poolside/Laguna-XS.2
Original33.4Bparameters
Fixture13.2Mparameters
Layer stacks1compared
Findings0to review
✓ Same architecture, smaller
Every module class, layer kind and checked property of the original is in the fixture.
Layer stacks
One square per layer, colored by kind: the layer's module classes and type attributes. Hover a square for its index and kind.
model.layers
Original · 40
Fixture · 2
| Kind | Original | Fixture | |
|---|---|---|---|
| no Experts | 1 | 1 | ✓ kept |
| Experts, SparseMoeBlock, TopKRouter | 39 | 1 | ✓ kept |
Where the parameters are
Share of each model's parameters per component. Vocabulary and patch sizes are kept, so embeddings usually take a larger share of the fixture.
Original
Fixture
| Component | Original | Share | Fixture | Share |
|---|---|---|---|---|
| model.layers | 33.0B | 98.8% | 319.0K | 2.4% |
| model.embed_tokens | 205.5M | 0.6% | 6.4M | 48.8% |
| lm_head | 205.5M | 0.6% | 6.4M | 48.8% |
| model.norm | 2.0K | 0.0% | 64 | 0.0% |
Preserved properties
| Config | Property | Original | Fixture | |
|---|---|---|---|---|
| model | attention style | GQA (48 heads, 8 KV) | GQA (4 heads, 2 KV) | ✓ kept |
| model | MoE routing | top-8 of 256 experts | top-2 of 4 experts | ✓ kept |
| model | model_type | laguna | laguna | ✓ kept |
| model | vocab_size | 100352 | 100352 | ✓ kept |
| model | tie_word_embeddings | False | False | ✓ kept |
| model | hidden_act | silu | silu | ✓ kept |
| model | max_position_embeddings | 262144 | 262144 | ✓ kept |
Module classes
Instances of each class in the built model (prefix Laguna omitted).
| Class | Original | Fixture |
|---|---|---|
| Linear | 321 | 17 |
| RMSNorm | 161 | 9 |
| SiLUActivation | 79 | 3 |
| Attention | 40 | 2 |
| DecoderLayer | 40 | 2 |
| MLP | 40 | 2 |
| Experts | 39 | 1 |
| SparseMoeBlock | 39 | 1 |
| TopKRouter | 39 | 1 |
| Embedding | 1 | 1 |
| ForCausalLM | 1 | 1 |
| Model | 1 | 1 |
| RotaryEmbedding | 1 | 1 |
| ModuleList | 1 | 1 |
Config changes
13 fields changed, 31 unchanged.
| Field | Original | Fixture |
|---|---|---|
| head_dim | 128 | 16 |
| hidden_size | 2048 | 64 |
| intermediate_size | 8192 | 128 |
| layer_types | 40 entries: 30× sliding_attention, 10× full_attention | ["full_attention", "sliding_attention"] |
| mlp_layer_types | 40 entries: 39× sparse, 1× dense | ["dense", "sparse"] |
| moe_intermediate_size | 512 | 32 |
| num_attention_heads | 48 | 4 |
| num_attention_heads_per_layer | 40 entries: 30× 64, 10× 48 | [48, 64] |
| num_experts | 256 | 4 |
| num_experts_per_tok | 8 | 2 |
| num_hidden_layers | 40 | 2 |
| num_key_value_heads | 8 | 2 |
| shared_expert_intermediate_size | 512 | 128 |