vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Respect tokenizer.ggml.add_bos_token value when tokenizing (#4040)

Browse source 
* 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
This commit is contained in:
Kerfuffle 2023-11-16 19:14:37 -07:00 • committed by GitHub
parent 8da46278e1
commit 91f6499393
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
12 changed files with 85 additions and 29 deletions
Download patch file Download diff file Expand all files Collapse all files

3
examples/llava/llava-cli.cpp
View file

@ -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 "<system_prompt>\nUSER:<image_embeddings>\n<textual_prompt>\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);