Text Generation
Transformers
Burmese
English
myanmar
burmese
llm
chat
instruction-following
conversational
autoregressive
Instructions to use amkyawdev/myanmar-ghost with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use amkyawdev/myanmar-ghost with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="amkyawdev/myanmar-ghost")# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("amkyawdev/myanmar-ghost", device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use amkyawdev/myanmar-ghost with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "amkyawdev/myanmar-ghost" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "amkyawdev/myanmar-ghost", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/amkyawdev/myanmar-ghost
- SGLang
How to use amkyawdev/myanmar-ghost with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "amkyawdev/myanmar-ghost" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "amkyawdev/myanmar-ghost", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "amkyawdev/myanmar-ghost" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "amkyawdev/myanmar-ghost", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use amkyawdev/myanmar-ghost with Docker Model Runner:
docker model run hf.co/amkyawdev/myanmar-ghost
File size: 8,835 Bytes
cfb5e7f | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 | """LIME explainer for Myanmar Ghost model.
Uses LIME (Local Interpretable Model-agnostic Explanations)
to explain individual predictions.
"""
import logging
from typing import Any, Callable, Dict, List, Optional, Tuple
import numpy as np
import torch
from lime.lime_text import LimeTextExplainer
logger = logging.getLogger(__name__)
class ThankingLIMEExplainer:
"""LIME-based explainer for Myanmar text classification."""
def __init__(
self,
model,
tokenizer,
class_names: Optional[List[str]] = None,
kernel_width: float = 25.0,
):
"""
Args:
model: PyTorch model
tokenizer: Tokenizer
class_names: Names for output classes
kernel_width: Kernel width for LIME
"""
self.model = model
self.tokenizer = tokenizer
self.class_names = class_names or [
"negative", "neutral", "positive", "sarcastic"
]
self.model.eval()
# Create LIME explainer
self.explainer = LimeTextExplainer(
class_names=self.class_names,
kernel_width=kernel_width,
verbose=False,
)
def _predict_proba(self, texts: List[str]) -> np.ndarray:
"""Prediction function for LIME.
Args:
texts: List of text strings
Returns:
Probability array (n_samples, n_classes)
"""
# Tokenize
encoding = self.tokenizer(
texts,
padding=True,
truncation=True,
max_length=128,
return_tensors="pt",
)
input_ids = encoding["input_ids"]
attention_mask = encoding["attention_mask"]
with torch.no_grad():
outputs = self.model(input_ids, attention_mask)
if hasattr(outputs, "logits"):
logits = outputs.logits
else:
logits = outputs
probs = torch.softmax(logits, dim=-1)
return probs.cpu().numpy()
def explain(
self,
text: str,
num_features: int = 10,
num_samples: int = 5000,
top_labels: int = 4,
) -> Any:
"""Explain a single text prediction.
Args:
text: Myanmar text to explain
num_features: Number of features to show
num_samples: Number of samples for LIME
top_labels: Number of top labels to explain
Returns:
LIME Explanation object
"""
logger.info(f"Explaining with LIME: {text[:50]}...")
explanation = self.explainer.explain_instance(
text,
self._predict_proba,
num_features=num_features,
num_samples=num_samples,
top_labels=top_labels,
)
return explanation
def get_word_importance(
self,
text: str,
class_index: Optional[int] = None,
num_features: int = 10,
) -> List[Tuple[str, float]]:
"""Get word importance for a specific class.
Args:
text: Myanmar text
class_index: Class index (None = predicted class)
num_features: Number of top features
Returns:
List of (word, importance) tuples
"""
explanation = self.explain(text, num_features=num_features)
if class_index is None:
# Use predicted class
class_index = explanation.available_labels()[0]
exp_list = explanation.as_list(label=class_index)
return exp_list
def visualize(
self,
explanation: Any,
output_path: Optional[str] = None,
) -> str:
"""Generate text visualization of explanation.
Args:
explanation: LIME Explanation object
output_path: Optional path to save
Returns:
Visualization text
"""
output = "\n" + "=" * 60 + "\n"
output += "LIME EXPLANATION\n"
output += "=" * 60 + "\n"
for label in explanation.available_labels()[:3]:
label_name = self.class_names[label]
output += f"\n{label_name.upper()}:\n"
output += "-" * 40 + "\n"
for word, weight in explanation.as_list(label=label):
sign = "+" if weight > 0 else ""
output += f" {word}: {sign}{weight:.4f}\n"
print(output)
if output_path:
with open(output_path, "w", encoding="utf-8") as f:
f.write(output)
logger.info(f"Visualization saved to {output_path}")
return output
def batch_explain(
self,
texts: List[str],
num_features: int = 10,
) -> List[Dict[str, Any]]:
"""Explain multiple texts.
Args:
texts: List of texts
num_features: Number of features per explanation
Returns:
List of explanation dictionaries
"""
results = []
for text in texts:
explanation = self.explain(text, num_features=num_features)
result = {
"text": text,
"predicted_class": self.class_names[
explanation.available_labels()[0]
],
"explanations": {},
}
for label in explanation.available_labels():
result["explanations"][self.class_names[label]] = {
word: weight
for word, weight in explanation.as_list(label=label)
}
results.append(result)
return results
class SegmentLevelLIME:
"""LIME with Myanmar-specific segmentation."""
def __init__(
self,
model,
tokenizer,
segment_syllables: bool = True,
):
"""
Args:
model: PyTorch model
tokenizer: Tokenizer
segment_syllables: Segment by Myanmar syllables
"""
self.model = model
self.tokenizer = tokenizer
self.segment_syllables = segment_syllables
def _segment_text(self, text: str) -> List[str]:
"""Segment text into interpretable units.
For Myanmar, this can be syllable or word level.
"""
if self.segment_syllables:
# Simple syllable segmentation
# Myanmar syllables end with vowel markers or consonants
segments = []
current = ""
for char in text:
current += char
# Check for syllable boundary (simplified)
if char in "း့်ှင်း":
segments.append(current)
current = ""
if current:
segments.append(current)
return segments if segments else [text]
else:
return text.split()
def _join_segments(self, segments: List[str]) -> str:
"""Join segments back to text."""
return "".join(segments)
def explain(
self,
text: str,
num_samples: int = 1000,
) -> Dict[str, Any]:
"""Explain text with syllable-level segmentation.
Args:
text: Myanmar text
num_samples: Number of LIME samples
Returns:
Explanation dictionary
"""
segments = self._segment_text(text)
logger.info(f"Segmented into {len(segments)} units")
# Use standard LIME with segmented text
base_explainer = ThankingLIMEExplainer(
self.model,
self.tokenizer,
)
explanation = base_explainer.explain(
text,
num_features=len(segments),
num_samples=num_samples,
)
return {
"text": text,
"segments": segments,
"explanation": explanation,
"word_importance": base_explainer.get_word_importance(text),
}
def create_lime_explainer(
model,
tokenizer,
class_names: Optional[List[str]] = None,
) -> ThankingLIMEExplainer:
"""Factory function to create LIME explainer."""
return ThankingLIMEExplainer(
model=model,
tokenizer=tokenizer,
class_names=class_names,
)
if __name__ == "__main__":
print("ThankingLIMEExplainer loaded")
print("Use create_lime_explainer() to create an explainer")
|