├── .gitignore
├── LICENSE
├── README.md
├── assets
├── alpaca_main.jpg
├── alpaca_right_email.png
├── alpaca_right_llama.png
├── alpaca_wrong_42.png
├── alpaca_wrong_capital.png
├── logo.png
└── parse_analysis.png
├── configs
└── default_offload_opt_param.json
├── requirements.txt
├── train.py
└── utils.py
/.gitignore:
--------------------------------------------------------------------------------
1 | # Byte-compiled / optimized / DLL files
2 | __pycache__/
3 | *.py[cod]
4 | *$py.class
5 |
6 | # C extensions
7 | *.so
8 |
9 | # Distribution / packaging
10 | .Python
11 | build/
12 | develop-eggs/
13 | dist/
14 | downloads/
15 | eggs/
16 | .eggs/
17 | lib/
18 | lib64/
19 | parts/
20 | sdist/
21 | var/
22 | wheels/
23 | pip-wheel-metadata/
24 | share/python-wheels/
25 | *.egg-info/
26 | .installed.cfg
27 | *.egg
28 | MANIFEST
29 |
30 | # PyInstaller
31 | # Usually these files are written by a python script from a template
32 | # before PyInstaller builds the exe, so as to inject date/other infos into it.
33 | *.manifest
34 | *.spec
35 |
36 | # Installer logs
37 | pip-log.txt
38 | pip-delete-this-directory.txt
39 |
40 | # Unit test / coverage reports
41 | htmlcov/
42 | .tox/
43 | .nox/
44 | .coverage
45 | .coverage.*
46 | .cache
47 | nosetests.xml
48 | coverage.xml
49 | *.cover
50 | *.py,cover
51 | .hypothesis/
52 | .pytest_cache/
53 |
54 | # Translations
55 | *.mo
56 | *.pot
57 |
58 | # Django stuff:
59 | *.log
60 | local_settings.py
61 | db.sqlite3
62 | db.sqlite3-journal
63 |
64 | # Flask stuff:
65 | instance/
66 | .webassets-cache
67 |
68 | # Scrapy stuff:
69 | .scrapy
70 |
71 | # Sphinx documentation
72 | docs/_build/
73 |
74 | # PyBuilder
75 | target/
76 |
77 | # Jupyter Notebook
78 | .ipynb_checkpoints
79 |
80 | # IPython
81 | profile_default/
82 | ipython_config.py
83 |
84 | # pyenv
85 | .python-version
86 |
87 | # pipenv
88 | # According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
89 | # However, in case of collaboration, if having platform-specific dependencies or dependencies
90 | # having no cross-platform support, pipenv may install dependencies that don't work, or not
91 | # install all needed dependencies.
92 | #Pipfile.lock
93 |
94 | # PEP 582; used by e.g. github.com/David-OConnor/pyflow
95 | __pypackages__/
96 |
97 | # Celery stuff
98 | celerybeat-schedule
99 | celerybeat.pid
100 |
101 | # SageMath parsed files
102 | *.sage.py
103 |
104 | # Environments
105 | .env
106 | .venv
107 | env/
108 | venv/
109 | ENV/
110 | env.bak/
111 | venv.bak/
112 |
113 | # Spyder project settings
114 | .spyderproject
115 | .spyproject
116 |
117 | # Rope project settings
118 | .ropeproject
119 |
120 | # mkdocs documentation
121 | /site
122 |
123 | # mypy
124 | .mypy_cache/
125 | .dmypy.json
126 | dmypy.json
127 |
128 | # Pyre type checker
129 | .pyre/
130 |
131 | .DS_Store
132 | .idea
133 |
--------------------------------------------------------------------------------
/LICENSE:
--------------------------------------------------------------------------------
1 | Apache License
2 | Version 2.0, January 2004
3 | http://www.apache.org/licenses/
4 |
5 | TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
6 |
7 | 1. Definitions.
8 |
9 | "License" shall mean the terms and conditions for use, reproduction,
10 | and distribution as defined by Sections 1 through 9 of this document.
11 |
12 | "Licensor" shall mean the copyright owner or entity authorized by
13 | the copyright owner that is granting the License.
14 |
15 | "Legal Entity" shall mean the union of the acting entity and all
16 | other entities that control, are controlled by, or are under common
17 | control with that entity. For the purposes of this definition,
18 | "control" means (i) the power, direct or indirect, to cause the
19 | direction or management of such entity, whether by contract or
20 | otherwise, or (ii) ownership of fifty percent (50%) or more of the
21 | outstanding shares, or (iii) beneficial ownership of such entity.
22 |
23 | "You" (or "Your") shall mean an individual or Legal Entity
24 | exercising permissions granted by this License.
25 |
26 | "Source" form shall mean the preferred form for making modifications,
27 | including but not limited to software source code, documentation
28 | source, and configuration files.
29 |
30 | "Object" form shall mean any form resulting from mechanical
31 | transformation or translation of a Source form, including but
32 | not limited to compiled object code, generated documentation,
33 | and conversions to other media types.
34 |
35 | "Work" shall mean the work of authorship, whether in Source or
36 | Object form, made available under the License, as indicated by a
37 | copyright notice that is included in or attached to the work
38 | (an example is provided in the Appendix below).
39 |
40 | "Derivative Works" shall mean any work, whether in Source or Object
41 | form, that is based on (or derived from) the Work and for which the
42 | editorial revisions, annotations, elaborations, or other modifications
43 | represent, as a whole, an original work of authorship. For the purposes
44 | of this License, Derivative Works shall not include works that remain
45 | separable from, or merely link (or bind by name) to the interfaces of,
46 | the Work and Derivative Works thereof.
47 |
48 | "Contribution" shall mean any work of authorship, including
49 | the original version of the Work and any modifications or additions
50 | to that Work or Derivative Works thereof, that is intentionally
51 | submitted to Licensor for inclusion in the Work by the copyright owner
52 | or by an individual or Legal Entity authorized to submit on behalf of
53 | the copyright owner. For the purposes of this definition, "submitted"
54 | means any form of electronic, verbal, or written communication sent
55 | to the Licensor or its representatives, including but not limited to
56 | communication on electronic mailing lists, source code control systems,
57 | and issue tracking systems that are managed by, or on behalf of, the
58 | Licensor for the purpose of discussing and improving the Work, but
59 | excluding communication that is conspicuously marked or otherwise
60 | designated in writing by the copyright owner as "Not a Contribution."
61 |
62 | "Contributor" shall mean Licensor and any individual or Legal Entity
63 | on behalf of whom a Contribution has been received by Licensor and
64 | subsequently incorporated within the Work.
65 |
66 | 2. Grant of Copyright License. Subject to the terms and conditions of
67 | this License, each Contributor hereby grants to You a perpetual,
68 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable
69 | copyright license to reproduce, prepare Derivative Works of,
70 | publicly display, publicly perform, sublicense, and distribute the
71 | Work and such Derivative Works in Source or Object form.
72 |
73 | 3. Grant of Patent License. Subject to the terms and conditions of
74 | this License, each Contributor hereby grants to You a perpetual,
75 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable
76 | (except as stated in this section) patent license to make, have made,
77 | use, offer to sell, sell, import, and otherwise transfer the Work,
78 | where such license applies only to those patent claims licensable
79 | by such Contributor that are necessarily infringed by their
80 | Contribution(s) alone or by combination of their Contribution(s)
81 | with the Work to which such Contribution(s) was submitted. If You
82 | institute patent litigation against any entity (including a
83 | cross-claim or counterclaim in a lawsuit) alleging that the Work
84 | or a Contribution incorporated within the Work constitutes direct
85 | or contributory patent infringement, then any patent licenses
86 | granted to You under this License for that Work shall terminate
87 | as of the date such litigation is filed.
88 |
89 | 4. Redistribution. You may reproduce and distribute copies of the
90 | Work or Derivative Works thereof in any medium, with or without
91 | modifications, and in Source or Object form, provided that You
92 | meet the following conditions:
93 |
94 | (a) You must give any other recipients of the Work or
95 | Derivative Works a copy of this License; and
96 |
97 | (b) You must cause any modified files to carry prominent notices
98 | stating that You changed the files; and
99 |
100 | (c) You must retain, in the Source form of any Derivative Works
101 | that You distribute, all copyright, patent, trademark, and
102 | attribution notices from the Source form of the Work,
103 | excluding those notices that do not pertain to any part of
104 | the Derivative Works; and
105 |
106 | (d) If the Work includes a "NOTICE" text file as part of its
107 | distribution, then any Derivative Works that You distribute must
108 | include a readable copy of the attribution notices contained
109 | within such NOTICE file, excluding those notices that do not
110 | pertain to any part of the Derivative Works, in at least one
111 | of the following places: within a NOTICE text file distributed
112 | as part of the Derivative Works; within the Source form or
113 | documentation, if provided along with the Derivative Works; or,
114 | within a display generated by the Derivative Works, if and
115 | wherever such third-party notices normally appear. The contents
116 | of the NOTICE file are for informational purposes only and
117 | do not modify the License. You may add Your own attribution
118 | notices within Derivative Works that You distribute, alongside
119 | or as an addendum to the NOTICE text from the Work, provided
120 | that such additional attribution notices cannot be construed
121 | as modifying the License.
122 |
123 | You may add Your own copyright statement to Your modifications and
124 | may provide additional or different license terms and conditions
125 | for use, reproduction, or distribution of Your modifications, or
126 | for any such Derivative Works as a whole, provided Your use,
127 | reproduction, and distribution of the Work otherwise complies with
128 | the conditions stated in this License.
129 |
130 | 5. Submission of Contributions. Unless You explicitly state otherwise,
131 | any Contribution intentionally submitted for inclusion in the Work
132 | by You to the Licensor shall be under the terms and conditions of
133 | this License, without any additional terms or conditions.
134 | Notwithstanding the above, nothing herein shall supersede or modify
135 | the terms of any separate license agreement you may have executed
136 | with Licensor regarding such Contributions.
137 |
138 | 6. Trademarks. This License does not grant permission to use the trade
139 | names, trademarks, service marks, or product names of the Licensor,
140 | except as required for reasonable and customary use in describing the
141 | origin of the Work and reproducing the content of the NOTICE file.
142 |
143 | 7. Disclaimer of Warranty. Unless required by applicable law or
144 | agreed to in writing, Licensor provides the Work (and each
145 | Contributor provides its Contributions) on an "AS IS" BASIS,
146 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
147 | implied, including, without limitation, any warranties or conditions
148 | of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
149 | PARTICULAR PURPOSE. You are solely responsible for determining the
150 | appropriateness of using or redistributing the Work and assume any
151 | risks associated with Your exercise of permissions under this License.
152 |
153 | 8. Limitation of Liability. In no event and under no legal theory,
154 | whether in tort (including negligence), contract, or otherwise,
155 | unless required by applicable law (such as deliberate and grossly
156 | negligent acts) or agreed to in writing, shall any Contributor be
157 | liable to You for damages, including any direct, indirect, special,
158 | incidental, or consequential damages of any character arising as a
159 | result of this License or out of the use or inability to use the
160 | Work (including but not limited to damages for loss of goodwill,
161 | work stoppage, computer failure or malfunction, or any and all
162 | other commercial damages or losses), even if such Contributor
163 | has been advised of the possibility of such damages.
164 |
165 | 9. Accepting Warranty or Additional Liability. While redistributing
166 | the Work or Derivative Works thereof, You may choose to offer,
167 | and charge a fee for, acceptance of support, warranty, indemnity,
168 | or other liability obligations and/or rights consistent with this
169 | License. However, in accepting such obligations, You may act only
170 | on Your own behalf and on Your sole responsibility, not on behalf
171 | of any other Contributor, and only if You agree to indemnify,
172 | defend, and hold each Contributor harmless for any liability
173 | incurred by, or claims asserted against, such Contributor by reason
174 | of your accepting any such warranty or additional liability.
175 |
176 | END OF TERMS AND CONDITIONS
177 |
178 | APPENDIX: How to apply the Apache License to your work.
179 |
180 | To apply the Apache License to your work, attach the following
181 | boilerplate notice, with the fields enclosed by brackets "[]"
182 | replaced with your own identifying information. (Don't include
183 | the brackets!) The text should be enclosed in the appropriate
184 | comment syntax for the file format. We also recommend that a
185 | file or class name and description of purpose be included on the
186 | same "printed page" as the copyright notice for easier
187 | identification within third-party archives.
188 |
189 | Copyright 2023 Rohan Taori, Ishaan Gulrajani, Tianyi Zhang, Yann Dubois, Xuechen Li
190 |
191 | Licensed under the Apache License, Version 2.0 (the "License");
192 | you may not use this file except in compliance with the License.
193 | You may obtain a copy of the License at
194 |
195 | http://www.apache.org/licenses/LICENSE-2.0
196 |
197 | Unless required by applicable law or agreed to in writing, software
198 | distributed under the License is distributed on an "AS IS" BASIS,
199 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
200 | See the License for the specific language governing permissions and
201 | limitations under the License.
202 |
--------------------------------------------------------------------------------
/README.md:
--------------------------------------------------------------------------------
1 |
2 |
3 |
4 |
5 |
6 | # Stanford Alpaca Trainer - Updated for use training Replit's Code Model
7 |
8 | [](https://github.com/tatsu-lab/stanford_alpaca/blob/main/LICENSE)
9 | [](https://www.python.org/downloads/release/python-390/)
10 | [](https://github.com/psf/black)
11 |
12 | This was the repo for the Stanford Alpaca project, which is edited to become a trainer for Alpaca-format datasets over Replit's 3B Code Model:
13 |
14 | - The Base Model: [Replit 3B Code](https://huggingface.co/replit/replit-code-v1-3b)
15 |
16 | - The code for [fine-tuning the model](#fine-tuning).
17 |
18 |
19 | ## Overview
20 |
21 | A trainer for Replit's 3B parameter code model.
22 |
23 | ## Dataset Format
24 |
25 | Alpaca format datasets should be in the following format, in json:
26 |
27 | - `instruction`: `str`, describes the task the model should perform. Each of the 52K instructions is unique.
28 | - `input`: `str`, optional context or input for the task. For example, when the instruction is "Summarize the following article", the input is the article. Around 40% of the examples have an input.
29 | - `output`: `str`, the answer to the instruction as generated by `text-davinci-003`.
30 |
31 | ### Here is an example of a dataset:
32 | ```
33 | [
34 | {
35 | "instruction": "Give three tips for staying healthy.",
36 | "input": "",
37 | "output": "1. Eat a balanced and nutritious diet: Make sure your meals are inclusive of a variety of fruits and vegetables, lean protein, whole grains, and healthy fats. This helps to provide your body with the essential nutrients to function at its best and can help prevent chronic diseases.\n\n2. Engage in regular physical activity: Exercise is crucial for maintaining strong bones, muscles, and cardiovascular health. Aim for at least 150 minutes of moderate aerobic exercise or 75 minutes of vigorous exercise each week.\n\n3. Get enough sleep: Getting enough quality sleep is crucial for physical and mental well-being. It helps to regulate mood, improve cognitive function, and supports healthy growth and immune function. Aim for 7-9 hours of sleep each night."
38 | },
39 | {
40 | "instruction": "What are the three primary colors?",
41 | "input": "",
42 | "output": "The three primary colors are red, blue, and yellow. These colors are called primary because they cannot be created by mixing other colors and all other colors can be made by combining them in various proportions. In the additive color system, used for light, the primary colors are red, green, and blue (RGB)."
43 | },
44 | ]
45 | ```
46 |
47 | We used the following prompts for fine-tuning the Replit model:
48 |
49 | - for examples with a non-empty input field:
50 |
51 | ```
52 | ### Instruction:
53 | {instruction}
54 |
55 | ### Input:
56 | {input}
57 |
58 | ### Response:
59 | ```
60 |
61 | - for examples with an empty input field:
62 |
63 | ```
64 | ### Instruction:
65 | {instruction}
66 |
67 | ### Response:
68 | ```
69 |
70 | ## Fine-tuning
71 |
72 | To fine-tune for Replit's model, first install the requirements
73 |
74 | ```bash
75 | pip install -r requirements.txt
76 | ```
77 |
78 | The train.py script defaults to 2000 sequence length for training. It runs in small batch size at this sequence length on an a100 80gb. You will save a significant amount of vram, and thus, can train faster, with a smaller sequence length. Training on 2x a100 80gb with what is possible with 2000 token sequence length takes about 2.5 hours, with 512 token length, only 45~ minutes.
79 |
80 | Below is a command that fine-tunes Replit-3B with an alpaca-formated dataset on a machine with 2 A100 80G GPUs with 2000 token sequence length.
81 |
82 | Replace `` with a port of your own, `` with the path to your converted checkpoint and tokenizer or leave default for Replit's base code model, and `` with where you want to store your outputs.
83 |
84 | ```bash
85 | torchrun --nproc_per_node=2 --master_port= train.py \
86 | --model_name_or_path \
87 | --data_path ./.json \
88 | --bf16 True \
89 | --output_dir \
90 | --num_train_epochs 3 \
91 | --per_device_train_batch_size 1 \
92 | --gradient_accumulation_steps 4 \
93 | --evaluation_strategy "no" \
94 | --save_strategy "steps" \
95 | --save_steps 50 \
96 | --save_total_limit 2 \
97 | --learning_rate 2e-5 \
98 | --weight_decay 0. \
99 | --warmup_ratio 0.03 \
100 | --lr_scheduler_type "cosine" \
101 | --logging_steps 1 \
102 | ```
103 |
104 | Note the given training script is meant to be simple and easy to use, and is not particularly optimized.
105 | To run on more gpus, you may prefer to turn down `gradient_accumulation_steps` to keep a global batch size of 128. Global batch size has not been tested for optimality.
106 |
107 | ### Addressing OOM
108 |
109 | Naively, fine-tuning a 7B model requires about 7 x 4 x 4 = 112 GB of VRAM. Commands given above enable parameter sharding, so no redundant model copy is stored on any GPU.
110 | If you'd like to further reduce the memory footprint, here are some options:
111 |
112 | - Turn on CPU offload for FSDP with `--fsdp "full_shard auto_wrap offload"`. This saves VRAM at the cost of longer runtime.
113 | - In our experience, DeepSpeed stage-3 (with offload) can at times be more memory efficient than FSDP with offload. Here's an example to use DeepSpeed stage-3 with 4 GPUs with both parameter and optimizer offload:
114 | ```bash
115 | pip install deepspeed
116 | torchrun --nproc_per_node=4 --master_port= train.py \
117 | --model_name_or_path \
118 | --data_path ./alpaca_data.json \
119 | --bf16 True \
120 | --output_dir \
121 | --num_train_epochs 3 \
122 | --per_device_train_batch_size 4 \
123 | --per_device_eval_batch_size 4 \
124 | --gradient_accumulation_steps 8 \
125 | --evaluation_strategy "no" \
126 | --save_strategy "steps" \
127 | --save_steps 2000 \
128 | --save_total_limit 1 \
129 | --learning_rate 2e-5 \
130 | --weight_decay 0. \
131 | --warmup_ratio 0.03 \
132 | --deepspeed "./configs/default_offload_opt_param.json" \
133 | --tf32 True
134 | ```
135 | - The DeepSpeed library also provides some [helpful functions](https://deepspeed.readthedocs.io/en/latest/memory.html) to estimate memory usage.
136 | - [LoRA](https://arxiv.org/abs/2106.09685) fine-tunes low-rank slices of the query, key, and value embedding heads. This can reduce the total memory footprint from 112GB to about 7x4=28GB. We may release our re-implemention of this in the future, but for now the [peft](https://github.com/huggingface/peft) codebase can be a useful resource.
137 |
138 | ### Original Authors of the Alpaca paper
139 |
140 | All grad students below contributed equally and the order is determined by random draw.
141 |
142 | - [Rohan Taori](https://www.rohantaori.com/)
143 | - [Ishaan Gulrajani](https://ishaan.io/)
144 | - [Tianyi Zhang](https://tiiiger.github.io/)
145 | - [Yann Dubois](https://yanndubs.github.io/)
146 | - [Xuechen Li](https://www.lxuechen.com/)
147 |
148 | All advised by [Tatsunori B. Hashimoto](https://thashim.github.io/). Yann is also advised by [Percy Liang](https://cs.stanford.edu/~pliang/) and Xuechen is also advised by [Carlos Guestrin](https://guestrin.su.domains/).
149 |
150 | ### Citation
151 |
152 | Please cite the repo if you use the data or code in this repo.
153 |
154 | ```
155 | @misc{alpaca,
156 | author = {Rohan Taori and Ishaan Gulrajani and Tianyi Zhang and Yann Dubois and Xuechen Li and Carlos Guestrin and Percy Liang and Tatsunori B. Hashimoto },
157 | title = {Stanford Alpaca: An Instruction-following LLaMA model},
158 | year = {2023},
159 | publisher = {GitHub},
160 | journal = {GitHub repository},
161 | howpublished = {\url{https://github.com/tatsu-lab/stanford_alpaca}},
162 | }
163 | ```
164 |
165 | Naturally, you should also cite the original LLaMA paper [1] and the Self-Instruct paper [2].
166 |
167 | ### Acknowledgements
168 |
169 | We thank Yizhong Wang for his help in explaining the data generation pipeline in Self-Instruct and providing the code for the parse analysis plot.
170 | We thank Yifan Mai for helpful support, and members of the Stanford NLP Group as well as the Center for Research on Foundation Models (CRFM) for their helpful feedback.
171 |
--------------------------------------------------------------------------------
/assets/alpaca_main.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/teknium1/stanford_alpaca-replit/855ccdf2184f185e6c4a9c1ddbca46f000a8d9fc/assets/alpaca_main.jpg
--------------------------------------------------------------------------------
/assets/alpaca_right_email.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/teknium1/stanford_alpaca-replit/855ccdf2184f185e6c4a9c1ddbca46f000a8d9fc/assets/alpaca_right_email.png
--------------------------------------------------------------------------------
/assets/alpaca_right_llama.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/teknium1/stanford_alpaca-replit/855ccdf2184f185e6c4a9c1ddbca46f000a8d9fc/assets/alpaca_right_llama.png
--------------------------------------------------------------------------------
/assets/alpaca_wrong_42.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/teknium1/stanford_alpaca-replit/855ccdf2184f185e6c4a9c1ddbca46f000a8d9fc/assets/alpaca_wrong_42.png
--------------------------------------------------------------------------------
/assets/alpaca_wrong_capital.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/teknium1/stanford_alpaca-replit/855ccdf2184f185e6c4a9c1ddbca46f000a8d9fc/assets/alpaca_wrong_capital.png
--------------------------------------------------------------------------------
/assets/logo.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/teknium1/stanford_alpaca-replit/855ccdf2184f185e6c4a9c1ddbca46f000a8d9fc/assets/logo.png
--------------------------------------------------------------------------------
/assets/parse_analysis.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/teknium1/stanford_alpaca-replit/855ccdf2184f185e6c4a9c1ddbca46f000a8d9fc/assets/parse_analysis.png
--------------------------------------------------------------------------------
/configs/default_offload_opt_param.json:
--------------------------------------------------------------------------------
1 | {
2 | "bf16": {
3 | "enabled": "auto"
4 | },
5 | "optimizer": {
6 | "type": "AdamW",
7 | "params": {
8 | "lr": "auto",
9 | "betas": "auto",
10 | "eps": "auto",
11 | "weight_decay": "auto"
12 | }
13 | },
14 | "scheduler": {
15 | "type": "WarmupDecayLR",
16 | "params": {
17 | "total_num_steps": "auto",
18 | "warmup_min_lr": "auto",
19 | "warmup_max_lr": "auto",
20 | "warmup_num_steps": "auto"
21 | }
22 | },
23 | "zero_optimization": {
24 | "stage": 3,
25 | "offload_optimizer": {
26 | "device": "cpu",
27 | "pin_memory": true
28 | },
29 | "offload_param": {
30 | "device": "cpu",
31 | "pin_memory": true
32 | },
33 | "overlap_comm": true,
34 | "contiguous_gradients": true,
35 | "sub_group_size": 1e9,
36 | "reduce_bucket_size": "auto",
37 | "stage3_prefetch_bucket_size": "auto",
38 | "stage3_param_persistence_threshold": "auto",
39 | "stage3_max_live_parameters": 1e9,
40 | "stage3_max_reuse_distance": 1e9,
41 | "stage3_gather_16bit_weights_on_model_save": false
42 | },
43 | "gradient_accumulation_steps": "auto",
44 | "gradient_clipping": "auto",
45 | "steps_per_print": 5,
46 | "train_batch_size": "auto",
47 | "train_micro_batch_size_per_gpu": "auto",
48 | "wall_clock_breakdown": false
49 | }
50 |
--------------------------------------------------------------------------------
/requirements.txt:
--------------------------------------------------------------------------------
1 | numpy
2 | rouge_score
3 | fire
4 | transformers>=4.28.1
5 | torch
6 | sentencepiece
7 | tokenizers>=0.13.3
8 | wandb
9 | einops
10 |
--------------------------------------------------------------------------------
/train.py:
--------------------------------------------------------------------------------
1 | # Copyright 2023 Rohan Taori, Ishaan Gulrajani, Tianyi Zhang, Yann Dubois, Xuechen Li
2 | #
3 | # Licensed under the Apache License, Version 2.0 (the "License");
4 | # you may not use this file except in compliance with the License.
5 | # You may obtain a copy of the License at
6 | #
7 | # http://www.apache.org/licenses/LICENSE-2.0
8 | #
9 | # Unless required by applicable law or agreed to in writing, software
10 | # distributed under the License is distributed on an "AS IS" BASIS,
11 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 | # See the License for the specific language governing permissions and
13 | # limitations under the License.
14 |
15 | import copy
16 | import logging
17 | from dataclasses import dataclass, field
18 | from typing import Dict, Optional, Sequence
19 |
20 | import torch
21 | import transformers
22 | import utils
23 | from torch.utils.data import Dataset
24 | from transformers import Trainer
25 |
26 | IGNORE_INDEX = -100
27 | DEFAULT_PAD_TOKEN = "<|pad|>"
28 | DEFAULT_EOS_TOKEN = "<|endoftext|>"
29 | DEFAULT_UNK_TOKEN = "<|unk|>"
30 | PROMPT_DICT = {
31 | "prompt_input": (
32 | "### Instruction:\n{instruction}\n\n### Input:\n{input}\n\n### Response:"
33 | ),
34 | "prompt_no_input": (
35 | "### Instruction:\n{instruction}\n\n### Response:"
36 | ),
37 | }
38 |
39 |
40 | @dataclass
41 | class ModelArguments:
42 | model_name_or_path: Optional[str] = field(default="replit/replit-code-v1-3b")
43 |
44 |
45 | @dataclass
46 | class DataArguments:
47 | data_path: str = field(default=None, metadata={"help": "Path to the training data."})
48 |
49 |
50 | @dataclass
51 | class TrainingArguments(transformers.TrainingArguments):
52 | cache_dir: Optional[str] = field(default=None)
53 | optim: str = field(default="adamw_torch_fused")
54 | model_max_length: int = field(
55 | default=2000,
56 | metadata={"help": "Maximum sequence length. Sequences will be right padded (and possibly truncated)."},
57 | )
58 |
59 |
60 | def smart_tokenizer_and_embedding_resize(
61 | special_tokens_dict: Dict,
62 | tokenizer: transformers.PreTrainedTokenizer,
63 | model: transformers.PreTrainedModel,
64 | ):
65 | """Resize tokenizer and embedding.
66 |
67 | Note: This is the unoptimized version that may make your embedding size not be divisible by 64.
68 | """
69 | num_new_tokens = tokenizer.add_special_tokens(special_tokens_dict)
70 | model.resize_token_embeddings(len(tokenizer))
71 |
72 | if num_new_tokens > 0:
73 | input_embeddings = model.get_input_embeddings().weight.data
74 | output_embeddings = model.get_output_embeddings().weight.data
75 |
76 | input_embeddings_avg = input_embeddings[:-num_new_tokens].mean(dim=0, keepdim=True)
77 | output_embeddings_avg = output_embeddings[:-num_new_tokens].mean(dim=0, keepdim=True)
78 |
79 | input_embeddings[-num_new_tokens:] = input_embeddings_avg
80 | output_embeddings[-num_new_tokens:] = output_embeddings_avg
81 |
82 |
83 | def _tokenize_fn(strings: Sequence[str], tokenizer: transformers.PreTrainedTokenizer) -> Dict:
84 | """Tokenize a list of strings."""
85 | tokenized_list = [
86 | tokenizer(
87 | text,
88 | return_tensors="pt",
89 | padding="longest",
90 | max_length=tokenizer.model_max_length,
91 | truncation=True,
92 | )
93 | for text in strings
94 | ]
95 | input_ids = labels = [tokenized.input_ids[0] for tokenized in tokenized_list]
96 | input_ids_lens = labels_lens = [
97 | tokenized.input_ids.ne(tokenizer.pad_token_id).sum().item() for tokenized in tokenized_list
98 | ]
99 | return dict(
100 | input_ids=input_ids,
101 | labels=labels,
102 | input_ids_lens=input_ids_lens,
103 | labels_lens=labels_lens,
104 | )
105 |
106 |
107 | def preprocess(
108 | sources: Sequence[str],
109 | targets: Sequence[str],
110 | tokenizer: transformers.PreTrainedTokenizer,
111 | ) -> Dict:
112 | """Preprocess the data by tokenizing."""
113 | examples = [s + t for s, t in zip(sources, targets)]
114 | examples_tokenized, sources_tokenized = [_tokenize_fn(strings, tokenizer) for strings in (examples, sources)]
115 | input_ids = examples_tokenized["input_ids"]
116 | labels = copy.deepcopy(input_ids)
117 | for label, source_len in zip(labels, sources_tokenized["input_ids_lens"]):
118 | label[:source_len] = IGNORE_INDEX
119 | return dict(input_ids=input_ids, labels=labels)
120 |
121 |
122 | class SupervisedDataset(Dataset):
123 | """Dataset for supervised fine-tuning."""
124 |
125 | def __init__(self, data_path: str, tokenizer: transformers.PreTrainedTokenizer):
126 | super(SupervisedDataset, self).__init__()
127 | logging.warning("Loading data...")
128 | list_data_dict = utils.jload(data_path)
129 |
130 | logging.warning("Formatting inputs...")
131 | prompt_input, prompt_no_input = PROMPT_DICT["prompt_input"], PROMPT_DICT["prompt_no_input"]
132 | sources = [
133 | prompt_input.format_map(example) if example.get("input", "") != "" else prompt_no_input.format_map(example)
134 | for example in list_data_dict
135 | ]
136 | targets = [f"{example['output']}{tokenizer.eos_token}" for example in list_data_dict]
137 |
138 | logging.warning("Tokenizing inputs... This may take some time...")
139 | data_dict = preprocess(sources, targets, tokenizer)
140 |
141 | self.input_ids = data_dict["input_ids"]
142 | self.labels = data_dict["labels"]
143 |
144 | def __len__(self):
145 | return len(self.input_ids)
146 |
147 | def __getitem__(self, i) -> Dict[str, torch.Tensor]:
148 | return dict(input_ids=self.input_ids[i], labels=self.labels[i])
149 |
150 |
151 | @dataclass
152 | class DataCollatorForSupervisedDataset(object):
153 | """Collate examples for supervised fine-tuning."""
154 |
155 | tokenizer: transformers.PreTrainedTokenizer
156 |
157 | def __call__(self, instances: Sequence[Dict]) -> Dict[str, torch.Tensor]:
158 | input_ids, labels = tuple([instance[key] for instance in instances] for key in ("input_ids", "labels"))
159 | input_ids = torch.nn.utils.rnn.pad_sequence(
160 | input_ids, batch_first=True, padding_value=self.tokenizer.pad_token_id
161 | )
162 | labels = torch.nn.utils.rnn.pad_sequence(labels, batch_first=True, padding_value=IGNORE_INDEX)
163 | return dict(
164 | input_ids=input_ids,
165 | labels=labels,
166 | attention_mask=input_ids.ne(self.tokenizer.pad_token_id),
167 | )
168 |
169 |
170 | def make_supervised_data_module(tokenizer: transformers.PreTrainedTokenizer, data_args) -> Dict:
171 | """Make dataset and collator for supervised fine-tuning."""
172 | train_dataset = SupervisedDataset(tokenizer=tokenizer, data_path=data_args.data_path)
173 | data_collator = DataCollatorForSupervisedDataset(tokenizer=tokenizer)
174 | return dict(train_dataset=train_dataset, eval_dataset=None, data_collator=data_collator)
175 |
176 |
177 | def train():
178 | parser = transformers.HfArgumentParser((ModelArguments, DataArguments, TrainingArguments))
179 | model_args, data_args, training_args = parser.parse_args_into_dataclasses()
180 |
181 | model = transformers.AutoModelForCausalLM.from_pretrained(
182 | model_args.model_name_or_path,
183 | cache_dir=training_args.cache_dir,
184 | trust_remote_code=True,
185 | )
186 |
187 | tokenizer = transformers.AutoTokenizer.from_pretrained(
188 | model_args.model_name_or_path,
189 | cache_dir=training_args.cache_dir,
190 | model_max_length=training_args.model_max_length,
191 | padding_side="right",
192 | use_fast=False,
193 | trust_remote_code=True,
194 | )
195 | special_tokens_dict = dict()
196 | if tokenizer.pad_token is None:
197 | special_tokens_dict["pad_token"] = DEFAULT_PAD_TOKEN
198 | if tokenizer.eos_token is None:
199 | special_tokens_dict["eos_token"] = DEFAULT_EOS_TOKEN
200 | if tokenizer.unk_token is None:
201 | special_tokens_dict["unk_token"] = DEFAULT_UNK_TOKEN
202 |
203 | smart_tokenizer_and_embedding_resize(
204 | special_tokens_dict=special_tokens_dict,
205 | tokenizer=tokenizer,
206 | model=model,
207 | )
208 |
209 | data_module = make_supervised_data_module(tokenizer=tokenizer, data_args=data_args)
210 | trainer = Trainer(model=model, tokenizer=tokenizer, args=training_args, **data_module)
211 | trainer.train()
212 | trainer.save_state()
213 | trainer.save_model(output_dir=training_args.output_dir)
214 |
215 |
216 | if __name__ == "__main__":
217 | train()
218 |
--------------------------------------------------------------------------------
/utils.py:
--------------------------------------------------------------------------------
1 | import dataclasses
2 | import logging
3 | import math
4 | import os
5 | import io
6 | import sys
7 | import time
8 | import json
9 | from typing import Optional, Sequence, Union
10 | import tqdm
11 | import copy
12 |
13 | def _make_w_io_base(f, mode: str):
14 | if not isinstance(f, io.IOBase):
15 | f_dirname = os.path.dirname(f)
16 | if f_dirname != "":
17 | os.makedirs(f_dirname, exist_ok=True)
18 | f = open(f, mode=mode)
19 | return f
20 |
21 |
22 | def _make_r_io_base(f, mode: str):
23 | if not isinstance(f, io.IOBase):
24 | f = open(f, mode=mode)
25 | return f
26 |
27 |
28 | def jdump(obj, f, mode="w", indent=4, default=str):
29 | """Dump a str or dictionary to a file in json format.
30 |
31 | Args:
32 | obj: An object to be written.
33 | f: A string path to the location on disk.
34 | mode: Mode for opening the file.
35 | indent: Indent for storing json dictionaries.
36 | default: A function to handle non-serializable entries; defaults to `str`.
37 | """
38 | f = _make_w_io_base(f, mode)
39 | if isinstance(obj, (dict, list)):
40 | json.dump(obj, f, indent=indent, default=default)
41 | elif isinstance(obj, str):
42 | f.write(obj)
43 | else:
44 | raise ValueError(f"Unexpected type: {type(obj)}")
45 | f.close()
46 |
47 |
48 | def jload(f, mode="r"):
49 | """Load a .json file into a dictionary."""
50 | f = _make_r_io_base(f, mode)
51 | jdict = json.load(f)
52 | f.close()
53 | return jdict
54 |
--------------------------------------------------------------------------------