Unifying TBE API using List (Frontend) #3711
Closed
Conversation
This file contains hidden or 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
|
This pull request was exported from Phabricator. Differential Revision: D68055168 |
✅ Deploy Preview for pytorch-fbgemm-docs ready!
To edit notification comments on pull requests, go to your Netlify site configuration. |
spcyppt
added a commit
to spcyppt/torchrec
that referenced
this pull request
Feb 19, 2025
Summary: X-link: pytorch/FBGEMM#3711 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap Differential Revision: D68055168
spcyppt
added a commit
to spcyppt/FBGEMM
that referenced
this pull request
Feb 19, 2025
Summary: X-link: pytorch/torchrec#2751 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap Differential Revision: D68055168
spcyppt
added a commit
to spcyppt/FBGEMM
that referenced
this pull request
Feb 21, 2025
Summary: X-link: pytorch/torchrec#2751 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap Differential Revision: D68055168
spcyppt
added a commit
to spcyppt/torchrec
that referenced
this pull request
Feb 21, 2025
Summary: X-link: pytorch/FBGEMM#3711 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap Differential Revision: D68055168
spcyppt
added a commit
to spcyppt/FBGEMM
that referenced
this pull request
Feb 21, 2025
Summary: X-link: pytorch/torchrec#2751 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap Differential Revision: D68055168
spcyppt
force-pushed
the
export-D68055168
branch
from
34a17c6 to
83fc96d
Compare
|
This pull request was exported from Phabricator. Differential Revision: D68055168 |
spcyppt
added a commit
to spcyppt/torchrec
that referenced
this pull request
Feb 26, 2025
Summary: X-link: pytorch/FBGEMM#3711 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap Differential Revision: D68055168
spcyppt
force-pushed
the
export-D68055168
branch
from
83fc96d to
6406ea2
Compare
spcyppt
added a commit
to spcyppt/FBGEMM
that referenced
this pull request
Feb 26, 2025
Summary: X-link: pytorch/torchrec#2751 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap Differential Revision: D68055168
spcyppt
added a commit
to spcyppt/torchrec
that referenced
this pull request
Feb 26, 2025
Summary: X-link: pytorch/FBGEMM#3711 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap Differential Revision: D68055168
spcyppt
added a commit
to spcyppt/FBGEMM
that referenced
this pull request
Feb 26, 2025
Summary: X-link: pytorch/torchrec#2751 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap Differential Revision: D68055168
spcyppt
force-pushed
the
export-D68055168
branch
from
6406ea2 to
7a1b97f
Compare
spcyppt
added a commit
to spcyppt/FBGEMM
that referenced
this pull request
Feb 26, 2025
Summary: X-link: pytorch/torchrec#2751 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap Differential Revision: D68055168
spcyppt
added a commit
to spcyppt/FBGEMM
that referenced
this pull request
Feb 26, 2025
Summary: X-link: pytorch/torchrec#2751 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap Differential Revision: D68055168
spcyppt
added a commit
to spcyppt/torchrec
that referenced
this pull request
Mar 3, 2025
Summary: X-link: pytorch/FBGEMM#3711 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap Differential Revision: D68055168
spcyppt
force-pushed
the
export-D68055168
branch
from
7a1b97f to
4256f60
Compare
spcyppt
added a commit
to spcyppt/FBGEMM
that referenced
this pull request
Mar 3, 2025
Summary: X-link: pytorch/torchrec#2751 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap Differential Revision: D68055168
spcyppt
added a commit
to spcyppt/torchrec
that referenced
this pull request
Mar 3, 2025
Summary: X-link: pytorch/FBGEMM#3711 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap Differential Revision: D68055168
spcyppt
added a commit
to spcyppt/FBGEMM
that referenced
this pull request
Mar 3, 2025
Summary: X-link: pytorch/torchrec#2751 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap Differential Revision: D68055168
spcyppt
force-pushed
the
export-D68055168
branch
from
4256f60 to
6b31758
Compare
|
This pull request was exported from Phabricator. Differential Revision: D68055168 |
spcyppt
force-pushed
the
export-D68055168
branch
from
6b31758 to
c8b67fd
Compare
|
This pull request was exported from Phabricator. Differential Revision: D68055168 |
spcyppt
force-pushed
the
export-D68055168
branch
from
c8b67fd to
c4c83cc
Compare
spcyppt
added a commit
to spcyppt/torchrec
that referenced
this pull request
Mar 6, 2025
Summary: X-link: pytorch/FBGEMM#3711 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap, nautsimon Differential Revision: D68055168
spcyppt
added a commit
to spcyppt/FBGEMM
that referenced
this pull request
Mar 6, 2025
Summary: X-link: pytorch/torchrec#2751 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap, nautsimon Differential Revision: D68055168
spcyppt
force-pushed
the
export-D68055168
branch
from
525db0b to
ce34aac
Compare
spcyppt
added a commit
to spcyppt/torchrec
that referenced
this pull request
Mar 6, 2025
Summary: Pull Request resolved: pytorch#2751 X-link: pytorch/FBGEMM#3711 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap, nautsimon Differential Revision: D68055168
|
This pull request was exported from Phabricator. Differential Revision: D68055168 |
spcyppt
added a commit
to spcyppt/FBGEMM
that referenced
this pull request
Mar 6, 2025
Summary: X-link: pytorch/torchrec#2751 Pull Request resolved: pytorch#3711 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap, nautsimon Differential Revision: D68055168
spcyppt
force-pushed
the
export-D68055168
branch
from
ce34aac to
8b221be
Compare
|
This pull request was exported from Phabricator. Differential Revision: D68055168 |
spcyppt
added a commit
to spcyppt/FBGEMM
that referenced
this pull request
Mar 6, 2025
Summary: X-link: pytorch/torchrec#2751 Pull Request resolved: pytorch#3711 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap, nautsimon Differential Revision: D68055168
spcyppt
force-pushed
the
export-D68055168
branch
from
8b221be to
de5ca5c
Compare
spcyppt
added a commit
to spcyppt/torchrec
that referenced
this pull request
Mar 6, 2025
Summary: Pull Request resolved: pytorch#2751 X-link: pytorch/FBGEMM#3711 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap, nautsimon Differential Revision: D68055168
spcyppt
added a commit
to spcyppt/torchrec
that referenced
this pull request
Mar 6, 2025
Summary: X-link: pytorch/FBGEMM#3711 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap, nautsimon Differential Revision: D68055168
spcyppt
force-pushed
the
export-D68055168
branch
from
de5ca5c to
5114484
Compare
spcyppt
added a commit
to spcyppt/FBGEMM
that referenced
this pull request
Mar 6, 2025
Summary: X-link: pytorch/torchrec#2751 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap, nautsimon Differential Revision: D68055168
|
This pull request was exported from Phabricator. Differential Revision: D68055168 |
spcyppt
added a commit
to spcyppt/FBGEMM
that referenced
this pull request
Mar 6, 2025
Summary: X-link: pytorch/torchrec#2751 Pull Request resolved: pytorch#3711 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap, nautsimon Differential Revision: D68055168
spcyppt
force-pushed
the
export-D68055168
branch
from
5114484 to
b95d848
Compare
spcyppt
added a commit
to spcyppt/torchrec
that referenced
this pull request
Mar 6, 2025
Summary: Pull Request resolved: pytorch#2751 X-link: pytorch/FBGEMM#3711 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap, nautsimon Differential Revision: D68055168
Summary: X-link: pytorch/torchrec#2751 Pull Request resolved: pytorch#3711 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap, nautsimon Differential Revision: D68055168
|
This pull request was exported from Phabricator. Differential Revision: D68055168 |
spcyppt
force-pushed
the
export-D68055168
branch
from
b95d848 to
415e284
Compare
spcyppt
added a commit
to spcyppt/torchrec
that referenced
this pull request
Mar 6, 2025
Summary: Pull Request resolved: pytorch#2751 X-link: pytorch/FBGEMM#3711 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap, nautsimon Differential Revision: D68055168
facebook-github-bot
pushed a commit
to pytorch/torchrec
that referenced
this pull request
Mar 7, 2025
Summary: Pull Request resolved: #2751 X-link: pytorch/FBGEMM#3711 X-link: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap, nautsimon Differential Revision: D68055168 fbshipit-source-id: 80b8dcdd8c2996fedbf26f4fb057abcded760b8d
|
This pull request has been merged in 1dab53e. |
q10
pushed a commit
to q10/FBGEMM
that referenced
this pull request
Apr 10, 2025
Summary: X-link: pytorch/torchrec#2751 X-link: pytorch#3711 Pull Request resolved: facebookresearch/FBGEMM#793 **Backend**: D68054868 --- As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration. **For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.** We pack the arguments as a list for each type. For **common** arguments, we pack - weights and arguments of type `Momentum` into TensorList - other tensors and optional tensors to list of optional tensors `aux_tensor` - `int` arguments into `aux_int` - `float` arguments into `aux_float` - `bool` arguments into `aux_bool`. Similarly for **optimizer-specific** arguments, we pack - arguments of type `Momentum` that are *__not__ optional* into TensorList - *optional* tensors to list of optional tensors `optim_tensor` - `int` arguments into `optim_int` - `float` arguments into `optim_float` - `bool` arguments into `optim_bool`. We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually. **This significantly reduces number of arguments.** For example, `split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function`, which currently has 61 arguments only have 26 arguments with this API design. Please refer to the design doc on which arguments are packed and signature. Design doc: https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb Full signature for each optimizer lookup function will be provided shortly. Reviewed By: sryap, nautsimon Differential Revision: D68055168 fbshipit-source-id: 80b8dcdd8c2996fedbf26f4fb057abcded760b8d
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.
Summary:
X-link: https://github.com/facebookresearch/FBGEMM/pull/793
Backend: D68054868
As the number of arguments in TBE keeps growing, some of the optimizers run into number of arguments limitation (i.e., 64) during pytorch operation registration.
For long-term growth and maintenance, we hence redesign TBE API by packing some of the arguments into list. Note that not all arguments are packed.
We pack the arguments as a list for each type.
For common arguments, we pack
Momentuminto TensorListaux_tensorintarguments intoaux_intfloatarguments intoaux_floatboolarguments intoaux_bool.Similarly for optimizer-specific arguments, we pack
Momentumthat are not optional into TensorListoptim_tensorintarguments intooptim_intfloatarguments intooptim_floatboolarguments intooptim_bool.We see issues with pytorch registration across packing SymInt in python-C++, so we unroll and pass SymInt arguments individually.
This significantly reduces number of arguments. For example,
split_embedding_codegen_lookup_rowwise_adagrad_with_counter_function, which currently has 61 arguments only have 26 arguments with this API design.Please refer to the design doc on which arguments are packed and signature.
Design doc:
https://docs.google.com/document/d/1dCBg7dcf7Yq9FHVrvXsAmFtBxkDi9o6u0r-Ptd4UDPE/edit?tab=t.0#heading=h.6bip5pwqq8xb
Full signature for each optimizer lookup function will be provided shortly.
Reviewed By: sryap
Differential Revision: D68055168