From 91f6499393d2d999331fbfdba47a7f8b9f913f0d Mon Sep 17 00:00:00 2001 From: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com> Date: Thu, 16 Nov 2023 19:14:37 -0700 Subject: Respect tokenizer.ggml.add_bos_token value when tokenizing (#4040) * gguf-py: gguf-dump: Respect --no-tensor flag in JSON mode. * Respect add_bos_token GGUF metadata value * gguf-py: Try to fix SpecialVocab giving up too easily for the Nth time --- examples/llava/llava-cli.cpp | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'examples/llava') diff --git a/examples/llava/llava-cli.cpp b/examples/llava/llava-cli.cpp index 633afd1d..31f8cd8e 100644 --- a/examples/llava/llava-cli.cpp +++ b/examples/llava/llava-cli.cpp @@ -208,9 +208,10 @@ static void process_prompt(struct llava_context * ctx_llava, struct llava_image_ int n_past = 0; const int max_tgt_len = params->n_predict < 0 ? 256 : params->n_predict; + const bool add_bos = llama_should_add_bos_token(llama_get_model(ctx_llava->ctx_llama)); // llava chat format is "\nUSER:\n\nASSISTANT:" - eval_string(ctx_llava->ctx_llama, "A chat between a curious human and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the human's questions.\nUSER:", params->n_batch, &n_past, true); + eval_string(ctx_llava->ctx_llama, "A chat between a curious human and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the human's questions.\nUSER:", params->n_batch, &n_past, add_bos); llava_eval_image_embed(ctx_llava->ctx_llama, image_embed, params->n_batch, &n_past); eval_string(ctx_llava->ctx_llama, (prompt + "\nASSISTANT:").c_str(), params->n_batch, &n_past, false); -- cgit v1.2.3