Angshul commited on
Commit
29e60f4
·
verified ·
1 Parent(s): b4636d7

Upload 4 files

Browse files
Files changed (4) hide show
  1. LICENSE +177 -0
  2. NOTICE +4 -0
  3. README.md +149 -3
  4. pyproject.toml +31 -0
LICENSE ADDED
@@ -0,0 +1,177 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+
2
+ Apache License
3
+ Version 2.0, January 2004
4
+ http://www.apache.org/licenses/
5
+
6
+ TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
7
+
8
+ 1. Definitions.
9
+
10
+ "License" shall mean the terms and conditions for use, reproduction,
11
+ and distribution as defined by Sections 1 through 9 of this document.
12
+
13
+ "Licensor" shall mean the copyright owner or entity authorized by
14
+ the copyright owner that is granting the License.
15
+
16
+ "Legal Entity" shall mean the union of the acting entity and all
17
+ other entities that control, are controlled by, or are under common
18
+ control with that entity. For the purposes of this definition,
19
+ "control" means (i) the power, direct or indirect, to cause the
20
+ direction or management of such entity, whether by contract or
21
+ otherwise, or (ii) ownership of fifty percent (50%) or more of the
22
+ outstanding shares, or (iii) beneficial ownership of such entity.
23
+
24
+ "You" (or "Your") shall mean an individual or Legal Entity
25
+ exercising permissions granted by this License.
26
+
27
+ "Source" form shall mean the preferred form for making modifications,
28
+ including but not limited to software source code, documentation
29
+ source, and configuration files.
30
+
31
+ "Object" form shall mean any form resulting from mechanical
32
+ transformation or translation of a Source form, including but
33
+ not limited to compiled object code, generated documentation,
34
+ and conversions to other media types.
35
+
36
+ "Work" shall mean the work of authorship, whether in Source or
37
+ Object form, made available under the License, as indicated by a
38
+ copyright notice that is included in or attached to the work
39
+ (an example is provided in the Appendix below).
40
+
41
+ "Derivative Works" shall mean any work, whether in Source or Object
42
+ form, that is based on (or derived from) the Work and for which the
43
+ editorial revisions, annotations, elaborations, or other modifications
44
+ represent, as a whole, an original work of authorship. For the purposes
45
+ of this License, Derivative Works shall not include works that remain
46
+ separable from, or merely link (or bind by name) to the interfaces of,
47
+ the Work and Derivative Works thereof.
48
+
49
+ "Contribution" shall mean any work of authorship, including
50
+ the original version of the Work and any modifications or additions
51
+ to that Work or Derivative Works thereof, that is intentionally
52
+ submitted to Licensor for inclusion in the Work by the copyright owner
53
+ or by an individual or Legal Entity authorized to submit on behalf of
54
+ the copyright owner. For the purposes of this definition, "submitted"
55
+ means any form of electronic, verbal, or written communication sent
56
+ to the Licensor or its representatives, including but not limited to
57
+ communication on electronic mailing lists, source code control systems,
58
+ and issue tracking systems that are managed by, or on behalf of, the
59
+ Licensor for the purpose of discussing and improving the Work, but
60
+ excluding communication that is conspicuously marked or otherwise
61
+ designated in writing by the copyright owner as "Not a Contribution."
62
+
63
+ "Contributor" shall mean Licensor and any individual or Legal Entity
64
+ on behalf of whom a Contribution has been received by Licensor and
65
+ subsequently incorporated within the Work.
66
+
67
+ 2. Grant of Copyright License. Subject to the terms and conditions of
68
+ this License, each Contributor hereby grants to You a perpetual,
69
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
70
+ copyright license to reproduce, prepare Derivative Works of,
71
+ publicly display, publicly perform, sublicense, and distribute the
72
+ Work and such Derivative Works in Source or Object form.
73
+
74
+ 3. Grant of Patent License. Subject to the terms and conditions of
75
+ this License, each Contributor hereby grants to You a perpetual,
76
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
77
+ (except as stated in this section) patent license to make, have made,
78
+ use, offer to sell, sell, import, and otherwise transfer the Work,
79
+ where such license applies only to those patent claims licensable
80
+ by such Contributor that are necessarily infringed by their
81
+ Contribution(s) alone or by combination of their Contribution(s)
82
+ with the Work to which such Contribution(s) was submitted. If You
83
+ institute patent litigation against any entity (including a
84
+ cross-claim or counterclaim in a lawsuit) alleging that the Work
85
+ or a Contribution incorporated within the Work constitutes direct
86
+ or contributory patent infringement, then any patent licenses
87
+ granted to You under this License for that Work shall terminate
88
+ as of the date such litigation is filed.
89
+
90
+ 4. Redistribution. You may reproduce and distribute copies of the
91
+ Work or Derivative Works thereof in any medium, with or without
92
+ modifications, and in Source or Object form, provided that You
93
+ meet the following conditions:
94
+
95
+ (a) You must give any other recipients of the Work or
96
+ Derivative Works a copy of this License; and
97
+
98
+ (b) You must cause any modified files to carry prominent notices
99
+ stating that You changed the files; and
100
+
101
+ (c) You must retain, in the Source form of any Derivative Works
102
+ that You distribute, all copyright, patent, trademark, and
103
+ attribution notices from the Source form of the Work,
104
+ excluding those notices that do not pertain to any part of
105
+ the Derivative Works; and
106
+
107
+ (d) If the Work includes a "NOTICE" text file as part of its
108
+ distribution, then any Derivative Works that You distribute must
109
+ include a readable copy of the attribution notices contained
110
+ within such NOTICE file, excluding those notices that do not
111
+ pertain to any part of the Derivative Works, in at least one
112
+ of the following places: within a NOTICE text file distributed
113
+ as part of the Derivative Works; within the Source form or
114
+ documentation, if provided along with the Derivative Works; or,
115
+ within a display generated by the Derivative Works, if and
116
+ wherever such third-party notices normally appear. The contents
117
+ of the NOTICE file are for informational purposes only and
118
+ do not modify the License. You may add Your own attribution
119
+ notices within Derivative Works that You distribute, alongside
120
+ or as an addendum to the NOTICE text from the Work, provided
121
+ that such additional attribution notices cannot be construed
122
+ as modifying the License.
123
+
124
+ You may add Your own copyright statement to Your modifications and
125
+ may provide additional or different license terms and conditions
126
+ for use, reproduction, or distribution of Your modifications, or
127
+ for any such Derivative Works as a whole, provided Your use,
128
+ reproduction, and distribution of the Work otherwise complies with
129
+ the conditions stated in this License.
130
+
131
+ 5. Submission of Contributions. Unless You explicitly state otherwise,
132
+ any Contribution intentionally submitted for inclusion in the Work
133
+ by You to the Licensor shall be under the terms and conditions of
134
+ this License, without any additional terms or conditions.
135
+ Notwithstanding the above, nothing herein shall supersede or modify
136
+ the terms of any separate license agreement you may have executed
137
+ with Licensor regarding such Contributions.
138
+
139
+ 6. Trademarks. This License does not grant permission to use the trade
140
+ names, trademarks, service marks, or product names of the Licensor,
141
+ except as required for reasonable and customary use in describing the
142
+ origin of the Work and reproducing the content of the NOTICE file.
143
+
144
+ 7. Disclaimer of Warranty. Unless required by applicable law or
145
+ agreed to in writing, Licensor provides the Work (and each
146
+ Contributor provides its Contributions) on an "AS IS" BASIS,
147
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
148
+ implied, including, without limitation, any warranties or conditions
149
+ of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
150
+ PARTICULAR PURPOSE. You are solely responsible for determining the
151
+ appropriateness of using or redistributing the Work and assume any
152
+ risks associated with Your exercise of permissions under this License.
153
+
154
+ 8. Limitation of Liability. In no event and under no legal theory,
155
+ whether in tort (including negligence), contract, or otherwise,
156
+ unless required by applicable law (such as deliberate and grossly
157
+ negligent acts) or agreed to in writing, shall any Contributor be
158
+ liable to You for damages, including any direct, indirect, special,
159
+ incidental, or consequential damages of any character arising as a
160
+ result of this License or out of the use or inability to use the
161
+ Work (including but not limited to damages for loss of goodwill,
162
+ work stoppage, computer failure or malfunction, or any and all
163
+ other commercial damages or losses), even if such Contributor
164
+ has been advised of the possibility of such damages.
165
+
166
+ 9. Accepting Warranty or Additional Liability. While redistributing
167
+ the Work or Derivative Works thereof, You may choose to offer,
168
+ and charge a fee for, acceptance of support, warranty, indemnity,
169
+ or other liability obligations and/or rights consistent with this
170
+ License. However, in accepting such obligations, You may act only
171
+ on Your own behalf and on Your sole responsibility, not on behalf
172
+ of any other Contributor, and only if You agree to indemnify,
173
+ defend, and hold each Contributor harmless for any liability
174
+ incurred by, or claims asserted against, such Contributor by reason
175
+ of your accepting any such warranty or additional liability.
176
+
177
+ END OF TERMS AND CONDITIONS
NOTICE ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ TensorPEFT
2
+ Copyright 2026 Angshul Majumdar
3
+
4
+ Licensed under the Apache License, Version 2.0 (see LICENSE).
README.md CHANGED
@@ -1,3 +1,149 @@
1
- ---
2
- license: apache-2.0
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # TensorPEFT
2
+
3
+ Parameter-efficient fine-tuning with **tensor gradient-pursuit adapters**. TensorPEFT is a drop-in
4
+ alternative to LoRA for Hugging Face models, designed for very small trainable-parameter budgets.
5
+
6
+ LoRA gives every weight matrix its own low-rank update. TensorPEFT instead stacks all
7
+ equal-shaped matrices of the network into one tensor, for example
8
+ `layers x {q,k,v,o} x d_in x d_out`, and learns a few **rank-one CP atoms** of that tensor:
9
+
10
+ dW = sum_i a_i u_i(layer) o u_i(role) o u_i(in) o u_i(out)
11
+
12
+ Each atom couples every layer and projection while costing only
13
+ `L + P + d_in + d_out + 1` parameters. The atoms are not random. They are the leading rank-one
14
+ CP components of the negative gradient tensor (higher-order power iteration with deflation),
15
+ computed from **one gradient probe** before training. The coefficients start at zero, so the
16
+ model is unchanged at step 0, and every atom starts on a steepest-descent direction. After that,
17
+ training is ordinary: any optimiser and the HF `Trainer` both work.
18
+
19
+ ## Install
20
+
21
+ pip install tensorpeft # once published; until then: pip install .
22
+
23
+ It requires `torch>=2.1`, `transformers>=4.40`, `safetensors` and `huggingface_hub`.
24
+
25
+ ## Quick start
26
+
27
+ ```python
28
+ from transformers import AutoModelForCausalLM
29
+ from tensorpeft import TensorPEFTConfig, get_tensorpeft_model, TensorPEFTModel
30
+
31
+ base = AutoModelForCausalLM.from_pretrained("gpt2")
32
+ model = get_tensorpeft_model(base, TensorPEFTConfig(atoms_per_group=4))
33
+
34
+ model.init_atoms(train_dataloader, n_batches=4) # the one step LoRA does not have
35
+ model.print_trainable_parameters() # trainable params: 37,036 || ... || 0.0298%
36
+
37
+ # train as usual: AdamW, lr ~3e-3, short warm-up, weight_decay 0 (or TensorPEFTTrainer below)
38
+
39
+ model.save_pretrained("my-adapter") # adapter_config.json + adapter_model.safetensors
40
+ model.push_to_hub("your-name/my-adapter")
41
+
42
+ base = AutoModelForCausalLM.from_pretrained("gpt2")
43
+ model = TensorPEFTModel.from_pretrained(base, "your-name/my-adapter") # local dir or Hub id
44
+ merged = model.merge_and_unload() # plain HF model, zero inference overhead
45
+ ```
46
+
47
+ `train_dataloader` can yield dicts (`input_ids`, optional `attention_mask` / `labels`) or plain
48
+ input-id tensors. For causal LMs, labels default to the inputs.
49
+
50
+ ### With the Hugging Face Trainer
51
+
52
+ ```python
53
+ from tensorpeft import TensorPEFTTrainer
54
+ trainer = TensorPEFTTrainer(model=model, args=training_args, train_dataset=ds, data_collator=collator)
55
+ trainer.train() # checkpoints contain only the adapter
56
+ ```
57
+
58
+ `TensorPEFTTrainer` is the stock `Trainer` except that it saves adapters instead of full model
59
+ weights. A complete script is in `examples/train_with_trainer.py`.
60
+
61
+ ## Configuration
62
+
63
+ | `TensorPEFTConfig` field | default | meaning |
64
+ |---|---|---|
65
+ | `atoms_per_group` | 4 | rank-one atoms per tensor group; this sets the budget |
66
+ | `target_modules` | `"all-linear"` | or a list of name suffixes, e.g. `["q_proj", "v_proj"]` (`"c_attn.q"` for GPT-2) |
67
+ | `exclude_modules` | `lm_head`, embeddings | never adapted |
68
+ | `split_fused_qkv` | True | split GPT-2's fused `c_attn` into q, k and v |
69
+ | `init_batches` | 4 | batches used by `init_atoms` |
70
+ | `hopm_iters` | 30 | power-iteration sweeps per atom |
71
+ | `alpha_scale` | 33.3 | coefficient reparametrisation so that one learning rate (~3e-3) fits coefficients and factors |
72
+
73
+ **Budget.** Groups are formed automatically by matrix shape. GPT-2 small has three groups:
74
+ `{q,k,v,o}` (12x4x768x768), MLP-in (12x768x3072) and MLP-out (12x3072x768). One atom per group
75
+ is 9,259 parameters, four per group is 37,036, and six per group is 55,554.
76
+ `print_trainable_parameters()` shows the exact count for any model.
77
+
78
+ **Recipe.** Use AdamW at lr 3e-3 with about 10 warm-up steps, then a constant schedule and
79
+ weight decay 0. Use all atoms from the start: adding atoms gradually during training was clearly
80
+ worse in our tests.
81
+
82
+ ## Results
83
+
84
+ Setup: GPT-2 124M (English base) fine-tuned on Bengali news (3M tokens), 300 AdamW steps,
85
+ batch 8 x 128 tokens. Validation cross-entropy is in nats/token, lower is better; the base model
86
+ scores 1.708. All methods saw identical data, and baseline learning rates follow their papers.
87
+
88
+ Three seeds, best baseline per budget:
89
+
90
+ | budget | TensorPEFT | best baseline | difference | seeds won | p (paired t) |
91
+ |---|---:|---|---:|:---:|---:|
92
+ | ~9k params | **1.3245 ± 0.004** | prompt tuning: 1.3684 ± 0.010 | **-0.044** | 3/3 | 0.008 |
93
+ | ~37k | 1.2558 ± 0.010 | LayerNorm tuning (38.4k): 1.2504 ± 0.003 | +0.005 | 1/3 | 0.37 (tie) |
94
+ | ~55k | **1.2368 ± 0.017** | (IA)³: 1.2880 ± 0.003 | **-0.051** | 3/3 | 0.026 |
95
+
96
+ All baselines at each budget (seed 0):
97
+
98
+ | budget | TensorPEFT | LoRA | QLoRA | VeRA | (IA)³ | prompt tuning | LayerNorm tuning | BitFit |
99
+ |---|---:|---:|---:|---:|---:|---:|---:|---:|
100
+ | ~1.5k | **1.461** | - | - | - | - | 1.493 (768 params) | 1.600 | - |
101
+ | ~9k | **1.327** | 1.430 | - | - | 1.432 | 1.370 | 1.407 | - |
102
+ | ~37k | 1.255 | 1.318 | 1.335 | 1.393 (24.6k) | - | 1.327 | **1.253** | - |
103
+ | ~55k | **1.231** | - | - | - | 1.289 | - | - | - |
104
+ | ~83k | - | - | - | - | - | - | - | 1.326 |
105
+
106
+ TensorPEFT budgets map to configs as follows: ~1.5k is one atom on the attention tensor only
107
+ (`target_modules=["c_attn.q", "c_attn.k", "c_attn.v", "attn.c_proj"]`, `atoms_per_group=1`),
108
+ ~9k is `atoms_per_group=1`, ~37k is 4 and ~55k is 6. A dash means the method cannot be
109
+ configured at that budget on GPT-2, or was not run there.
110
+ LoRA's smallest all-layer configuration is r=1 on q,v (37k parameters). At 9k it is LoRA r=1 on
111
+ q,v of the last 3 layers.
112
+
113
+ **What this does and does not show.** TensorPEFT beats every baseline from about 1.5k to
114
+ about 9k parameters and at about 55k, and ties LayerNorm tuning at about 37k. With larger
115
+ budgets (≥ 115k) LoRA was better in our tests. The evidence covers one base model, one task and
116
+ short training. Results on other models and downstream tasks are not yet available.
117
+
118
+ `examples/reproduce_gpt2_bengali.py` reruns these benchmarks with this package
119
+ (`--atoms-per-group 1 | 4 | 6`). The
120
+ data comes from the Bengali news corpus at github.com/soham96/Bengali_news_classifier,
121
+ tokenised with the GPT-2 tokenizer.
122
+
123
+ ## Supported models
124
+
125
+ Any model whose transformer blocks use `nn.Linear` or GPT-2-style `Conv1D` layers inside
126
+ numbered blocks (`...layers.N...` / `...h.N...`). Tested on GPT-2, Llama (GQA), Qwen2, Mistral,
127
+ Phi and Gemma configurations. Fused projections other than GPT-2's `c_attn` are treated as a
128
+ single matrix.
129
+
130
+ ## Limitations (v0.1)
131
+
132
+ - **GPU support is untested.** The code is device- and dtype-agnostic (adapter parameters in
133
+ fp32, bf16/fp16 bases supported), but all tests and benchmarks so far ran on CPU.
134
+ - **Probe memory.** `init_atoms` needs one backward pass with gradients on the targeted weights,
135
+ plus an fp32 copy of each group's gradient tensor. That is roughly 2x the size of the targeted
136
+ weights, temporarily. Use `target_modules` to limit it for large models.
137
+ - **Quantised bases** (bitsandbytes 4/8-bit) are not supported yet, because the probe needs
138
+ weight gradients.
139
+ - **No checkpoint resume.** `Trainer` resume-from-checkpoint is not implemented; reload with
140
+ `TensorPEFTModel.from_pretrained(..., is_trainable=True)`.
141
+
142
+ ## Tests
143
+
144
+ pip install .[test]
145
+ pytest tests
146
+
147
+ ## License
148
+
149
+ Apache-2.0. Copyright 2026 Angshul Majumdar.
pyproject.toml ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [build-system]
2
+ requires = ["setuptools>=64", "wheel"]
3
+ build-backend = "setuptools.build_meta"
4
+
5
+ [project]
6
+ name = "tensorpeft"
7
+ version = "0.1.0"
8
+ description = "Parameter-efficient fine-tuning with tensor gradient-pursuit (rank-one CP) adapters"
9
+ readme = "README.md"
10
+ license = "Apache-2.0"
11
+ license-files = ["LICENSE", "NOTICE"]
12
+ authors = [{ name = "Angshul Majumdar", email = "angshul@iiitd.ac.in" }]
13
+ requires-python = ">=3.9"
14
+ dependencies = [
15
+ "torch>=2.1",
16
+ "transformers>=4.40",
17
+ "safetensors>=0.4",
18
+ "huggingface_hub>=0.20",
19
+ ]
20
+ keywords = ["peft", "fine-tuning", "lora", "tensor", "transformers", "huggingface"]
21
+ classifiers = [
22
+ "Programming Language :: Python :: 3",
23
+ "Intended Audience :: Science/Research",
24
+ "Topic :: Scientific/Engineering :: Artificial Intelligence",
25
+ ]
26
+
27
+ [project.optional-dependencies]
28
+ test = ["pytest", "accelerate"]
29
+
30
+ [tool.setuptools.packages.find]
31
+ where = ["src"]