Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,63 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language: en
|
| 3 |
+
license: apache-2.0
|
| 4 |
+
tags:
|
| 5 |
+
- minicpm
|
| 6 |
+
- minicpm-o
|
| 7 |
+
- multimodal
|
| 8 |
+
- vision
|
| 9 |
+
- audio
|
| 10 |
+
- text-to-speech
|
| 11 |
+
- tiny-model
|
| 12 |
+
- random-weights
|
| 13 |
+
model-index:
|
| 14 |
+
- name: MiniCPM-o-2_6
|
| 15 |
+
results: []
|
| 16 |
+
---
|
| 17 |
+
|
| 18 |
+
# Tiny Random MiniCPM-o-2_6
|
| 19 |
+
|
| 20 |
+
This is a tiny random variant of the MiniCPM-o-2_6 model, specifically created for testing and validation purposes with Optimum-Intel.
|
| 21 |
+
|
| 22 |
+
## Model Details
|
| 23 |
+
|
| 24 |
+
- **Model Type**: Multimodal (Vision + Audio + Text-to-Speech)
|
| 25 |
+
- **Original Model**: [openbmb/MiniCPM-o-2_6](https://huggingface.co/openbmb/MiniCPM-o-2_6)
|
| 26 |
+
- **Size**: ~4.4MB (significantly reduced from original 160MB)
|
| 27 |
+
- **Architecture**: MiniCPMO (multimodal architecture)
|
| 28 |
+
- **Purpose**: Testing and validation with Optimum-Intel
|
| 29 |
+
|
| 30 |
+
## Configuration Changes
|
| 31 |
+
|
| 32 |
+
This tiny model has the following modifications from the original:
|
| 33 |
+
|
| 34 |
+
- `hidden_size`: 16 (reduced from original)
|
| 35 |
+
- `intermediate_size`: 32 (reduced from original)
|
| 36 |
+
- `num_hidden_layers`: 1 (reduced from original)
|
| 37 |
+
- `vocab_size`: 32000 (maintained for compatibility)
|
| 38 |
+
- Vision and Audio components significantly reduced in size
|
| 39 |
+
|
| 40 |
+
## Usage
|
| 41 |
+
|
| 42 |
+
```python
|
| 43 |
+
from transformers import AutoModel, AutoTokenizer, AutoProcessor
|
| 44 |
+
|
| 45 |
+
model = AutoModel.from_pretrained("your-username/tiny-random-MiniCPM-o-2_6", trust_remote_code=True)
|
| 46 |
+
tokenizer = AutoTokenizer.from_pretrained("your-username/tiny-random-MiniCPM-o-2_6", trust_remote_code=True)
|
| 47 |
+
processor = AutoProcessor.from_pretrained("your-username/tiny-random-MiniCPM-o-2_6", trust_remote_code=True)
|
| 48 |
+
```
|
| 49 |
+
|
| 50 |
+
## Validation
|
| 51 |
+
|
| 52 |
+
This model has been tested and validated with Optimum-Intel tests:
|
| 53 |
+
- ✅ `test_compare_to_transformers_14_minicpmo`
|
| 54 |
+
- ✅ `test_generate_utils_14_minicpmo`
|
| 55 |
+
- ✅ `test_model_can_be_loaded_after_saving_14_minicpmo`
|
| 56 |
+
|
| 57 |
+
## Limitations
|
| 58 |
+
|
| 59 |
+
⚠️ **Important**: This is a random weights model intended for testing purposes only. It will not produce meaningful outputs and should not be used for production applications.
|
| 60 |
+
|
| 61 |
+
## License
|
| 62 |
+
|
| 63 |
+
Apache 2.0
|