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

llama : fix pre-tokenizers

Browse source
This commit is contained in:
Georgi Gerganov 2024-05-08 19:43:26 +03:00
parent e59b54657a
commit b7ede48294
No known key found for this signature in database
GPG key ID: 449E073F9DC10735
3 changed files with 56 additions and 42 deletions
Download patch file Download diff file Expand all files Collapse all files

9
convert-hf-to-gguf-update.py
View file

@ -55,8 +55,8 @@ else:
# TODO: add models here, base models preferred
models = [
#{"name": "llama-spm", "tokt": TOKENIZER_TYPE.SPM, "repo": "https://huggingface.co/meta-llama/Llama-2-7b-hf", },
#{"name": "llama-bpe", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/meta-llama/Meta-Llama-3-8B", },
{"name": "llama-spm", "tokt": TOKENIZER_TYPE.SPM, "repo": "https://huggingface.co/meta-llama/Llama-2-7b-hf", },
{"name": "llama-bpe", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/meta-llama/Meta-Llama-3-8B", },
{"name": "phi-3", "tokt": TOKENIZER_TYPE.SPM, "repo": "https://huggingface.co/microsoft/Phi-3-mini-4k-instruct", },
{"name": "deepseek-llm", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/deepseek-ai/deepseek-llm-7b-base", },
{"name": "deepseek-coder", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/deepseek-ai/deepseek-coder-6.7b-base", },
@ -70,8 +70,9 @@ models = [
{"name": "qwen2", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/Qwen/Qwen1.5-7B", },
{"name": "olmo", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/allenai/OLMo-1.7-7B-hf", },
{"name": "dbrx", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/databricks/dbrx-base", },
{"name": "jina-embeddings-v2-base-es", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/jinaai/jina-embeddings-v2-base-es", },
{"name": "jina-embeddings-v2-base-de", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/jinaai/jina-embeddings-v2-base-de", },
{"name": "jina-en", "tokt": TOKENIZER_TYPE.WPM, "repo": "https://huggingface.co/jinaai/jina-embeddings-v2-base-en", }, # WPM!
{"name": "jina-es", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/jinaai/jina-embeddings-v2-base-es", },
{"name": "jina-de", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/jinaai/jina-embeddings-v2-base-de", },
]
# make directory "models/tokenizers" if it doesn't exist

11
convert-hf-to-gguf.py
View file

@ -322,12 +322,17 @@ class Model(ABC):
# ref: https://huggingface.co/allenai/OLMo-1.7-7B-hf
res = "olmo"
if chkhsh == "a8594e3edff7c29c003940395316294b2c623e09894deebbc65f33f1515df79e":
# ref: https://huggingface.co/databricks/dbrx-instruct
# ref: https://huggingface.co/databricks/dbrx-base
res = "dbrx"
if chkhsh == "0876d13b50744004aa9aeae05e7b0647eac9d801b5ba4668afc01e709c15e19f":
# ref: https://huggingface.co/jinaai/jina-embeddings-v2-base-en
res = "jina-en"
if chkhsh == "171aeeedd6fb548d418a7461d053f11b6f1f1fc9b387bd66640d28a4b9f5c643":
res = "default"
# ref: https://huggingface.co/jinaai/jina-embeddings-v2-base-es
res = "jina-es"
if chkhsh == "27949a2493fc4a9f53f5b9b029c82689cfbe5d3a1929bb25e043089e28466de6":
res = "default"
# ref: https://huggingface.co/jinaai/jina-embeddings-v2-base-de
res = "jina-de"
if res is None:
logger.warning("\n")

10
llama.cpp
View file

@ -3800,6 +3800,12 @@ static void llm_load_hparams(
// get hparams kv
ml.get_key(LLM_KV_VOCAB_SIZE, hparams.n_vocab, false) || ml.get_arr_n(LLM_KV_TOKENIZER_LIST, hparams.n_vocab);
// everything past this point is not vocab-related
if (hparams.vocab_only) {
return;
}
ml.get_key(LLM_KV_CONTEXT_LENGTH, hparams.n_ctx_train);
ml.get_key(LLM_KV_EMBEDDING_LENGTH, hparams.n_embd);
ml.get_key(LLM_KV_FEED_FORWARD_LENGTH, hparams.n_ff);
@ -4417,7 +4423,9 @@ static void llm_load_vocab(
tokenizer_pre == "starcoder") {
vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_STARCODER;
} else if (
tokenizer_pre == "gpt-2") {
tokenizer_pre == "gpt-2" ||
tokenizer_pre == "jina-es" ||
tokenizer_pre == "jina-de") {
vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_GPT2;
} else if (
tokenizer_pre == "refact") {