rlightning.models.model_utils¶ rlightning.models.model_utils.compute_entropy_from_logits(logits, epsilon=1e-10)[source]¶ Compute entropy by logits. Parameters: logits – [B, vocab-size, seq-len] epsilon – Small constant for numerical stability. Returns: [B, seq-len] Return type: entropy rlightning.models.model_utils.compute_logprobs_from_logits(logits, target)[source]¶