| name | reversible-sparse-moe-single-node |
| description | Train hundred-billion-parameter sparse MoE models on single nodes using reversible recurrence stacks and state-preserving growth principles with TQP optimizer strategy. |
| authors | ["Rohan Shravan"] |
| date | 2026-06-05T00:00:00.000Z |
| arxiv | 2606.07404v1 |
| tags | ["foundation-model","sparse-moe","single-node","reversible-computing","memory-efficient","scaling"] |
Reversible Sparse MoE Single-Node Training
Overview
Train 120B sparse mixture-of-experts (MoE) models on single eight-GPU nodes through reversible recurrence stacks, state-preserving growth, and TQP optimizer strategy. Achieves flat activation memory regardless of model growth through reversibility principle.
Key Innovation
Three Disciplines Integration:
- Reversibility: Recurrence stack reconstructs activations in backward pass
- State-Preserving Growth: Reproducible expansion principles with documented failure modes
- Single-Node Economics: TQP (quantized base + trained low-rank adapters) reduces optimizer state by ~45x
Memory Efficiency: Activation memory remains flat as model grows (reversibility eliminates storage)
Parameter Efficiency: Active params rise monotonically (1.78B → 5.93B, ~5% of 118.67B stored)
Methodology
Principle 1: Reversible Recurrence Stack
# Standard: Store all activations for backward pass
# Memory: O(depth * batch_size)
# Reversible: Reconstruct activations during backward
# Memory: O(batch_size) (flat!)
Forward:
state = initial_state
for layer in stack:
state = layer(state)
# No storage of intermediate states
Backward:
# Reconstruct from final state
state = final_state
for layer in reversed(stack):
state = layer.inverse(state) # Reversible operation
# Compute gradients locally
Principle 2: State-Preserving Growth
Four-Stage Growth Lineage:
- Dense seed (small dense model)
- 5B MoE (dense → MoE expansion)
- 9B MoE (shallow → deep expansion)
- 120B MoE (few → many experts)
Each Expansion Principle:
- Dense to MoE: Initialize experts from dense weights
- Shallow to deep: Copy layers, add depth
- Few to many: Duplicate experts, reinitialize routing
Failure Modes (Silent):
- Incorrect initialization → routing collapse
- Wrong growth factor → gradient instability
- Improper scaling → capacity mismatch
Principle 3: TQP Optimizer Strategy
# Traditional MoE: Optimizer state for all routed experts
# Memory: O(118B * optimizer_factor) = huge
# TQP: Quantized base + trained low-rank adapters
Base expert weights: Quantized (fixed)
Low-rank adapters: Full optimizer state (2.26B params)
Optimizer state reduction:
2.26B / 118.67B ≈ 1.9% of stored weights
~45x reduction vs. full expert optimizer state
Architecture Details
LightningLM 0.1V Specs:
- 120B total parameters
- 460 routed experts
- Top-12 routing
- 5.93B active parameters per forward pass
- 8K context length
- Single 8-GPU node training
Released Metrics:
- Training loss: 1.78 at 120B scale
- Indic language competence (targeted)
- Code capabilities (targeted)
Reusable Patterns
Pattern 1: Reversible Layer Implementation
Use when: Memory-constrained training of deep models
Key requirement: Layers must have inverse operations
Example reversible operations:
- Addition/subtraction
- Matrix multiplication with stored matrices
- Attention with reversible patterns (e.g., reversible residual blocks)
Pattern 2: State-Preserving Expansion
Use when: Growing model capacity without retraining from scratch
Steps:
- Identify expansion type (dense→MoE, shallow→deep, few→many)
- Define initialization from smaller model
- Document expected failure modes (silent failures critical)
- Validate state preservation with held-out loss
Pattern 3: TQP Optimizer for Sparse Models
Use when: Sparse MoE with prohibitive optimizer state
Implementation:
- Quantize base expert weights (freeze or minimal updates)
- Train low-rank adapters per expert
- Optimizer state only on adapters
- Result: ~45x optimizer state reduction
Pattern 4: Growth Lineage Training
Use when: Progressive scaling from seed to large model
Approach:
- Stage 1: Dense seed (fast training)
- Stage 2: MoE conversion (efficient capacity boost)
- Stage 3: Depth expansion (context/complexity)
- Stage 4: Expert multiplication (final capacity)
- Each stage: State-preserving, not retraining
Implementation Considerations
Reversibility Implementation
- Use reversible residual blocks (e.g., RevNet)
- Attention: Store only queries, reconstruct keys/values
- MoE routing: Stateless routing decisions
State-Preserving Checks
- Held-out loss before and after expansion
- Targeted capability metrics (multilingual, code)
- Per-domain validation
TQP Quantization
- Base weights: Int8 or Int4 quantization
- Adapters: Full precision, trainable
- Router: Full precision (critical for routing quality)
Single-Node Constraints
- 8 GPUs typically 80GB each (total 640GB)
- Reversibility: Activation memory flat
- TQP: Optimizer state minimal
- Result: 120B fits on single node
Extensions
Beyond Language Models
- Apply to vision transformers
- Reversible ViT architectures
- Sparse vision MoE
Multi-Node Scaling
- Reversibility + TQP for multi-node efficiency
- Distribute experts across nodes
- Router synchronization overhead
Continuous Growth
- Automated expansion triggers
- Dynamic expert addition
- Adaptive routing expansion
Pitfalls
- Reversible Layer Design: Not all operations have efficient inverses (e.g., softmax)
- State-Preserving Silent Failures: Routing collapse without obvious symptoms
- Quantization Degradation: Over-quantized base weights hurt quality
- Adapter Rank: Too low rank → insufficient expressiveness
- Expert Initialization: Wrong initialization → slow convergence or collapse
- Growth Timing: Premature expansion → unstable training
Related Methods
- RevNet (reversible residual networks)
- Gradient checkpointing (memory trade-off)
- Sparse MoE (Switch Transformer, GShard)
- Progressive training (curriculum)
- LoRA (low-rank adaptation)
Code & Resources
- Model family released
- Tokenizer released
- Training code released
- Paper: Experience report format
Applications
- Large model training on limited hardware
- Startup/academic training of 100B+ models
- Efficient MoE deployment
- Memory-constrained foundation model development
- Progressive scaling strategies
Activation Keywords
reversible training, sparse MoE, single-node training, memory-efficient, state-preserving growth, TQP optimizer, 120B model, low-rank adapters, expert quantization, reversible recurrence, growth lineage, flat activation memory