Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Merge 'main' into 'llama' #26

Merged
merged 18 commits into from
Nov 12, 2024
Merged

Merge 'main' into 'llama' #26

merged 18 commits into from
Nov 12, 2024

Conversation

tanganke
Copy link
Owner

No description provided.

tanganke and others added 18 commits October 31, 2024 00:30
@tanganke
update docs
cd61266
@tanganke
udpate README.md
13c9458
@tanganke
new file: tests/README.md
e08c570
@tanganke
Add argument documentation to functions
82bdd3a 
Add missing argument explanations to various functions across multiple files.

* **`fusion_bench/compat/taskpool/clip_image_classification.py`**:
  - Add explanation for `clip_model` argument in `evaluate` function.
  - Add explanation for `model` argument in `evaluate` function.

* **`fusion_bench/compat/taskpool/flan_t5_glue_text_generation.py`**:
  - Add explanation for `model` argument in `evaluate` function.

* **`fusion_bench/dataset/gsm8k.py`**:
  - Add explanation for `dataset_name` argument in `load_gsm8k_question_label_data` function.

* **`fusion_bench/dataset/nyuv2.py`**:
  - Add explanation for `index` argument in `__getitem__` function.

* **`fusion_bench/method/adamerging/task_wise_adamerging.py`**:
  - Add explanation for `module`, `batch`, and `task` arguments in `compute_logits` function.

* **`fusion_bench/method/dawe/dawe_for_clip.py`**:
  - Add explanation for `pretrained_model_name_or_path` argument in `load_resnet_processor` function.

* **`fusion_bench/method/dummy.py`**:
  - Add explanation for `modelpool` argument in `run` function.

* **`fusion_bench/method/ensemble.py`**:
  - Add explanation for `modelpool` argument in `run` function.

* **`fusion_bench/method/linear/expo.py`**:
  - Add explanation for `modelpool` argument in `run` function.

* **`fusion_bench/method/linear/linear_interpolation.py`**:
  - Add explanation for `modelpool` argument in `run` function.

---

For more details, open the [Copilot Workspace session](https://copilot-workspace.githubnext.com/tanganke/fusion_bench?shareId=XXXX-XXXX-XXXX-XXXX).
@tanganke
Enhance Llama model Integration (#22)
16dda8e
@tanganke
Squashed commit of the following:
e35649a 
commit 160eeaf
Author: tanganke <[email protected]>
Date:   Fri Nov 1 20:29:05 2024 +0800

    	modified:   fusion_bench/taskpool/llama/test_generation.py

commit 574447f
Author: tanganke <[email protected]>
Date:   Fri Nov 1 19:41:52 2024 +0800

    update expo.py

commit 4e613a4
Author: tanganke <[email protected]>
Date:   Fri Nov 1 18:32:17 2024 +0800

    update task vector cos similarity visualization

commit 201c81b
Author: Anke Tang <[email protected]>
Date:   Fri Nov 1 15:42:34 2024 +0800

    update docstring

commit 50b5974
Author: Anke Tang <[email protected]>
Date:   Fri Nov 1 15:41:09 2024 +0800

    fix bugs

commit 92548f6
Author: Anke Tang <[email protected]>
Date:   Fri Nov 1 15:34:51 2024 +0800

    update docs and test generation for llama taskpool

commit 7a4ec5d
Author: Anke Tang <[email protected]>
Date:   Fri Nov 1 15:27:04 2024 +0800

    update LlamaTestGenerationTaskPool

commit 04048f5
Author: Anke Tang <[email protected]>
Date:   Fri Nov 1 15:25:53 2024 +0800

    add LlamaTestGenerationTaskPool

commit f22ae6d
Author: tanganke <[email protected]>
Date:   Fri Nov 1 12:14:02 2024 +0800

    update functions related to llama models
    	new file:   fusion_bench/dataset/llama/openai.py
    	modified:   fusion_bench/modelpool/base_pool.py
    	modified:   fusion_bench/modelpool/causal_lm/causal_lm.py
    	new file:   fusion_bench/taskpool/llama/__init__.py
    	new file:   fusion_bench/taskpool/llama/test_generation.py

commit 378efd2
Author: tanganke <[email protected]>
Date:   Fri Nov 1 09:15:58 2024 +0800

    update
    	modified:   config/method/adamerging/llama_sft.yaml
    	modified:   fusion_bench/dataset/llama/alpaca.py

commit ea0cfa5
Author: tanganke <[email protected]>
Date:   Thu Oct 31 21:07:35 2024 +0800

    update

commit 5d3e2f8
Author: tanganke <[email protected]>
Date:   Thu Oct 31 20:57:44 2024 +0800

    remove unused files

commit aacdc97
Author: tanganke <[email protected]>
Date:   Thu Oct 31 20:56:00 2024 +0800

    update sharegpt.py

commit d0acaaa
Author: tanganke <[email protected]>
Date:   Thu Oct 31 20:26:00 2024 +0800

    add sharegpt.py

commit 37e1ab3
Author: tanganke <[email protected]>
Date:   Thu Oct 31 18:27:07 2024 +0800

    	new file:   fusion_bench/dataset/llama/alpaca.py

commit 67688c3
Author: tanganke <[email protected]>
Date:   Thu Oct 31 17:38:33 2024 +0800

    rename squad.py and wikitext.py

commit c53288c
Author: tanganke <[email protected]>
Date:   Thu Oct 31 17:35:28 2024 +0800

    add AdamW mixin

commit 2ec72e4
Author: tanganke <[email protected]>
Date:   Thu Oct 31 17:17:01 2024 +0800

    Add squad.py, update docs

commit aa3b318
Author: tanganke <[email protected]>
Date:   Thu Oct 31 16:03:40 2024 +0800

    optimize task_vector_violin_plot

commit 8a5806f
Author: tanganke <[email protected]>
Date:   Thu Oct 31 15:49:24 2024 +0800

    add TaskVectorViolinPlot
    	new file:   config/method/analysis/task_vector_violin_plot.yaml
    	new file:   fusion_bench/method/analysis/task_vector_violin_plot.py

commit 418a802
Author: tanganke <[email protected]>
Date:   Thu Oct 31 14:30:10 2024 +0800

    fix bugs

commit 855460a
Merge: 18717f0 b37a63a
Author: tanganke <[email protected]>
Date:   Thu Oct 31 14:12:07 2024 +0800

    Merge branch 'llama' of https://github.com/tanganke/fusion_bench into llama

commit b37a63a
Author: Anke Tang <[email protected]>
Date:   Thu Oct 31 11:34:58 2024 +0800

    Update llama_sft.yaml, llama_adamerging.py, layer_wise_fusion.py.

commit e306e38
Author: Anke Tang <[email protected]>
Date:   Thu Oct 31 11:10:23 2024 +0800

    refactor: rename BaseModelFusionAlgorithm to BaseAlgorithm

    **Key changes**:
    - Renamed BaseModelFusionAlgorithm to BaseAlgorithm for better clarity and consistency
    - Updated all references to BaseModelFusionAlgorithm across the codebase
    - Added BaseModelAlgorithm as an alias for BaseAlgorithm for backward compatibility
    - Added/improved docstrings for models/wrappers/ensemble.py
    - No functional changes - purely refactoring of class names

commit 8268d7a
Author: tanganke <[email protected]>
Date:   Wed Oct 30 20:50:39 2024 +0800

    update code

commit 18717f0
Author: tanganke <[email protected]>
Date:   Wed Oct 30 20:50:39 2024 +0800

    update code

commit be0442e
Author: tanganke <[email protected]>
Date:   Wed Oct 30 20:20:56 2024 +0800

    update config

commit 46ae4b6
Author: tanganke <[email protected]>
Date:   Wed Oct 30 20:20:56 2024 +0800

    update config

commit 3538534
Author: tanganke <[email protected]>
Date:   Wed Oct 30 20:16:55 2024 +0800

    update llama_sft adamerging

commit a4a6c89
Author: tanganke <[email protected]>
Date:   Wed Oct 30 20:16:55 2024 +0800

    update llama_sft adamerging

commit c68b7d2
Author: tanganke <[email protected]>
Date:   Wed Oct 30 14:52:22 2024 +0800

    update llama adamerging

commit a58cb45
Author: tanganke <[email protected]>
Date:   Wed Oct 30 10:49:37 2024 +0800

    update data for llama models

commit ccfc3c2
Author: tanganke <[email protected]>
Date:   Wed Oct 30 14:52:22 2024 +0800

    update llama adamerging

commit bcefd49
Author: tanganke <[email protected]>
Date:   Wed Oct 30 10:49:37 2024 +0800

    update data for llama models

commit a66f396
Author: tanganke <[email protected]>
Date:   Wed Oct 30 10:07:00 2024 +0800

    update data utils for llama models

commit d0b0884
Author: tanganke <[email protected]>
Date:   Wed Oct 30 09:54:05 2024 +0800

    add collator and template

commit d722c1a
Author: tanganke <[email protected]>
Date:   Wed Oct 30 10:07:00 2024 +0800

    update data utils for llama models

commit 9bad0eb
Author: tanganke <[email protected]>
Date:   Wed Oct 30 09:54:05 2024 +0800

    add collator and template

commit 71a9291
Author: Anke Tang <[email protected]>
Date:   Wed Oct 30 00:39:26 2024 +0800

    Update utils modified: fusion_bench/utils/devices.py modified: fusion_bench/utils/parameters.py

commit dcdeb1b
Author: Anke Tang <[email protected]>
Date:   Wed Oct 30 00:05:59 2024 +0800

    Enhance LLaMA model integration

commit d09ca14
Author: Anke Tang <[email protected]>
Date:   Wed Oct 30 00:39:26 2024 +0800

    Update utils
    	modified:   fusion_bench/utils/devices.py
    	modified:   fusion_bench/utils/parameters.py

commit a4cc970
Author: Anke Tang <[email protected]>
Date:   Wed Oct 30 00:05:59 2024 +0800

    Enhance LLaMA model integration

commit 68ee6be
Author: tanganke <[email protected]>
Date:   Tue Oct 29 21:38:08 2024 +0800

    partial update

commit 72c21bb
Author: tanganke <[email protected]>
Date:   Tue Oct 29 21:38:08 2024 +0800

    partial update

commit 7026685
Author: tanganke <[email protected]>
Date:   Tue Oct 29 20:13:32 2024 +0800

    add Dare simple_average
@hetailang @tanganke
fix calculate the average accuracy and loss in fusion_bench/fusion_be…
bd62831 
…nch/taskpool/clip_vision/taskpool.py
@tanganke
version 0.2.3
b7fb0a7
@tanganke
Merge branch 'main' of github.com:tanganke/fusion_bench
a7a33af
@tanganke
update README.md
0c10565
@tanganke
modified: README.md
3698efb
@tanganke
fix dataloader bug in layer-wise adamerging
ec90d51
@tanganke
fix dataloader bug in layer-wise adamerging
f796870
@tanganke
Merge branch main of github.com:tanganke/fusion_bench into main
7a1b2f9
@tanganke
fix bug
42dff55
@tanganke
Merge branch 'llama'
5d14b16
@tanganke
Update pyproject.toml
a95ca16
@github-actions
Optimised images with calibre/image-actions
48e8b26
@github-actions GitHub Actions
Copy link
Contributor

Images automagically compressed by Calibre's image-actions ✨

Compression reduced images by 42.1%, saving 455.32 KB.

Filename Before After Improvement Visual comparison
docs/algorithms/images/ewemoe_1.png 307.40 KB 201.93 KB -34.3% View diff
docs/algorithms/images/ewemoe_2.png 353.25 KB 184.03 KB -47.9% View diff
docs/algorithms/images/ewemoe.png 420.68 KB 240.04 KB -42.9% View diff

62 images did not require optimisation.

@tanganke tanganke marked this pull request as ready for review November 12, 2024 02:27
@tanganke tanganke merged commit b946fbe into llama Nov 12, 2024
Copilot
Copilot AI reviewed Nov 13, 2024
View reviewed changes
Copy link
Contributor

@Copilot Copilot AI left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Copilot reviewed 10 out of 25 changed files in this pull request and generated 1 suggestion.

Files not reviewed (15)
  • .vscode/.gitignore: Language not supported
  • fusion_bench/dataset/gsm8k.py: Evaluated as low risk
  • fusion_bench/method/analysis/task_vector_cos_similarity.py: Evaluated as low risk
  • fusion_bench/method/linear/linear_interpolation.py: Evaluated as low risk
  • fusion_bench/method/dummy.py: Evaluated as low risk
  • fusion_bench/compat/taskpool/clip_image_classification.py: Evaluated as low risk
  • fusion_bench/dataset/nyuv2.py: Evaluated as low risk
  • fusion_bench/method/dawe/dawe_for_clip.py: Evaluated as low risk
  • fusion_bench/compat/taskpool/flan_t5_glue_text_generation.py: Evaluated as low risk
  • fusion_bench/method/adamerging/clip_layer_wise_adamerging.py: Evaluated as low risk
  • fusion_bench/method/linear/expo.py: Evaluated as low risk
  • fusion_bench/method/adamerging/task_wise_adamerging.py: Evaluated as low risk
  • fusion_bench/method/ensemble.py: Evaluated as low risk
  • fusion_bench/taskpool/clip_vision/taskpool.py: Evaluated as low risk
  • README.md: Evaluated as low risk

fusion_bench/mixins/clip_classification.py
Returns:
Iterator: An iterator over the shuffled test DataLoader.
"""
# get dataloader kwargs
Copy link
Preview

Copilot AI Nov 13, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The word 'dataloader' should be 'DataLoader' to maintain consistency.

Suggested change
# get dataloader kwargs
DataLoader_kwargs = self._dataloader_kwargs.copy()

Copilot is powered by AI, so mistakes are possible. Review output carefully before use.

Positive Feedback
Negative Feedback

Provide additional feedback

Please help us improve GitHub Copilot by sharing more details about this comment.

Please select one or more of the options
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

Copilot Copilot left review comments

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@tanganke @hetailang