__.py                                              5      0   100%
colossalai/cluster/process_group_mesh.py                                   72      1    99%
colossalai/context/moe_context.py                                          60     30    50%
colossalai/engine/gradient_handler/__init__.py                              6      0   100%
colossalai/interface/optimizer.py                                          45      5    89%
colossalai/kernel/cuda_native/__init__.py                                   5      0   100%
colossalai/lazy/lazy_init.py                                              315     44    86%
colossalai/nn/layer/moe/__init__.py                                         6      0   100%
colossalai/nn/layer/moe/_operation.py                                     143     98    31%
colossalai/nn/layer/moe/checkpoint.py                                      56     37    34%
colossalai/nn/layer/moe/experts.py                                         84     65    23%
colossalai/nn/layer/moe/layers.py                                          81     60    26%
colossalai/nn/layer/moe/routers.py                                        143    119    17%
colossalai/nn/layer/moe/utils.py                                           58     38    34%
colossalai/pipeline/p2p.py                                                102      7    93%
colossalai/pipeline/schedule/__init__.py                                    3      0   100%
colossalai/pipeline/schedule/_utils.py                                     50      5    90%
colossalai/pipeline/schedule/base.py                                       10      1    90%
colossalai/pipeline/schedule/one_f_one_b.py                               116      4    97%
colossalai/pipeline/stage_manager.py                                       68      4    94%
colossalai/shardformer/_utils.py                                           54     15    72%
colossalai/shardformer/layer/__init__.py                                    8      0   100%
colossalai/shardformer/layer/embedding.py                                 130     24    82%
colossalai/shardformer/layer/linear.py                                    181     53    71%
colossalai/shardformer/layer/normalization.py                              51     10    80%
colossalai/shardformer/layer/qkv_fused_linear.py                          292     70    76%
colossalai/shardformer/layer/utils.py                                      84     17    80%
colossalai/shardformer/modeling/bert.py                                   431    128    70%
colossalai/shardformer/modeling/blip2.py                                   53      1    98%
colossalai/shardformer/modeling/bloom.py                                  387    122    68%
colossalai/shardformer/modeling/chatglm2_6b/configuration_chatglm.py       30      0   100%
colossalai/shardformer/modeling/chatglm2_6b/modeling_chatglm.py           571    239    58%
colossalai/shardformer/modeling/chatglm.py                                149     34    77%
colossalai/shardformer/modeling/gpt2.py                                   293     83    72%
colossalai/shardformer/modeling/jit.py                                     19      3    84%
colossalai/shardformer/modeling/llama.py                                  204     65    68%
colossalai/shardformer/modeling/opt.py                                    285     65    77%
colossalai/shardformer/modeling/sam.py                                     94      6    94%
colossalai/shardformer/modeling/t5.py                                     297     74    75%
colossalai/shardformer/modeling/vit.py                                    149     23    85%
colossalai/shardformer/modeling/whisper.py                                 95     13    86%
colossalai/shardformer/policies/auto_policy.py                             27      2    93%
colossalai/shardformer/policies/base_policy.py                             87     11    87%
colossalai/shardformer/policies/bert.py                                   257      0   100%
colossalai/shardformer/policies/blip2.py                                   54      2    96%
colossalai/shardformer/policies/bloom.py                                  151      2    99%
colossalai/shardformer/policies/chatglm.py                                100      6    94%
colossalai/shardformer/policies/gpt2.py                                   181      1    99%
colossalai/shardformer/policies/llama.py                                  114      3    97%
colossalai/shardformer/policies/opt.py                                    140      2    99%
colossalai/shardformer/policies/sam.py                                     32      0   100%
colossalai/shardformer/policies/t5.py                                     182      5    97%
colossalai/shardformer/policies/vit.py                                    108      1    99%
colossalai/shardformer/policies/whisper.py                                 61      2    97%
colossalai/shardformer/shard/shard_config.py                               28      0   100%
colossalai/shardformer/shard/sharder.py                                    95      3    97%
colossalai/shardformer/shard/shardformer.py                                15      0   100%
colossalai/shardformer/shard/utils.py                                      11      0   100%
colossalai/tensor/d_tensor/api.py                                         149     24    84%
colossalai/tensor/moe_tensor/__init__.py                                    0      0   100%
colossalai/tensor/moe_tensor/api.py                                        20      7    65%
colossalai/tensor/moe_tensor/moe_info.py                                   10      7    30%
colossalai/zero/low_level/low_level_optim.py                              330     30    91%
tests/kit/model_zoo/transformers/__init__.py                               12      0   100%
tests/kit/model_zoo/transformers/bert.py                                   50      0   100%
tests/kit/model_zoo/transformers/blip2.py                                  21      0   100%
tests/kit/model_zoo/transformers/bloom.py                                  36      0   100%
tests/kit/model_zoo/transformers/chatglm.py                                20      0   100%
tests/kit/model_zoo/transformers/gpt.py                                    39      0   100%
tests/kit/model_zoo/transformers/opt.py                                    32      0   100%
tests/kit/model_zoo/transformers/sam.py                                    14      0   100%
tests/kit/model_zoo/transformers/t5.py                                     25      0   100%
tests/kit/model_zoo/transformers/vit.py                                    24      0   100%
tests/kit/model_zoo/transformers/whisper.py                                23      0   100%
tests/test_booster/test_plugin/test_3d_plugin.py                           64      7    89%
tests/test_booster/test_plugin/test_gemini_plugin.py                       74     10    86%
tests/test_cluster/test_process_group_mesh.py                              86      1    99%
tests/test_fx/test_tracer/test_hf_model/hf_tracer_utils.py                 21      2    90%
tests/test_fx/test_tracer/test_hf_model/test_hf_bert.py                    17      1    94%
tests/test_fx/test_tracer/test_hf_model/test_hf_gpt.py                     17      1    94%
tests/test_lazy/test_models.py                                             14      1    93%
tests/test_pipeline/test_p2p_communication.py                              44      1    98%
tests/test_pipeline/test_schedule/test_oneF_oneB.py                        80      2    98%
tests/test_pipeline/test_schedule/test_pipeline_schedule_utils.py          40      0   100%
tests/test_pipeline/test_stage_manager.py                                  52      1    98%
tests/test_shardformer/test_layer/test_embedding.py                        37      1    97%
tests/test_shardformer/test_layer/test_gpt2_qkv_fused_linear_1d.py         89      1    99%
tests/test_shardformer/test_layer/test_layernorm.py                        35      1    97%
tests/test_shardformer/test_layer/test_linear_1d.py                       110      1    99%
tests/test_shardformer/test_layer/test_qkv_fused_linear_1d.py              51      1    98%
tests/test_shardformer/test_layer/test_vocab_parallel_embedding_1d.py      39      1    97%
tests/test_shardformer/test_model/_utils.py                               142     21    85%
tests/test_shardformer/test_model/test_shard_bert.py                       62      1    98%
tests/test_shardformer/test_model/test_shard_blip2.py                      40      1    98%
tests/test_shardformer/test_model/test_shard_bloom.py                      59      1    98%
tests/test_shardformer/test_model/test_shard_chatglm.py                    60      1    98%
tests/test_shardformer/test_model/test_shard_gpt2.py                       65      1    98%
tests/test_shardformer/test_model/test_shard_llama.py                      62      1    98%
tests/test_shardformer/test_model/test_shard_opt.py                        62      1    98%
tests/test_shardformer/test_model/test_shard_sam.py                        39      1    97%
tests/test_shardformer/test_model/test_shard_t5.py                         59      1    98%
tests/test_shardformer/test_model/test_shard_vit.py                        61      1    98%
tests/test_shardformer/test_model/test_shard_whisper.py                    46      1    98%
tests/test_shardformer/test_shard_utils.py                                 21      0   100%
tests/test_shardformer/test_with_torch_ddp.py                              52      1    98%
tests/test_utils/test_flash_attention.py                                   92      8    91%
-------------------------------------------------------------------------------------------
TOTAL                                                                    9861   1792    82%

* init * update moe ckpt * update config * support openmoe infernece * update config * remove pdb * support train * add ckpt download * update ckpt loading * use general ckpt * add loss and optim * update ci * update require

oahzxl added 13 commits September 4, 2023 13:25

init

bbae495

update moe ckpt

255cdae

update config

dbf634d

support openmoe infernece

a4d3c0a

update config

bb6eda0

remove pdb

1b51447

support train

a6768c5

Merge branch 'feature/moe' of https://github.com/hpcaitech/ColossalAI …

4346fa9

…into moe_train

add ckpt download

88f15be

update ckpt loading

1bd11cd

use general ckpt

2cacb8e

add loss and optim

faf5aaa

merge

af1731a

oahzxl added 2 commits September 8, 2023 10:48

update ci

148b030

update require

ca4cd89

ver217 approved these changes Sep 8, 2023

View reviewed changes

ver217 merged commit 1cd9e88 into hpcaitech:feature/moe Sep 8, 2023

oahzxl deleted the loss branch September 12, 2023 02:12

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[moe] align train settings and losses#4655

[moe] align train settings and losses#4655
ver217 merged 15 commits intohpcaitech:feature/moefrom
oahzxl:loss

oahzxl commented Sep 7, 2023

Uh oh!

github-actions Bot commented Sep 7, 2023

Uh oh!

github-actions Bot commented Sep 8, 2023

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

oahzxl commented Sep 7, 2023

Uh oh!

github-actions Bot commented Sep 7, 2023

Uh oh!

github-actions Bot commented Sep 8, 2023

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants