manta info

Show model info: tensor names, dtypes, shapes, and embedding capacity at a given LSB depth. Supports safetensors and GGUF formats (auto-detected by extension or magic bytes). For GGUF, capacity is reported only for GGML_TYPE_F32 tensors; quantized tensors are excluded and quantized-only files will show zero usable capacity.

Usage

manta info -m <MODEL> [-d <DEPTH>]

Options

Flag	Short	Default	Description
--model	-m	required	Path to model file (safetensors or GGUF)
--lsb-depth	-d	3	LSB depth for capacity calculation

Example

$ manta info -m phi-2.safetensors -d 3

Model: phi-2.safetensors
Target tensors: 64
LSB depth: 3

  model.layers.0.mlp.fc1.weight: 33554432 elements, 12582912 bytes capacity
  model.layers.0.mlp.fc2.weight: 16777216 elements, 6291456 bytes capacity
  ...

Total capacity: 402653184 bytes (384.00 MB)

Output

• Lists all target tensors (MLP layers, attention projections) with element count and byte capacity
• Shows total capacity across all tensors
• Non-target tensors (embeddings, norms, lm_head) are excluded
• GGUF quantized tensors are excluded because the embedder only supports F32 carriers

See Also

• capacity — machine-readable JSON output