[PyTorch TB] Write raw tensor as tensor_proto #104908

sethlee0111 · 2023-07-10T21:28:09Z

Stack from ghstack (oldest at bottom):

This is the first diff to support logging of raw tensors for TensorBoard Intermediate Logging

Ultimately, we aim to support the feature where store full tensor is stored as a tensor protobuf to TB. Protobuf contains shape, dtype, and elements of the given tensor.

add tensor_proto() to summary.py which takes a tensor and convert to protobuf
add add_tensor() to writer.py
formatting changes introduced by arc lint

Differential Revision: D47302017

This is the first diff to support logging of raw tensors for [TensorBoard Intermediate Logging](https://www.internalfb.com/intern/wiki/TensorBoard/Intermediate_Logging/) Ultimately, we aim to support the feature where store full tensor is stored as a tensor protobuf to TB. Protobuf contains shape, dtype, and elements of the given tensor. 1. add `tensor_proto()` to `summary.py` which takes a tensor and convert to protobuf 2. add `add_tensor()` to `writer.py` 3. formatting changes introduced by `arc lint` ------------- ghstack-source-id: 194155757 Differential Revision: [D47302017](https://our.internmc.facebook.com/intern/diff/D47302017/) [ghstack-poisoned]

pytorch-bot · 2023-07-10T21:28:12Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/104908

📄 Preview Python docs built from this PR
📄 Preview C++ docs built from this PR
❓ Need help or want to give feedback on the CI? Visit the bot commands wiki or our office hours

Note: Links to docs will display an error until the docs builds have been completed.

❌ 2 New Failures

As of commit 88e722f:

NEW FAILURES - The following jobs have failed:

This comment was automatically generated by Dr. CI and updates every 15 minutes.

vadimkantorov · 2023-07-10T21:32:46Z

torch/utils/tensorboard/summary.py

+            ],
+            proto_val_field: torch.view_as_real(tensor).tolist()
+            if proto_val_field == "scomplex_val" or proto_val_field == "dcomplex_val"
+            else tensor.reshape(-1).tolist(),


here a bit more idiomatic and modern might be .flatten() instead of reshape(-1)

also, the two branches are currently not equivalent. if multidim tensors can be accepted here, the view_as_real(...).tolist() can produce nested lists

Also beware that tensor.reshape(-1).tolist() can produce python scalar instead of the list for numel() == 1 inputs: #52262 - my proposal for that one is to introduce a tolist(force = True) which would always return a list, even for scalars (so .numel() == 1 and .numel() == 0 probably should be checked)

Updated, thanks!

Added checks for numel() == 1 and numel() == 0

This is the first diff to support logging of raw tensors for [TensorBoard Intermediate Logging](https://www.internalfb.com/intern/wiki/TensorBoard/Intermediate_Logging/) Ultimately, we aim to support the feature where store full tensor is stored as a tensor protobuf to TB. Protobuf contains shape, dtype, and elements of the given tensor. 1. add `tensor_proto()` to `summary.py` which takes a tensor and convert to protobuf 2. add `add_tensor()` to `writer.py` 3. formatting changes introduced by `arc lint` ------------- Differential Revision: [D47302017](https://our.internmc.facebook.com/intern/diff/D47302017/) [ghstack-poisoned]

Pull Request resolved: #104908 This diff applies changes to open-source PyTorch to support logging of raw tensors for [TensorBoard Intermediate Logging](https://www.internalfb.com/intern/wiki/TensorBoard/Intermediate_Logging/) Ultimately, we aim to support the feature where store full tensor is stored as a tensor protobuf to TB. Protobuf contains shape, dtype, and elements of the given tensor. 1. add `tensor_proto()` to `summary.py` which takes a tensor and convert to protobuf 2. add `add_tensor()` to `writer.py` 3. formatting changes introduced by `arc lint` ------------- ghstack-source-id: 194300671 ghstack-source-id: 194300671 Differential Revision: [D47357048](https://our.internmc.facebook.com/intern/diff/D47357048/)

This is the first diff to support logging of raw tensors for [TensorBoard Intermediate Logging](https://www.internalfb.com/intern/wiki/TensorBoard/Intermediate_Logging/) Ultimately, we aim to support the feature where store full tensor is stored as a tensor protobuf to TB. Protobuf contains shape, dtype, and elements of the given tensor. 1. add `tensor_proto()` to `summary.py` which takes a tensor and convert to protobuf 2. add `add_tensor()` to `writer.py` 3. formatting changes introduced by `arc lint` ------------- Differential Revision: [D47302017](https://our.internmc.facebook.com/intern/diff/D47302017/) [ghstack-poisoned]

kunalb · 2023-07-12T03:02:50Z

torch/utils/tensorboard/summary.py

+        )
+
+    # type maps: torch.Tensor type -> (protobuf type, protobuf val field)
+    type_map = {


make this a private constant map outside the function? _TENSOR_TYPE_MAP = ... or something?

Makes sense, applied changes.

kunalb

Generally looks good to me, thanks for adding tests! some nits inline

kunalb · 2023-07-12T03:05:13Z

torch/utils/tensorboard/summary.py

+
+    if tensor.dtype in type_map:
+        proto_val_field = type_map[tensor.dtype][1]
+        proto_val_contents = (


For readability, I'd still this make this explicit
proto_val_contents = []
if proto_val_field == ...:
proto_val_contents = ...
elif tensor.numel() == 1
proto_val_contents = ...

and so on

I see what you mean. Made changes.

This is the first diff to support logging of raw tensors for [TensorBoard Intermediate Logging](https://www.internalfb.com/intern/wiki/TensorBoard/Intermediate_Logging/) Ultimately, we aim to support the feature where store full tensor is stored as a tensor protobuf to TB. Protobuf contains shape, dtype, and elements of the given tensor. 1. add `tensor_proto()` to `summary.py` which takes a tensor and convert to protobuf 2. add `add_tensor()` to `writer.py` 3. formatting changes introduced by `arc lint` ------------- Differential Revision: [D47302017](https://our.internmc.facebook.com/intern/diff/D47302017/) [ghstack-poisoned]

Pull Request resolved: #104908 This diff applies changes to open-source PyTorch to support logging of raw tensors for [TensorBoard Intermediate Logging](https://www.internalfb.com/intern/wiki/TensorBoard/Intermediate_Logging/) Ultimately, we aim to support the feature where store full tensor is stored as a tensor protobuf to TB. Protobuf contains shape, dtype, and elements of the given tensor. 1. add `tensor_proto()` to `summary.py` which takes a tensor and convert to protobuf 2. add `add_tensor()` to `writer.py` 3. formatting changes introduced by `arc lint` ------------- ghstack-source-id: 194385501 ghstack-source-id: 194385501 Differential Revision: [D47357048](https://our.internmc.facebook.com/intern/diff/D47357048/)

facebook-github-bot · 2023-07-13T05:29:03Z

@pytorchbot merge

(Initiating merge automatically since Phabricator Diff has merged)

pytorchmergebot · 2023-07-13T05:30:44Z

Merge started

Your change will be merged once all checks pass (ETA 0-4 Hours).

Learn more about merging in the wiki.

Questions? Feedback? Please reach out to the PyTorch DevX Team

Advanced Debugging

Check the merge workflow status
here

atalman · 2023-07-13T15:03:55Z

@pytorchbot revert -c ignoredsignal -m "broke win-vs2019-cpu-py3 and macos-12-py3-arm64 "

pytorchmergebot · 2023-07-13T15:06:08Z

@pytorchbot successfully started a revert job. Check the current status here.
Questions? Feedback? Please reach out to the PyTorch DevX Team

pytorchmergebot · 2023-07-13T15:06:13Z

Can't revert PR that was landed via phabricator as D47302017. Please revert by going to the internal diff and clicking Unland.

facebook-github-bot · 2023-07-13T16:20:29Z

@pytorchbot revert -m="Diff reverted internally" -c="ghfirst"

This Pull Request has been reverted by a revert inside Meta. To re-land this change, please open another pull request, assign the same reviewers, fix the CI failures that caused the revert and make sure that the failing CI runs on the PR by applying the proper ciflow label (e.g., ciflow/trunk).)

pytorchmergebot · 2023-07-13T16:21:59Z

@pytorchbot successfully started a revert job. Check the current status here.
Questions? Feedback? Please reach out to the PyTorch DevX Team

pytorchmergebot · 2023-07-13T16:22:09Z

@sethlee0111 your PR has been successfully reverted.

This reverts commit dceae41. Reverted #104908 on behalf of https://github.com/facebook-github-bot due to Diff reverted internally ([comment](#104908 (comment)))

sethlee0111 mentioned this pull request Jul 10, 2023

Functions to support logging raw tensors on tensorboard #104881

Closed

vadimkantorov reviewed Jul 10, 2023

View reviewed changes

sethlee0111 mentioned this pull request Jul 11, 2023

[PyTorch TB] Refactor formatting #104978

Merged

kunalb reviewed Jul 12, 2023

View reviewed changes

kunalb approved these changes Jul 12, 2023

View reviewed changes

sethlee0111 mentioned this pull request Jul 12, 2023

[PyTorch TB] Refactor formatting #105027

Closed

pytorch-bot bot added the ciflow/trunk Trigger trunk jobs on your pull request label Jul 13, 2023

pytorchmergebot added the merging label Jul 13, 2023

pytorchmergebot added Merged and removed merging labels Jul 13, 2023

pytorchmergebot closed this in dceae41 Jul 13, 2023

pytorchmergebot added the Reverted label Jul 13, 2023

facebook-github-bot deleted the gh/sethlee0111/2/head branch July 16, 2023 14:17

seberg mentioned this pull request Oct 26, 2023

ENH: Create obvious way of getting a single 1-D list with all array elements numpy/numpy#24989

Open

[PyTorch TB] Write raw tensor as tensor_proto #104908

[PyTorch TB] Write raw tensor as tensor_proto #104908

Uh oh!

Conversation

sethlee0111 commented Jul 10, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pytorch-bot bot commented Jul 10, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/104908

❌ 2 New Failures

Uh oh!

vadimkantorov Jul 10, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

sethlee0111 Jul 10, 2023

Choose a reason for hiding this comment

Uh oh!

sethlee0111 Jul 10, 2023

Choose a reason for hiding this comment

Uh oh!

kunalb Jul 12, 2023

Choose a reason for hiding this comment

Uh oh!

sethlee0111 Jul 12, 2023

Choose a reason for hiding this comment

Uh oh!

kunalb left a comment

Choose a reason for hiding this comment

Uh oh!

kunalb Jul 12, 2023

Choose a reason for hiding this comment

Uh oh!

sethlee0111 Jul 12, 2023

Choose a reason for hiding this comment

Uh oh!

facebook-github-bot commented Jul 13, 2023

Uh oh!

pytorchmergebot commented Jul 13, 2023

Merge started

Uh oh!

atalman commented Jul 13, 2023

Uh oh!

pytorchmergebot commented Jul 13, 2023

Uh oh!

pytorchmergebot commented Jul 13, 2023

Uh oh!

facebook-github-bot commented Jul 13, 2023

Uh oh!

pytorchmergebot commented Jul 13, 2023

Uh oh!

pytorchmergebot commented Jul 13, 2023

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

sethlee0111 commented Jul 10, 2023 •

edited

Loading

pytorch-bot bot commented Jul 10, 2023 •

edited

Loading

vadimkantorov Jul 10, 2023 •

edited

Loading