notlikejoe commited on
Commit
3ff0e39
·
verified ·
1 Parent(s): 45af4b6

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +161 -0
README.md ADDED
@@ -0,0 +1,161 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ base_model: openbmb/MiniCPM-o-2_6
4
+ tags:
5
+ - vision
6
+ - text-generation
7
+ - multimodal
8
+ - minicpm
9
+ - tiny-model
10
+ - testing
11
+ - optimum-intel
12
+ pipeline_tag: text-generation
13
+ library_name: transformers
14
+ ---
15
+
16
+ # tiny-random-MiniCPM-o-2_6
17
+
18
+ A minimal, randomly initialized version of MiniCPM-o-2_6 designed for testing and development purposes. This model maintains the same architecture as the original MiniCPM-o-2_6 but with drastically reduced dimensions to create a lightweight test model.
19
+
20
+ ## Model Details
21
+
22
+ ### Model Description
23
+
24
+ This is a tiny, randomly initialized version of the MiniCPM-o-2_6 multimodal model. It was created by scaling down the original model's dimensions while preserving the architecture structure. The model is intended for:
25
+
26
+ - Testing and development workflows
27
+ - Integration testing with Optimum-Intel
28
+ - Quick prototyping and experimentation
29
+ - CI/CD pipelines requiring lightweight models
30
+
31
+ **⚠️ Important:** This model is randomly initialized and should NOT be used for production inference. It is designed solely for testing purposes.
32
+
33
+ ### Model Architecture
34
+
35
+ The model maintains the same architecture as MiniCPM-o-2_6 but with reduced dimensions:
36
+
37
+ **Language Model (LLM):**
38
+ - `hidden_size`: 40
39
+ - `num_hidden_layers`: 1
40
+ - `num_attention_heads`: 4
41
+ - `num_key_value_heads`: 2
42
+ - `intermediate_size`: 16
43
+ - `max_position_embeddings`: 128
44
+ - `vocab_size`: 151,700
45
+
46
+ **Vision Component:**
47
+ - `hidden_size`: 16
48
+ - `num_hidden_layers`: 1
49
+ - `num_attention_heads`: 4
50
+ - `intermediate_size`: 8
51
+ - `patch_size`: 14
52
+
53
+ **Audio/TTS Components:**
54
+ - Audio: Disabled (`init_audio: false`)
55
+ - TTS: Disabled (`init_tts: false`)
56
+
57
+ ### Model Size
58
+
59
+ - **Total Parameters**: ~6.17M
60
+ - **Model Size**: ~12.4 MB (on disk)
61
+ - **Precision**: bfloat16
62
+
63
+ ## Usage
64
+
65
+ ### Basic Usage
66
+
67
+ ```python
68
+ from transformers import AutoModel, AutoTokenizer, AutoProcessor
69
+ import torch
70
+ from PIL import Image
71
+
72
+ # Load model and tokenizer
73
+ model_id = "notlikejoe/tiny-random-MiniCPM-o-2_6"
74
+ model = AutoModel.from_pretrained(model_id, trust_remote_code=True, torch_dtype=torch.bfloat16)
75
+ tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
76
+ processor = AutoProcessor.from_pretrained(model_id, trust_remote_code=True)
77
+
78
+ # Prepare inputs
79
+ text = "Hello, how are you?"
80
+ image = Image.new('RGB', (224, 224), color='red') # Dummy image
81
+
82
+ # Process inputs
83
+ inputs = processor(text=text, images=image, return_tensors="pt")
84
+
85
+ # Forward pass
86
+ model.eval()
87
+ with torch.no_grad():
88
+ outputs = model(**inputs)
89
+ ```
90
+
91
+ ### With Optimum-Intel
92
+
93
+ This model is compatible with Optimum-Intel for OpenVINO optimization:
94
+
95
+ ```python
96
+ from optimum.intel import OVModelForCausalLM
97
+ from transformers import AutoTokenizer
98
+
99
+ model_id = "notlikejoe/tiny-random-MiniCPM-o-2_6"
100
+
101
+ # Export to OpenVINO format
102
+ ov_model = OVModelForCausalLM.from_pretrained(
103
+ model_id,
104
+ export=True,
105
+ trust_remote_code=True
106
+ )
107
+
108
+ # Use for inference
109
+ tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
110
+ ```
111
+
112
+ ## Model Validation
113
+
114
+ The model has been validated to ensure:
115
+
116
+ ✅ Model loads successfully from Hugging Face
117
+ ✅ Config, tokenizer, and processor load correctly
118
+ ✅ Model structure matches expected architecture
119
+ ✅ Compatible with Optimum-Intel export
120
+ ✅ Forward pass completes without errors
121
+
122
+ ## Limitations
123
+
124
+ 1. **Random Initialization**: This model is randomly initialized and will not produce meaningful outputs
125
+ 2. **Reduced Dimensions**: The model dimensions are minimal and may not capture complex patterns
126
+ 3. **Testing Only**: This model is intended for testing and development, not production use
127
+ 4. **Limited Vocabulary**: The vocabulary has been reduced to 2000 entries for size optimization
128
+
129
+ ## Training Details
130
+
131
+ This model was not trained. It is a randomly initialized, dimensionally-reduced version of MiniCPM-o-2_6 created for testing purposes.
132
+
133
+ ### Training Data
134
+
135
+ N/A - Model is randomly initialized.
136
+
137
+ ## Evaluation
138
+
139
+ This model is not intended for evaluation on standard benchmarks as it is randomly initialized.
140
+
141
+ ## Citation
142
+
143
+ If you use this model, please cite the original MiniCPM-o-2_6 model:
144
+
145
+ ```bibtex
146
+ @misc{minicpm-o-2_6,
147
+ title={MiniCPM-o-2_6},
148
+ author={OpenBMB},
149
+ year={2024},
150
+ howpublished={\url{https://huggingface.co/openbmb/MiniCPM-o-2_6}}
151
+ }
152
+ ```
153
+
154
+ ## Model Card Contact
155
+
156
+ For questions or issues related to this model, please open an issue in the repository.
157
+
158
+ ## License
159
+
160
+ This model is licensed under the Apache 2.0 License, same as the base model.
161
+