Text Classification
Transformers
Safetensors
English
basedecision_dyno
feature-extraction
decision-model
system-one
decision-index
multiple-choice
calibration
read-twice
custom-code
lora-merged
custom_code
Instructions to use onlyaady/BaseDecision-Dyno with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use onlyaady/BaseDecision-Dyno with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-classification", model="onlyaady/BaseDecision-Dyno", trust_remote_code=True)# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("onlyaady/BaseDecision-Dyno", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
File size: 3,366 Bytes
4c87263 | 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 | """Strict model-owned repetition, with no implicit switch to single-read mode."""
from .prompt_dyno import state_messages, options_of, render_value, DEFAULT_QUESTION, INSTRUCTION, ANSWER, render
READ_AGAIN = 'Read the same input again before answering:\n\n'
class LengthError(ValueError):
pass
def render_twice(processor, state, question, labels):
messages, media = state_messages(state)
if media: raise ValueError('This release is text-only')
keys, texts = options_of(question)
head = (render_value(question['instructions'])
if question.get('instructions') is not None else DEFAULT_QUESTION)
options = '\n'.join(f'{labels[i]}. {value}' for i, value in enumerate(texts))
ask = INSTRUCTION + f'\n\nQuestion: {head}\nOptions:\n{options}'
first = [dict(m) for m in messages] + [{'role': 'user', 'content': ask}]
again = [dict(m) for m in first if m.get('role') != 'system']
if again[0]['role'] == 'user':
content = again[0].get('content')
content = ([{'type': 'text', 'text': READ_AGAIN}] + content
if isinstance(content, list) else READ_AGAIN + (content or ''))
again[0] = {**again[0], 'content': content}
else: again.insert(0, {'role': 'user', 'content': READ_AGAIN.rstrip()})
text = processor.apply_chat_template(first + again, tokenize=False,
add_generation_prompt=True, enable_thinking=False)
return text + ANSWER, keys
def encode(processor, labels, label_ids, request, mode='default', limit=16384):
if mode not in ('default', 'flash'): raise ValueError('mode must be default or flash')
if not isinstance(request, dict) or 'state' not in request:
raise ValueError('Pass a structured decision request with state and question(s)')
if request.get('images'): raise ValueError('This release is text-only')
questions = request.get('questions')
if questions is None: questions = {'q': request['question']}
if not isinstance(questions, dict) or not questions: raise ValueError('Empty questions')
_, once, media = render(processor, request['state'], questions, labels)
if media: raise ValueError('This release is text-only')
output = []
for key, question in questions.items():
text, keys = (render_twice(processor, request['state'], question, labels)
if mode == 'default' else once[key])
if keys != once[key][1]: raise ValueError('Option order changed')
if len(keys) > len(label_ids): raise ValueError('Unsupported option count')
ids = processor.tokenizer.encode(text, add_special_tokens=False)
if not 1 <= len(ids) <= limit:
raise LengthError(f'{mode} rendered sequence has {len(ids)} tokens; limit {limit}. '
'No truncation or automatic flash fallback. Shorten input or explicitly request mode="flash".')
# Check the actual continuation boundary, including user-controlled text.
for label, token in zip(labels[:len(keys)], label_ids[:len(keys)]):
if processor.tokenizer.encode(text+' '+label, add_special_tokens=False) != ids+[token]:
raise ValueError('Answer label is not one token at the actual prompt boundary')
output.append({'qid': key, 'ids': ids, 'keys': keys, 'kind': question['type']})
return output
|