Initial Commit

README.md

@@ -0,0 +1,45 @@
+---
+license: cc-by-nc-nd-4.0
+---
+# Introduction
+TQCompressedGPT-2 is an advanced neural network model, offering a novel method for model compression through improved tensor decompositions. It addresses the challenges of computational and storage demands in NLP tasks, introducing a permutation-based enhancement to Kronecker decomposition, significantly reducing model size while maintaining performance.\
+TQCompressedGPT2 © 2024 by Terra Quantum AG is licensed under CC BY-NC-ND 4.0. To view a copy of this license, visit http://creativecommons.org/licenses/by-nc-nd/4.0/ \
+Any entity who wishes to use this library for commercial purposes should contact info@terraquantum.swiss for more information.\
+[![License: CC BY-NC-ND 4.0](https://img.shields.io/badge/License-CC%20BY--NC--ND%204.0-lightgrey.svg)](https://creativecommons.org/licenses/by-nc-nd/4.0/)\
+<img src="https://cdn-uploads.huggingface.co/production/uploads/6476003bbed7adbb05f8441f/jEKdKKFoEzlAbbI4NnokH.png" width="500">
+# Features
+**Model Size Reduction:** Compresses the GPT-2small model from 124 million to 81 million parameters.\
+**Permutation-Based Enhancement:** Introduces a new permutation algorithm for matrix factorization, minimizing performance degradation.\
+**Efficient Training Strategy:** Employs multi-step knowledge distillation with a fraction (3.1%) of the OpenWebText dataset.\
+**Performance:** Outperforms DistilGPT-2 in comparative evaluations.\
+<img src="https://cdn-uploads.huggingface.co/production/uploads/6476003bbed7adbb05f8441f/x1krVBC2RTZNDR0dynbRp.png" width="500">
+
+## Permutation-Based Enhancement
+In our work we employ permutation-based algorithm, which allows to achieve better decomposition approximation for weight matrices:\
+<img src="https://cdn-uploads.huggingface.co/production/uploads/6476003bbed7adbb05f8441f/bM6KwfKWYBJjeX_xGw83C.png" width="500">
+
+# Methodology
+For more details about the techniques of TQCompressedGPT-2, refer to our paper: **(ADD LINK)TQCompressor: Improving Tensor Decomposition in Neural Networks via Permutations**\
+**TQCompressed Decomposition:** Focuses on optimal permutation of weight matrices followed by Kronecker decomposition.\
+**Knowledge Distillation:** Uses an iterative compression method coupled with knowledge distillation, enhancing performance.\
+**Application:** Demonstrated on the GPT-2 model, showing its versatility and applicability to various neural network architectures.
+
+# Usage
+The model and code are publicly available at:
+- [GitHub Repository](https://github.com/terra-quantum-io/TQCompressedGPT2)
+- [HuggingFace Repository](https://huggingface.co/tq-ai-research/TQCompressedGPT2)
+
+# Citation
+If you find TQCompressedGPT-2 useful in your research, please cite the following paper:
+```
+@article{tqcompressedgpt2,
+  title={TQCompressor: Improving Tensor Decomposition in Neural Networks via Permutations},
+  author={Abronin, V., Naumov, A., Mazur, D., Bystrov, D., Tsarova, K., Melnikov, Ar., Oseledets, I., Dolgov, S., Brasher, R., Perelshtein, M.},
+  journal={arXiv preprint arXiv:[insert_arxiv_id]},
+  year={2023}
+}
+```
+
+# Acknowledgments
+- [Terra Quantum AG](https://terraquantum.swiss/), Kornhausstrasse 25, 9000 St. Gallen, Switzerland
+- Project contributors and researchers.

config.json

@@ -0,0 +1,39 @@
+{
+  "_name_or_path": "gpt2",
+  "activation_function": "gelu_new",
+  "architectures": [
+    "GPT2LMHeadModel"
+  ],
+  "attn_pdrop": 0.1,
+  "bos_token_id": 50256,
+  "embd_pdrop": 0.1,
+  "eos_token_id": 50256,
+  "initializer_range": 0.02,
+  "layer_norm_epsilon": 1e-05,
+  "model_type": "gpt2",
+  "n_ctx": 1024,
+  "n_embd": 768,
+  "n_head": 12,
+  "n_inner": null,
+  "n_layer": 12,
+  "n_positions": 1024,
+  "reorder_and_upcast_attn": false,
+  "resid_pdrop": 0.1,
+  "scale_attn_by_inverse_layer_idx": false,
+  "scale_attn_weights": true,
+  "summary_activation": null,
+  "summary_first_dropout": 0.1,
+  "summary_proj_to_labels": true,
+  "summary_type": "cls_index",
+  "summary_use_proj": true,
+  "task_specific_params": {
+    "text-generation": {
+      "do_sample": true,
+      "max_length": 50
+    }
+  },
+  "torch_dtype": "float32",
+  "transformers_version": "4.31.0",
+  "use_cache": true,
+  "vocab_size": 50257
+}

generation_config.json

@@ -0,0 +1,6 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 50256,
+  "eos_token_id": 50256,
+  "transformers_version": "4.31.0"
+}

gitattributes

@@ -0,0 +1,35 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c65274bbf42607d7451261fc0a5a671bc28dc5f7ac86b77adf0e5e684ea4c75c
+size 480171457