diff options
Diffstat (limited to 'llama.h')
| -rw-r--r-- | llama.h | 12 |
1 files changed, 12 insertions, 0 deletions
@@ -309,6 +309,18 @@ extern "C" { /// @details Frequency and presence penalties described in OpenAI API https://platform.openai.com/docs/api-reference/parameter-details. LLAMA_API void llama_sample_frequency_and_presence_penalties(struct llama_context * ctx, llama_token_data_array * candidates, const llama_token * last_tokens, size_t last_tokens_size, float alpha_frequency, float alpha_presence); + /// @details Apply classifier-free guidance to the logits as described in academic paper "Stay on topic with Classifier-Free Guidance" https://arxiv.org/abs/2306.17806 + /// @param candidates A vector of `llama_token_data` containing the candidate tokens, the logits must be directly extracted from the original generation context without being sorted. + /// @params guidance_ctx A separate context from the same model. Other than a negative prompt at the beginning, it should have all generated and user input tokens copied from the main context. + /// @params scale Guidance strength. 1.0f means no guidance. Higher values mean stronger guidance. + /// @params smooth_factor Smooth factor between guidance logits and original logits. 1.0f means only use guidance logits. 0.0f means only original logits. + LLAMA_API void llama_sample_classifier_free_guidance( + struct llama_context * ctx, + llama_token_data_array * candidates, + struct llama_context * guidance_ctx, + float scale, + float smooth_factor); + /// @details Sorts candidate tokens by their logits in descending order and calculate probabilities based on logits. LLAMA_API void llama_sample_softmax(struct llama_context * ctx, llama_token_data_array * candidates); |