add et export with gguf with test #245
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
facebook-github-bot
added
the
CLA Signed
mikekgfb
approved these changes
Apr 17, 2024
| model_to_pte = model | ||
| model_to_dso = model | ||
| else: | ||
| if output_pte_path: |
There was a problem hiding this comment.
This is very kludgy and I would prefer to export to int4 and then handle it from there. Basing front end decisions on backend is a very bad practice because we're going to end up in a world of hurt.
Kimish and I had discussed doing a transform from int4 ->a8w4dq. Right now we just get a de-quantized model.
Please plan to land that asap, Kimish?
cc: @kimishpatel
| with torch.no_grad(): | ||
| if output_pte_path: | ||
| output_pte_path = str(os.path.abspath(output_pte_path)) | ||
| print(f">{output_pte_path}<") | ||
| if executorch_export_available: | ||
| print(f"Exporting model using Executorch to {output_pte_path}") | ||
| export_model_et(model, builder_args.device, args.output_pte_path, args) | ||
| export_model_et(model_to_pte, builder_args.device, args.output_pte_path, args) |
There was a problem hiding this comment.
I don't like this at all :( But we're out of runway, so I will approve for now.
| @@ -68,7 +90,7 @@ def main(args): | |||
| if output_dso_path: | |||
| output_dso_path = str(os.path.abspath(output_dso_path)) | |||
| print(f"Exporting model using AOT Inductor to {output_dso_path}") | |||
| export_model_aoti(model, builder_args.device, output_dso_path, args) | |||
| export_model_aoti(model_to_dso, builder_args.device, output_dso_path, args) | |||
metascroy
force-pushed
the
gguf-export-ext
branch
from
f8884e6
to
bc92599
Compare
metascroy
force-pushed
the
gguf-export-ext
branch
from
bc92599
to
9563191
Compare
malfet
pushed a commit
that referenced
this pull request
Jul 17, 2024
* add et export with gguf with test * fix generate too * add gguf path to generate
malfet
pushed a commit
that referenced
this pull request
Jul 17, 2024
* add et export with gguf with test * fix generate too * add gguf path to generate
malfet
pushed a commit
that referenced
this pull request
Jul 17, 2024
* add et export with gguf with test * fix generate too * add gguf path to generate
malfet
pushed a commit
that referenced
this pull request
Jul 17, 2024
* add et export with gguf with test * fix generate too * add gguf path to generate
malfet
pushed a commit
that referenced
this pull request
Jul 17, 2024
* add et export with gguf with test * fix generate too * add gguf path to generate
malfet
pushed a commit
that referenced
this pull request
Jul 17, 2024
* add et export with gguf with test * fix generate too * add gguf path to generate
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.
ET does not support _weight_int4pack_mm, so this adds gguf_kwargs that can be passed to building that control whether GGUF should be load_as_quantized. If load_as_quantized=False, GGUF is converted to floating point.
Also adds test for torchchat export + generate to et.yml with gguf file.