generation/stopping_criteria

• generation/stopping_criteria
  • .StoppingCriteria
    • ._call(input_ids, scores) ⇒ Array.<boolean>
  • .StoppingCriteriaList
    • new StoppingCriteriaList()
    • .push(item)
    • .extend(items)
  • .MaxLengthCriteria
    • new MaxLengthCriteria(max_length, [max_position_embeddings])
  • .EosTokenCriteria
    • new EosTokenCriteria(eos_token_id)
    • ._call(input_ids, scores) ⇒ Array.<boolean>
  • .InterruptableStoppingCriteria

---

generation/stopping_criteria.StoppingCriteria

Abstract base class for all stopping criteria that can be applied during generation.

Kind: static class of generation/stopping_criteria

---

stoppingCriteria._call(input_ids, scores) ⇒ <code> Array. < boolean > </code>

Kind: instance method of StoppingCriteria
Returns: Array.<boolean> - A list of booleans indicating whether each sequence should be stopped.

Param	Type	Description
input_ids	Array.<Array<number>>	(number[][] of shape (batch_size, sequence_length)): Indices of input sequence tokens in the vocabulary.
scores	Array.<Array<number>>	scores (number[][] of shape (batch_size, config.vocab_size)): Prediction scores of a language modeling head. These can be scores for each vocabulary token before SoftMax or scores for each vocabulary token after SoftMax.

---

generation/stopping_criteria.StoppingCriteriaList

Kind: static class of generation/stopping_criteria

• .StoppingCriteriaList
  • new StoppingCriteriaList()
  • .push(item)
  • .extend(items)

---

new StoppingCriteriaList()

Constructs a new instance of StoppingCriteriaList.

---

stoppingCriteriaList.push(item)

Adds a new stopping criterion to the list.

Kind: instance method of StoppingCriteriaList

Param	Type	Description
item	StoppingCriteria	The stopping criterion to add.

---

stoppingCriteriaList.extend(items)

Adds multiple stopping criteria to the list.

Kind: instance method of StoppingCriteriaList

Param	Type	Description
items	StoppingCriteria | StoppingCriteriaList | Array<StoppingCriteria>	The stopping criteria to add.

---

generation/stopping_criteria.MaxLengthCriteria

This class can be used to stop generation whenever the full generated number of tokens exceeds max_length. Keep in mind for decoder-only type of transformers, this will include the initial prompted tokens.

Kind: static class of generation/stopping_criteria

---

new MaxLengthCriteria(max_length, [max_position_embeddings])

Param	Type	Default	Description
max_length	number		The maximum length that the output sequence can have in number of tokens.
[max_position_embeddings]	number		The maximum model length, as defined by the model's config.max_position_embeddings attribute.

---

generation/stopping_criteria.EosTokenCriteria

This class can be used to stop generation whenever the “end-of-sequence” token is generated. By default, it uses the model.generation_config.eos_token_id.

Kind: static class of generation/stopping_criteria

• .EosTokenCriteria
  • new EosTokenCriteria(eos_token_id)
  • ._call(input_ids, scores) ⇒ Array.<boolean>

---

new EosTokenCriteria(eos_token_id)

Param	Type	Description
eos_token_id	number | Array<number>	The id of the end-of-sequence token. Optionally, use a list to set multiple end-of-sequence tokens.

---

eosTokenCriteria._call(input_ids, scores) ⇒ <code> Array. < boolean > </code>

Kind: instance method of EosTokenCriteria

Param	Type
input_ids	Array.<Array<number>>
scores	Array.<Array<number>>

---

generation/stopping_criteria.InterruptableStoppingCriteria

This class can be used to stop generation whenever the user interrupts the process.

Kind: static class of generation/stopping_criteria

---

< > Update on GitHub