← MIC coverage of Transformers model types

nomic-ai/nomic-embed-text-v1.5

huggingface-mic/nomic-embed-text-v1.5-mic@2cbaf2971908 ← nomic-ai/nomic-embed-text-v1.5@e9b6763023c6 · NomicBertModel · random weights · built before the shrink rule (no recorded recipe)

CI: no failure on main in 30 days · MIC failures dashboard

Original136.7Mparameters
Fixture2.0Mparameters
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.

layers

Original · 12
Fixture · 2
KindOriginalFixture
Layer122✓ 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
ComponentOriginalShareFixtureShare
layers113.3M82.9%82.4K4.0%
embeddings.word_embeddings23.4M17.1%2.0M95.9%
embeddings.LayerNorm1.5K0.0%1280.0%
embeddings.token_type_embeddings1.5K0.0%1280.0%

Preserved properties

ConfigPropertyOriginalFixture
modelRoPE typedefaultdefault✓ kept
modelmodel_typenomic_bertnomic_bert✓ kept
modelvocab_size3052830528✓ kept
modelhidden_actsilusilu✓ kept
modelmax_position_embeddings20482048✓ kept

Module classes

Instances of each class in the built model (prefix NomicBert omitted).

ClassOriginalFixture
Linear8414
Dropout255
LayerNorm255
Attention122
Layer122
MLP122
SiLUActivation122
Embedding22
ModuleList11
Embeddings11
Model11
RotaryEmbedding11

Config changes

9 fields changed, 57 unchanged.

FieldOriginalFixture
head_dim6416
hidden_size76864
intermediate_size3072128
n_embd76864
n_head124
n_inner3072128
n_layer122
num_attention_heads124
num_hidden_layers122