Update README.md
Browse files
README.md
CHANGED
|
@@ -25,25 +25,33 @@ pipeline_tag: text-generation
|
|
| 25 |
|
| 26 |
### Model Description
|
| 27 |
|
| 28 |
-
|
| 29 |
|
| 30 |
This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
|
| 31 |
|
| 32 |
-
- **Developed by:** [
|
| 33 |
-
- **Funded by [
|
| 34 |
-
|
| 35 |
-
- **
|
| 36 |
-
|
| 37 |
-
- **
|
| 38 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 39 |
|
| 40 |
### Model Sources [optional]
|
| 41 |
|
| 42 |
<!-- Provide the basic links for the model. -->
|
| 43 |
|
| 44 |
-
- **Repository:** [
|
| 45 |
-
- **Paper [optional]:** [
|
| 46 |
-
|
|
|
|
|
|
|
|
|
|
| 47 |
|
| 48 |
## Uses
|
| 49 |
|
|
@@ -53,29 +61,75 @@ This is the model card of a 🤗 transformers model that has been pushed on the
|
|
| 53 |
|
| 54 |
<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
|
| 55 |
|
| 56 |
-
[
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 57 |
|
| 58 |
### Downstream Use [optional]
|
| 59 |
|
| 60 |
<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
|
| 61 |
|
| 62 |
-
[
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 63 |
|
| 64 |
### Out-of-Scope Use
|
| 65 |
|
| 66 |
<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
|
| 67 |
|
| 68 |
-
[
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 69 |
|
| 70 |
## Bias, Risks, and Limitations
|
| 71 |
|
| 72 |
<!-- This section is meant to convey both technical and sociotechnical limitations. -->
|
| 73 |
|
| 74 |
-
[
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 75 |
|
| 76 |
### Recommendations
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 77 |
|
| 78 |
-
<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
|
| 79 |
|
| 80 |
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
|
| 81 |
|
|
@@ -83,38 +137,76 @@ Users (both direct and downstream) should be made aware of the risks, biases and
|
|
| 83 |
|
| 84 |
Use the code below to get started with the model.
|
| 85 |
|
| 86 |
-
[
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 87 |
|
| 88 |
## Training Details
|
| 89 |
|
| 90 |
### Training Data
|
| 91 |
|
| 92 |
-
<!--
|
|
|
|
|
|
|
|
|
|
|
|
|
| 93 |
|
| 94 |
-
[More Information Needed]
|
| 95 |
|
| 96 |
### Training Procedure
|
| 97 |
|
| 98 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 99 |
|
| 100 |
#### Preprocessing [optional]
|
| 101 |
|
| 102 |
-
[
|
| 103 |
|
| 104 |
|
| 105 |
#### Training Hyperparameters
|
| 106 |
|
| 107 |
-
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 108 |
|
| 109 |
#### Speeds, Sizes, Times [optional]
|
| 110 |
|
| 111 |
<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
|
| 112 |
|
| 113 |
-
[
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 114 |
|
| 115 |
## Evaluation
|
| 116 |
|
| 117 |
-
|
|
|
|
|
|
|
|
|
|
| 118 |
|
| 119 |
### Testing Data, Factors & Metrics
|
| 120 |
|
|
@@ -128,17 +220,24 @@ Use the code below to get started with the model.
|
|
| 128 |
|
| 129 |
<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
|
| 130 |
|
| 131 |
-
[
|
|
|
|
|
|
|
| 132 |
|
| 133 |
#### Metrics
|
| 134 |
|
| 135 |
<!-- These are the evaluation metrics being used, ideally with a description of why. -->
|
| 136 |
|
| 137 |
-
[More Information Needed]
|
| 138 |
-
|
| 139 |
### Results
|
| 140 |
|
| 141 |
-
[
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 142 |
|
| 143 |
#### Summary
|
| 144 |
|
|
@@ -152,7 +251,15 @@ Use the code below to get started with the model.
|
|
| 152 |
|
| 153 |
## Environmental Impact
|
| 154 |
|
| 155 |
-
<!--
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 156 |
|
| 157 |
Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
|
| 158 |
|
|
@@ -162,7 +269,16 @@ Carbon emissions can be estimated using the [Machine Learning Impact calculator]
|
|
| 162 |
- **Compute Region:** [More Information Needed]
|
| 163 |
- **Carbon Emitted:** [More Information Needed]
|
| 164 |
|
| 165 |
-
## Technical Specifications [
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 166 |
|
| 167 |
### Model Architecture and Objective
|
| 168 |
|
|
@@ -170,7 +286,10 @@ Carbon emissions can be estimated using the [Machine Learning Impact calculator]
|
|
| 170 |
|
| 171 |
### Compute Infrastructure
|
| 172 |
|
| 173 |
-
[
|
|
|
|
|
|
|
|
|
|
| 174 |
|
| 175 |
#### Hardware
|
| 176 |
|
|
@@ -186,17 +305,30 @@ Carbon emissions can be estimated using the [Machine Learning Impact calculator]
|
|
| 186 |
|
| 187 |
**BibTeX:**
|
| 188 |
|
| 189 |
-
[
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 190 |
|
| 191 |
**APA:**
|
| 192 |
|
| 193 |
-
[
|
|
|
|
| 194 |
|
| 195 |
## Glossary [optional]
|
| 196 |
|
| 197 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 198 |
|
| 199 |
-
|
|
|
|
| 200 |
|
| 201 |
## More Information [optional]
|
| 202 |
|
|
@@ -204,8 +336,18 @@ Carbon emissions can be estimated using the [Machine Learning Impact calculator]
|
|
| 204 |
|
| 205 |
## Model Card Authors [optional]
|
| 206 |
|
| 207 |
-
[
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 208 |
|
| 209 |
## Model Card Contact
|
| 210 |
|
| 211 |
-
[
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 25 |
|
| 26 |
### Model Description
|
| 27 |
|
| 28 |
+
Provide a longer summary of what this model is. -->
|
| 29 |
|
| 30 |
This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
|
| 31 |
|
| 32 |
+
- **Developed by:** [Zubair Arshad Raoter CEO of SafeGuard.AI]
|
| 33 |
+
- **Funded by [optionl]:** [Self-funded using 100% free and open-source resources
|
| 34 |
+
]
|
| 35 |
+
- **Shared by [Zubair Arshad Raoter
|
| 36 |
+
]:** [More Information Needed]
|
| 37 |
+
- **Model type:** [ (Text Generation)
|
| 38 |
+
]
|
| 39 |
+
- **Language(s) (NLP):** [ (English)
|
| 40 |
+
]
|
| 41 |
+
- **License:** [MIT]
|
| 42 |
+
- **Finetuned from model [optional]:** [distilGPT2 by Hugging Face
|
| 43 |
+
]
|
| 44 |
|
| 45 |
### Model Sources [optional]
|
| 46 |
|
| 47 |
<!-- Provide the basic links for the model. -->
|
| 48 |
|
| 49 |
+
- **Repository:** [https://huggingface.co/Zubiiiiiii294/textbuddy]
|
| 50 |
+
- **Paper [optional]:** [No formal paper yet. This is an independent experimental build.
|
| 51 |
+
]
|
| 52 |
+
- **Demo [optional]:** [Demo under development. Will be shared soon.
|
| 53 |
+
|
| 54 |
+
]
|
| 55 |
|
| 56 |
## Uses
|
| 57 |
|
|
|
|
| 61 |
|
| 62 |
<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
|
| 63 |
|
| 64 |
+
[For learning how LLMs work
|
| 65 |
+
|
| 66 |
+
For generating text, stories, content ideas
|
| 67 |
+
|
| 68 |
+
For experimenting with text prompts
|
| 69 |
+
|
| 70 |
+
For research, education & testing small-scale NLP models
|
| 71 |
+
|
| 72 |
+
]
|
| 73 |
|
| 74 |
### Downstream Use [optional]
|
| 75 |
|
| 76 |
<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
|
| 77 |
|
| 78 |
+
[This model can be further fine-tuned for:
|
| 79 |
+
|
| 80 |
+
Chatbots
|
| 81 |
+
|
| 82 |
+
Educational writing assistants
|
| 83 |
+
|
| 84 |
+
Text-based games
|
| 85 |
+
|
| 86 |
+
Idea generation tools
|
| 87 |
+
]
|
| 88 |
|
| 89 |
### Out-of-Scope Use
|
| 90 |
|
| 91 |
<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
|
| 92 |
|
| 93 |
+
[Not recommended for sensitive content
|
| 94 |
+
|
| 95 |
+
Not for decision-making in medical, legal, or security domains
|
| 96 |
+
|
| 97 |
+
Not for multi-language tasks (supports English only)
|
| 98 |
+
]
|
| 99 |
|
| 100 |
## Bias, Risks, and Limitations
|
| 101 |
|
| 102 |
<!-- This section is meant to convey both technical and sociotechnical limitations. -->
|
| 103 |
|
| 104 |
+
[This model may reflect biases present in the Wikitext dataset
|
| 105 |
+
|
| 106 |
+
Text output may be inaccurate or incomplete
|
| 107 |
+
|
| 108 |
+
Doesn’t understand emotional or moral context
|
| 109 |
+
|
| 110 |
+
Doesn’t support multilingual tasks
|
| 111 |
+
|
| 112 |
+
Not suitable for commercial or mission-critical use (yet!)
|
| 113 |
+
This model may reflect biases present in the Wikitext dataset
|
| 114 |
+
|
| 115 |
+
Text output may be inaccurate or incomplete
|
| 116 |
+
|
| 117 |
+
Doesn’t understand emotional or moral context
|
| 118 |
+
|
| 119 |
+
Doesn’t support multilingual tasks
|
| 120 |
+
|
| 121 |
+
Not suitable for commercial or mission-critical use (yet!)
|
| 122 |
+
]
|
| 123 |
|
| 124 |
### Recommendations
|
| 125 |
+
Use in controlled environments (research, testing, education)
|
| 126 |
+
|
| 127 |
+
Don’t rely on its outputs as factual
|
| 128 |
+
|
| 129 |
+
Carefully evaluate before using in any product or service
|
| 130 |
+
|
| 131 |
+
Encourage transparency on limitations if shared publicly
|
| 132 |
|
|
|
|
| 133 |
|
| 134 |
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
|
| 135 |
|
|
|
|
| 137 |
|
| 138 |
Use the code below to get started with the model.
|
| 139 |
|
| 140 |
+
[from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 141 |
+
|
| 142 |
+
model_name = "Zubiiiiiii294/textbuddy"
|
| 143 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
| 144 |
+
model = AutoModelForCausalLM.from_pretrained(model_name)
|
| 145 |
+
|
| 146 |
+
prompt = "The future of AI in Pakistan is"
|
| 147 |
+
inputs = tokenizer(prompt, return_tensors="pt")
|
| 148 |
+
outputs = model.generate(**inputs, max_length=50)
|
| 149 |
+
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
|
| 150 |
+
]
|
| 151 |
|
| 152 |
## Training Details
|
| 153 |
|
| 154 |
### Training Data
|
| 155 |
|
| 156 |
+
<!-- Dataset: WikiText-2 (raw, preprocessed)
|
| 157 |
+
|
| 158 |
+
Source: Hugging Face Datasets
|
| 159 |
+
|
| 160 |
+
Type: General English knowledge, Wikipedia-style text
|
| 161 |
|
|
|
|
| 162 |
|
| 163 |
### Training Procedure
|
| 164 |
|
| 165 |
+
Platform: Google Colab
|
| 166 |
+
|
| 167 |
+
Tokenizer: distilgpt2 (EOS padded)
|
| 168 |
+
|
| 169 |
+
Model: distilgpt2
|
| 170 |
+
|
| 171 |
+
Run-time: ~30 minutes total
|
| 172 |
+
|
| 173 |
+
Finetuning: Light-touch to demonstrate model building
|
| 174 |
+
|
| 175 |
+
|
| 176 |
|
| 177 |
#### Preprocessing [optional]
|
| 178 |
|
| 179 |
+
[]
|
| 180 |
|
| 181 |
|
| 182 |
#### Training Hyperparameters
|
| 183 |
|
| 184 |
+
- Precision: fp32
|
| 185 |
+
|
| 186 |
+
Epochs: 1-2 (basic)
|
| 187 |
+
|
| 188 |
+
GPU: Colab T4 GPU
|
| 189 |
+
|
| 190 |
+
Batch size: Small (default)
|
| 191 |
+
|
| 192 |
|
| 193 |
#### Speeds, Sizes, Times [optional]
|
| 194 |
|
| 195 |
<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
|
| 196 |
|
| 197 |
+
[Model size: 0.1B (117M parameters)
|
| 198 |
+
|
| 199 |
+
File size: ~328MB
|
| 200 |
+
|
| 201 |
+
Upload: Hugging Face model repo
|
| 202 |
+
]
|
| 203 |
|
| 204 |
## Evaluation
|
| 205 |
|
| 206 |
+
Testing Data
|
| 207 |
+
|
| 208 |
+
Manual test prompts used during Colab testing
|
| 209 |
+
|
| 210 |
|
| 211 |
### Testing Data, Factors & Metrics
|
| 212 |
|
|
|
|
| 220 |
|
| 221 |
<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
|
| 222 |
|
| 223 |
+
[Focused on simple generation
|
| 224 |
+
|
| 225 |
+
No subpopulation or fairness breakdown]
|
| 226 |
|
| 227 |
#### Metrics
|
| 228 |
|
| 229 |
<!-- These are the evaluation metrics being used, ideally with a description of why. -->
|
| 230 |
|
|
|
|
|
|
|
| 231 |
### Results
|
| 232 |
|
| 233 |
+
[Model is functional and responsive
|
| 234 |
+
|
| 235 |
+
Good for basic prompt-based text generation
|
| 236 |
+
|
| 237 |
+
Excellent for learning & showcasing LLMs from Pakistan
|
| 238 |
+
|
| 239 |
+
|
| 240 |
+
]
|
| 241 |
|
| 242 |
#### Summary
|
| 243 |
|
|
|
|
| 251 |
|
| 252 |
## Environmental Impact
|
| 253 |
|
| 254 |
+
<!-- Hardware Type: Google Colab (shared cloud GPU)
|
| 255 |
+
|
| 256 |
+
Hours used: ~2 hours
|
| 257 |
+
|
| 258 |
+
Cloud Provider: Google
|
| 259 |
+
|
| 260 |
+
Compute Region: Not specified
|
| 261 |
+
|
| 262 |
+
Carbon Emitted: Minimal (low-resource training)
|
| 263 |
|
| 264 |
Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
|
| 265 |
|
|
|
|
| 269 |
- **Compute Region:** [More Information Needed]
|
| 270 |
- **Carbon Emitted:** [More Information Needed]
|
| 271 |
|
| 272 |
+
## Technical Specifications [ Model Architecture and Objective
|
| 273 |
+
|
| 274 |
+
Base model: distilgpt2
|
| 275 |
+
|
| 276 |
+
Task: Text generation
|
| 277 |
+
|
| 278 |
+
Objective: Predict next word/token in prompt-based context
|
| 279 |
+
|
| 280 |
+
|
| 281 |
+
]
|
| 282 |
|
| 283 |
### Model Architecture and Objective
|
| 284 |
|
|
|
|
| 286 |
|
| 287 |
### Compute Infrastructure
|
| 288 |
|
| 289 |
+
[Cloud: Google Colab (T4 GPU)
|
| 290 |
+
|
| 291 |
+
Software: Hugging Face Transformers, PyTorch, Datasets
|
| 292 |
+
]
|
| 293 |
|
| 294 |
#### Hardware
|
| 295 |
|
|
|
|
| 305 |
|
| 306 |
**BibTeX:**
|
| 307 |
|
| 308 |
+
[@misc{zubair2025textbuddy,
|
| 309 |
+
title={TextBuddy: Pakistan’s 1st Open-Source Chat AI Model},
|
| 310 |
+
author={Zubair Arshad Raoter},
|
| 311 |
+
year={2025},
|
| 312 |
+
howpublished={\url{https://huggingface.co/Zubiiiiiii294/textbuddy}},
|
| 313 |
+
note={Self-trained via Colab}
|
| 314 |
+
}
|
| 315 |
+
]
|
| 316 |
|
| 317 |
**APA:**
|
| 318 |
|
| 319 |
+
[Zubair Arshad Raoter. (2025). TextBuddy: Pakistan’s 1st Open-Source Chat AI Model. Hugging Face. Retrieved from https://huggingface.co/Zubiiiiiii294/textbuddy
|
| 320 |
+
]
|
| 321 |
|
| 322 |
## Glossary [optional]
|
| 323 |
|
| 324 |
+
[LLM – Large Language Model
|
| 325 |
+
|
| 326 |
+
Tokenization – Breaking down sentences into model-readable pieces
|
| 327 |
+
|
| 328 |
+
Finetuning – Training a model from an existing base on a custom dataset
|
| 329 |
|
| 330 |
+
|
| 331 |
+
]
|
| 332 |
|
| 333 |
## More Information [optional]
|
| 334 |
|
|
|
|
| 336 |
|
| 337 |
## Model Card Authors [optional]
|
| 338 |
|
| 339 |
+
[Zubair Arshad Raoter – CEO, SafeGuard.AI
|
| 340 |
+
|
| 341 |
+
Contributor to: Youth AI Movement in Pakistan
|
| 342 |
+
|
| 343 |
+
Model inspired by ChatGPT & Grok logic
|
| 344 |
+
|
| 345 |
+
]
|
| 346 |
|
| 347 |
## Model Card Contact
|
| 348 |
|
| 349 |
+
[Email: [[email protected]]
|
| 350 |
+
|
| 351 |
+
LinkedIn: www.linkedin.com/in/zubair-arshad-raoter-7b1210289
|
| 352 |
+
|
| 353 |
+
Location: Karachi, Pakistan ]
|