File size: 16,019 Bytes
4c2f9d6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d4f884d
 
5732644
 
 
 
4c2f9d6
 
d4f884d
 
 
 
5732644
 
 
 
 
d4f884d
 
 
 
 
 
 
 
 
5732644
 
d4f884d
 
 
 
 
5732644
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4c2f9d6
 
d4f884d
 
4c2f9d6
 
d4f884d
 
9bb9f71
05b3a30
d4f884d
9bb9f71
d4f884d
 
9bb9f71
d4f884d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4c2f9d6
d4f884d
4c2f9d6
d4f884d
4c2f9d6
d4f884d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4c2f9d6
d4f884d
 
 
 
 
 
 
 
 
4c2f9d6
d4f884d
4c2f9d6
d4f884d
 
 
 
 
 
 
4c2f9d6
d4f884d
4c2f9d6
d4f884d
 
 
 
 
 
 
 
 
 
05b3a30
 
 
 
 
 
 
 
 
 
d4f884d
 
 
 
 
 
 
 
 
 
 
 
4c2f9d6
 
d4f884d
4c2f9d6
 
 
 
 
 
 
 
 
 
 
d4f884d
4c2f9d6
 
 
d4f884d
 
 
4c2f9d6
d4f884d
 
 
 
 
 
 
 
4c2f9d6
 
d4f884d
 
 
 
 
 
 
4c2f9d6
d4f884d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4c2f9d6
 
 
 
 
d4f884d
 
 
 
9bb9f71
 
 
d4f884d
9bb9f71
 
4c2f9d6
 
 
d4f884d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4c2f9d6
 
d4f884d
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
---
language:
- en
license: mit
tags:
- pytorch
- transformers
- text-generation
- language-model
- graph-neural-network
- sparse-attention
- adaptive-depth
- temporal-decay
- mesh-attention
- efficient-transformer
- novel-architecture
- causal-lm
- research
- preprint
- mesh-transformer
- dynamic-graph
- early-exit
- per-token-routing
library_name: pytorch
pipeline_tag: text-generation
datasets:
- vigneshwar234/TMT-Benchmarks
metrics:
- perplexity
doi: 10.5281/zenodo.20287390
extra_gated_prompt: |
  Paper DOI: https://doi.org/10.5281/zenodo.20287390
  Zenodo: https://zenodo.org/records/20287390
  GitHub: https://github.com/vignesh2027/TemporalMesh-Transformer
model-index:
- name: TemporalMesh Transformer (TMT-Base)
  results:
  - task:
      type: text-generation
      name: Language Modelling
    dataset:
      type: wikitext
      name: WikiText-2
      config: wikitext-2-raw-v1
      split: validation
    metrics:
    - type: perplexity
      value: 29.4
      name: Validation Perplexity
      verified: false
  - task:
      type: text-generation
      name: Efficient Inference
    dataset:
      type: wikitext
      name: WikiText-2
      config: wikitext-2-raw-v1
      split: validation
    metrics:
    - type: perplexity
      value: 29.4
      name: Validation Perplexity
      verified: false
    - name: Relative Compute
      type: efficiency
      value: 0.48
      verified: false
    - name: Avg Exit Layer
      type: efficiency
      value: 5.5
      verified: false
---
---

<div align="center">

# TemporalMesh Transformer (TMT)

### *Dynamic Graph Attention · Temporal Semantic Decay · Per-Token Adaptive Depth Routing*

[![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.20287390.svg)](https://doi.org/10.5281/zenodo.20287390)
[![Space](https://img.shields.io/badge/🤗%20Space-Live%20Demo-orange?style=flat-square)](https://huggingface.co/spaces/vigneshwar234/TemporalMesh-Transformer-Demo)
[![GitHub](https://img.shields.io/badge/GitHub-vignesh2027%2FTemporalMesh--Transformer-181717?style=flat-square&logo=github)](https://github.com/vignesh2027/TemporalMesh-Transformer)
[![Paper PDF](https://img.shields.io/badge/Paper-PDF%2020%20pages-red?style=flat-square&logo=adobeacrobatreader)](https://doi.org/10.5281/zenodo.20287390)
[![Dataset](https://img.shields.io/badge/Dataset-TMT--Benchmarks-FFD21E?style=flat-square&logo=huggingface)](https://huggingface.co/datasets/vigneshwar234/TMT-Benchmarks)
[![License: MIT](https://img.shields.io/badge/License-MIT-green?style=flat-square)](https://github.com/vignesh2027/TemporalMesh-Transformer/blob/main/LICENSE)
[![Zenodo](https://img.shields.io/badge/Zenodo-Published-blue?style=flat-square&logo=zenodo)](https://zenodo.org/records/20287390)

**Val. Perplexity: 29.4** · **~50% compute reduction** · **~120M parameters** · **WikiText-2**

</div>

---

## Overview

The **TemporalMesh Transformer (TMT)** is a novel autoregressive language model architecture that breaks the three fundamental assumptions shared by every standard transformer:

| Assumption Every Transformer Makes | How TMT Breaks It |
|:---|:---|
| Every token attends to every other — O(S²) cost | **Mesh Attention**: Dynamic kNN graph rebuilt each layer — O(S·k) |
| Attention topology is flat and fixed | **Mesh Graph**: Topology changes every forward pass from token similarity |
| Every token uses identical compute (all N layers) | **Adaptive Depth**: Easy tokens exit after 2 layers; hard tokens use all 12 |

No single prior paper combines all three. That unification is the TMT research contribution.

---

## Architecture at a Glance

```
Input Tokens (B, S)


TokenEmbedding           ← Standard learned embedding × √d_model


TemporalPositionEncoder  ← RoPE + learned decay scalars per token


MeshBuilder              ← Cosine similarity → top-k graph  O(S·k)

      ▼  [× 12 layers]
┌─────────────────────────────────────────────────────┐
│  MeshAttention     ← Attention over graph edges only │
│  DualStreamFFN     ← Syntax stream + Semantic stream │
│  ExitGate          ← Freeze token if confidence>0.85 │
│  MemoryAnchorCross ← Cross-attend 16 EMA anchors     │
│  → Rebuild graph from updated representations        │
└─────────────────────────────────────────────────────┘


LayerNorm + OutputProjection (weight-tied to embedding)


TMTOutput: logits · exit_masks · confidences · graph_edges · memory_state
```

---

## The Five Innovations

### 1. Mesh Attention — Dynamic kNN Graph

At every layer, tokens are nodes. Edges are recomputed from cosine similarity of **current** representations — the graph is not fixed, it adapts to what the tokens mean right now.

```
sim(i,j) = Xᵢ · Xⱼ / (‖Xᵢ‖ · ‖Xⱼ‖)
N_k(i)   = top-k { j ≠ i : sim(i,j) }
Attention flows only along N_k edges  →  O(S·k) vs O(S²)
```

At S=1024, k=8: **128× fewer attention operations** than standard transformers.

### 2. Temporal Decay Encoding

A learned per-head scalar multiplied into post-softmax attention weights. Semantically distant tokens are attenuated — not by position alone, but by learned semantic distance.

```
δ_h(i,j) = σ( W_decay_h · |t_i − t_j| )
ã_ij      = α_ij · δ_h(i,j)
```

Unlike ALiBi (additive to logits, fixed schedule), TMT decay is **multiplicative, post-softmax, and fully learned**.

### 3. Adaptive Depth Routing — Per-Token Early Exit

Each token gets a confidence score after each layer. Confident tokens freeze and skip remaining layers.

```python
confidence = sigmoid(W_gate · x_token)   # ∈ (0,1)
if confidence > 0.85:
    token frozen — no more layers         # ~50% of tokens exit by layer 5
```

**Result**: ~50% average compute reduction. Punctuation exits at layer 2; rare technical terms use all 12.

### 4. Dual-Stream Feed-Forward Network

```
h_syntax   = GeLU(W_syn2 · GeLU(W_syn1 · x))   ← structural features
h_semantic = GeLU(W_sem2 · GeLU(W_sem1 · x))   ← meaning features
gate       = σ(W_gate_ffn · x)
output     = gate ⊙ h_syntax + (1−gate) ⊙ h_semantic
```

### 5. EMA Memory Anchors

16 persistent key-value vectors updated by EMA during training. Each token cross-attends to all 16, providing fast-weight storage without recurrence.

```
MemAttn(x)  = softmax(x·W_Q · K_mem^T / √d) · V_mem
k_m        ←  0.99 · k_m + 0.01 · mean(attending tokens)
```

---

## Performance

### WikiText-2 Benchmark (all models ~120M params, 10k steps)

| Model | Val PPL ↓ | Avg Layers/Token | Relative Compute |
|:---|:---:|:---:|:---:|
| Vanilla Transformer | 42.1 | 12.0 | 100% |
| + Mesh Attention only | 37.8 | 12.0 | 62% |
| + Temporal Decay only | 40.3 | 12.0 | 98% |
| + Adaptive Depth only | 39.6 | 5.8 | 51% |
| Mesh + Decay | 34.2 | 12.0 | 61% |
| Mesh + Exit | 35.1 | 5.7 | 50% |
| **Full TMT (all 3)** | **29.4** | **5.5** | **48%** |

### Compute Scaling

| Sequence Length | Standard Attn Ops | TMT Mesh Ops (k=8) | Reduction |
|:---:|:---:|:---:|:---:|
| 128 | 16,384 | 1,024 | 16× |
| 256 | 65,536 | 2,048 | 32× |
| 512 | 262,144 | 4,096 | 64× |
| 1024 | 1,048,576 | 8,192 | **128×** |
| 2048 | 4,194,304 | 16,384 | **256×** |

### Exit Gate Distribution (TMT-Base, step 10k)

| Token Type | Example | Avg Exit Layer | Compute Used |
|:---|:---|:---:|:---:|
| Punctuation | `. , ! ?` | 2.1 / 12 | 17% |
| Articles/Determiners | `a the an` | 3.4 / 12 | 28% |
| Common Nouns | `dog city` | 5.8 / 12 | 48% |
| Technical Terms | `neural FFN` | 9.3 / 12 | 78% |
| Rare Words | `palimpsest` | 11.7 / 12 | 98% |

---

## 🚀 Live Demo

Try TMT interactively — no install needed:

👉 **[huggingface.co/spaces/vigneshwar234/TemporalMesh-Transformer-Demo](https://huggingface.co/spaces/vigneshwar234/TemporalMesh-Transformer-Demo)**

Visualise exit gates, dynamic attention graphs, and per-token compute depth on any sentence you type.

---

## Quick Start

### Installation

```bash
git clone https://github.com/vignesh2027/TemporalMesh-Transformer.git
cd TemporalMesh-Transformer
python3 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
```

### Forward Pass

```python
import torch
from tmt.model.config import TMTConfig
from tmt.model.model import TMTModel

cfg = TMTConfig(
    vocab_size=50258,
    d_model=512,
    n_heads=8,
    n_layers=12,
    graph_k=8,
    exit_threshold=0.85,
    memory_anchors=16,
    max_seq_len=256,
)

model = TMTModel(cfg)
model.eval()

input_ids = torch.randint(0, 50258, (1, 64))  # batch=1, seq_len=64

with torch.no_grad():
    output = model(input_ids)

print("Logits shape:    ", output.logits.shape)          # (1, 64, 50258)
print("Exit masks:      ", len(output.exit_masks))       # 12 — one per layer
print("Tokens per layer:", [m.sum().item() for m in output.exit_masks])
print("Memory state:    ", output.memory_state.shape)    # (16, 512)
print("Graph edges:     ", output.graph_edges[0].shape)  # (2, E)
```

### Inspect Exit Behaviour

```python
# Which tokens exited at which layer?
for layer_idx, mask in enumerate(output.exit_masks):
    n_exited = mask.sum().item()
    print(f"Layer {layer_idx+1:2d}: {n_exited} tokens exited")

# Confidence scores per token
for layer_idx, conf in enumerate(output.confidences):
    print(f"Layer {layer_idx+1:2d}: avg confidence = {conf.mean():.3f}")
```

### Training (Quick CPU Run)

```python
from tmt.model.config import TMTConfig
from tmt.training.trainer import TMTTrainer, TrainConfig
from tmt.data.dataset import load_text_dataset

cfg = TMTConfig(vocab_size=50258, d_model=256, n_heads=4, n_layers=4,
                graph_k=4, ffn_stream_dim=128, memory_anchors=8, max_seq_len=128)

loaders = load_text_dataset('wikitext-2', seq_len=128, batch_size=8)

trainer = TMTTrainer(
    cfg,
    TrainConfig(total_steps=500, warmup_steps=50, use_wandb=False, eval_every=100),
    loaders['train'], loaders.get('validation')
)
trainer.train()
```

### Full GPU Training (Publication Quality)

```python
cfg = TMTConfig(
    vocab_size=50258, d_model=512, n_heads=8, n_layers=12,
    graph_k=8, decay_rate=0.1, exit_threshold=0.85,
    dual_stream=True, memory_anchors=16, ffn_stream_dim=256, max_seq_len=256,
)
train_cfg = TrainConfig(
    total_steps=10_000, warmup_steps=500, lr=3e-4, batch_size=16,
    eval_every=500, save_every=1000, use_wandb=True,
)
```

### Checkpoint Loading

```python
import torch
from tmt.model.config import TMTConfig
from tmt.model.model import TMTModel

cfg = TMTConfig(...)   # must match training config
model = TMTModel(cfg)
ckpt = torch.load('checkpoints/ckpt_step10000.pt', map_location='cpu')
model.load_state_dict(ckpt['model_state'])
model.eval()
```

---

## Configuration Reference

```python
TMTConfig(
    vocab_size      = 32000,   # vocabulary size
    d_model         = 512,     # hidden dimension
    n_heads         = 8,       # attention heads
    n_layers        = 12,      # transformer layers
    max_seq_len     = 1024,    # max sequence length

    # ── Mesh Attention ──────────────────────────────
    graph_k         = 8,       # kNN neighbourhood size (4–16)

    # ── Temporal Decay ──────────────────────────────
    decay_rate      = 0.1,     # base decay rate (0.05–0.4)

    # ── Adaptive Depth ──────────────────────────────
    exit_threshold  = 0.85,    # token exit confidence (0.70–0.95)

    # ── Dual-Stream FFN ─────────────────────────────
    dual_stream     = True,    # enable parallel syntax+semantic streams
    ffn_stream_dim  = 256,     # width per stream (total=512 for d_model=512)

    # ── Memory Anchors ──────────────────────────────
    memory_anchors  = 16,      # EMA anchor count (8–32)

    dropout         = 0.1,
)
```

### Model Scales

| Variant | d_model | Layers | Heads | k | Params | VRAM |
|:---|:---:|:---:|:---:|:---:|:---:|:---:|
| TMT-Small | 256 | 4 | 4 | 4 | ~16M | ~2 GB |
| TMT-Medium | 512 | 6 | 6 | 6 | ~60M | ~6 GB |
| **TMT-Base** | **512** | **12** | **8** | **8** | **~120M** | **~12 GB** |
| TMT-Large | 1024 | 24 | 16 | 16 | ~350M | ~40 GB |

---

## TMTOutput Fields

Every forward pass returns a rich structured output:

| Field | Shape | Description |
|:---|:---|:---|
| `logits` | `(B, S, V)` | Next-token logits — use for loss/generation |
| `exit_masks` | `list[(B, S) bool]` | True where token exited at that layer |
| `confidences` | `list[(B, S) float]` | Gate confidence per token per layer |
| `graph_edges` | `(edge_index, weights)` | Live sparse graph from final layer |
| `memory_state` | `(M, D)` | Final EMA memory anchor state |
| `decay_scalars` | `(B, S, D)` | Temporal decay weights applied |

---

## Test Dataset

The companion dataset **[vigneshwar234/TMT-Benchmarks](https://huggingface.co/datasets/vigneshwar234/TMT-Benchmarks)** contains:

- `complexity_test` — 1,000 sequences annotated by token complexity category
- `length_scaling` — sequences from S=32 to S=1024 for throughput benchmarking
- `ablation_reference` — canonical perplexity reference values for all 8 ablation configs
- `exit_gate_reference` — expected exit layer distributions per token type
- `edge_case_inputs` — boundary inputs for robustness testing (empty, max-length, all-same)

```python
from datasets import load_dataset
ds = load_dataset("vigneshwar234/TMT-Benchmarks", "complexity_test")
print(ds['test'][0])
# {'input_ids': [...], 'token_types': [...], 'expected_exit_layers': [...], 'text': '...'}
```

---

## Figures

| Figure | Description |
|:---|:---|
| [`fig_architecture.png`](paper/fig_architecture.png) | Full TMT architecture block diagram |
| [`fig_graph.png`](paper/fig_graph.png) | Dynamic graph evolution across 3 layers |
| [`fig_decay.png`](paper/fig_decay.png) | Temporal decay function curves + RoPE comparison |
| [`fig_exit.png`](paper/fig_exit.png) | Exit gate distribution by layer and token type |
| [`fig_training.png`](paper/fig_training.png) | Training loss + validation perplexity curves |
| [`fig_ablation.png`](paper/fig_ablation.png) | Ablation bar chart + Pareto frontier |
| [`fig_complexity.png`](paper/fig_complexity.png) | O(S²) vs O(S·k) operation count + memory |

---

## Citation

```bibtex
@misc{tmt2026,
  title        = {TemporalMesh Transformer: Dynamic Graph Attention with
                  Temporal Decay and Adaptive Depth Routing},
  author       = {Vignesh},
  year         = {2026},
  doi          = {10.5281/zenodo.20287390},
  url          = {https://doi.org/10.5281/zenodo.20287390},
  publisher    = {Zenodo},
  note         = {Preprint. Novel architecture combining mesh attention, temporal
                  decay encoding, and per-token adaptive depth routing.
                  Code: https://github.com/vignesh2027/TemporalMesh-Transformer}
}
```

---

## Related Work

| Paper | Relation to TMT |
|:---|:---|
| Vaswani et al. 2017 — *Attention Is All You Need* | Base architecture |
| Su et al. 2021 — *RoFormer (RoPE)* | TMT extends RoPE with learned decay |
| Elbayad et al. 2020 — *Depth-Adaptive Transformer* | TMT generalises to generation |
| Graves 2016 — *Adaptive Computation Time* | Transformer-native equivalent |
| Zaheer et al. 2020 — *BigBird* | Fixed sparse patterns vs TMT's dynamic graph |
| Shi et al. 2021 — *Graph Transformer* | Static graph vs TMT's rebuilt-per-layer graph |

---

## License

MIT — free to use, modify, and build upon. Citation appreciated for published work.