tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-06-27 20:05:20 +00:00
Code Issues Packages Projects Releases Wiki Activity

llama : add Trillion 7B model support (#12556)

Browse Source 
* Support Trillion 7B

* Update llama.h

* Update llama.h

* Update llama-vocab.cpp for Trillion

* Update llama-vocab.cpp
This commit is contained in:
Juyoung Suk
2025-03-31 03:38:33 +09:00
committed by GitHub
parent 7242dd9675
commit b3de7cac73
5 changed files with 11 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
README.md
View File

@ -112,6 +112,7 @@ Instructions for adding support for new models: [HOWTO-add-model.md](docs/develo
- [x] [RWKV-6](https://github.com/BlinkDL/RWKV-LM) - [x] [RWKV-6](https://github.com/BlinkDL/RWKV-LM)
- [x] [QRWKV-6](https://huggingface.co/recursal/QRWKV6-32B-Instruct-Preview-v0.1) - [x] [QRWKV-6](https://huggingface.co/recursal/QRWKV6-32B-Instruct-Preview-v0.1)
- [x] [GigaChat-20B-A3B](https://huggingface.co/ai-sage/GigaChat-20B-A3B-instruct) - [x] [GigaChat-20B-A3B](https://huggingface.co/ai-sage/GigaChat-20B-A3B-instruct)
- [X] [Trillion-7B-preview](https://huggingface.co/trillionlabs/Trillion-7B-preview)
#### Multimodal #### Multimodal

3
convert_hf_to_gguf.py
View File

@ -708,6 +708,9 @@ class Model:
if chkhsh == "7dec86086fcc38b66b7bc1575a160ae21cf705be7718b9d5598190d7c12db76f": if chkhsh == "7dec86086fcc38b66b7bc1575a160ae21cf705be7718b9d5598190d7c12db76f":
# ref: https://huggingface.co/UW/OLMo2-8B-SuperBPE-t180k # ref: https://huggingface.co/UW/OLMo2-8B-SuperBPE-t180k
res = "superbpe" res = "superbpe"
if chkhsh == "1994ffd01900cfb37395608534236ecd63f2bd5995d6cb1004dda1af50240f15":
# ref: https://huggingface.co/trillionlabs/Trillion-7B-preview
res = "trillion"
if res is None: if res is None:
logger.warning("\n") logger.warning("\n")

1
convert_hf_to_gguf_update.py
View File

@ -111,6 +111,7 @@ models = [
{"name": "deepseek-r1-qwen", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B"}, {"name": "deepseek-r1-qwen", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B"},
{"name": "gpt-4o", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/Xenova/gpt-4o", }, {"name": "gpt-4o", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/Xenova/gpt-4o", },
{"name": "superbpe", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/UW/OLMo2-8B-SuperBPE-t180k", }, {"name": "superbpe", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/UW/OLMo2-8B-SuperBPE-t180k", },
{"name": "trillion", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/trillionlabs/Trillion-7B-preview", },
] ]

1
include/llama.h
View File

@ -108,6 +108,7 @@ extern "C" {
LLAMA_VOCAB_PRE_TYPE_DEEPSEEK3_LLM = 28, LLAMA_VOCAB_PRE_TYPE_DEEPSEEK3_LLM = 28,
LLAMA_VOCAB_PRE_TYPE_GPT4O = 29, LLAMA_VOCAB_PRE_TYPE_GPT4O = 29,
LLAMA_VOCAB_PRE_TYPE_SUPERBPE = 30, LLAMA_VOCAB_PRE_TYPE_SUPERBPE = 30,
LLAMA_VOCAB_PRE_TYPE_TRILLION = 31,
}; };
enum llama_rope_type { enum llama_rope_type {

5
src/llama-vocab.cpp
View File

@ -342,6 +342,7 @@ struct llm_tokenizer_bpe : llm_tokenizer {
case LLAMA_VOCAB_PRE_TYPE_MPT: case LLAMA_VOCAB_PRE_TYPE_MPT:
case LLAMA_VOCAB_PRE_TYPE_OLMO: case LLAMA_VOCAB_PRE_TYPE_OLMO:
case LLAMA_VOCAB_PRE_TYPE_JAIS: case LLAMA_VOCAB_PRE_TYPE_JAIS:
case LLAMA_VOCAB_PRE_TYPE_TRILLION:
regex_exprs = { regex_exprs = {
"'s|'t|'re|'ve|'m|'ll|'d| ?\\p{L}+| ?\\p{N}+| ?[^\\s\\p{L}\\p{N}]+|\\s+(?!\\S)", "'s|'t|'re|'ve|'m|'ll|'d| ?\\p{L}+| ?\\p{N}+| ?[^\\s\\p{L}\\p{N}]+|\\s+(?!\\S)",
}; };
@ -1614,6 +1615,10 @@ void llama_vocab::impl::load(llama_model_loader & ml, const LLM_KV & kv) {
tokenizer_pre == "superbpe") { tokenizer_pre == "superbpe") {
pre_type = LLAMA_VOCAB_PRE_TYPE_SUPERBPE; pre_type = LLAMA_VOCAB_PRE_TYPE_SUPERBPE;
clean_spaces = false; clean_spaces = false;
} else if (
tokenizer_pre == "trillion") {
pre_type = LLAMA_VOCAB_PRE_TYPE_TRILLION;
clean_spaces = false;
} else { } else {
throw std::runtime_error(format("unknown pre-tokenizer type: '%s'", tokenizer_pre.c_str())); throw std::runtime_error(format("unknown pre-tokenizer type: '%s'", tokenizer_pre.c_str()));
} }