hidden_size: 128
num_hidden_layers: 2
num_attention_heads: 2
intermediate_size: 256
vocab_size: 1000
max_position_embeddings: 512
torch_dtype: bfloat16

Usage

Architecture: MiniCPM-o-2.6
Size: 37.70MB
Purpose: Testing and validation only (not trained)
Created for: Optimum-Intel integration tests

python

from transformers import AutoModel, AutoTokenizer

model = AutoModel.from_pretrained(
    "ahmedzk11/tiny-random-minicpm-o-2_6",
    trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained(
    "ahmedzk11/tiny-random-minicpm-o-2_6",
    trust_remote_code=True
)

Testing

This model is used in Optimum-Intel validation tests for MiniCPM-o-2.6 architecture.

Note: This model has random weights and is NOT suitable for actual inference tasks.