IBM Granite Architecture #9412
ggerganov
merged
9 commits into
ggml-org:master
from
gabe-l-hart:GraniteLM
ggerganov
approved these changes
on 2024-09-12
feat(gguf-py): Add Granite model and params to gguf-py
5ebc5ef5
feat(convert_hf_to_gguf): Add registration and param setup for Granite
406833d7
feat(llama.cpp): Add config parsing for Granite multiplier params
383065ad
feat(llama.cpp): First pass at full port of granite deviations from l…
ec13f29b
fix(llama.cpp): Determine granite language 3b instruct by vocab size
e73d795e
fix(convert_hf_to_gguf): Use LlamaModel as base for GraniteModel
80863806
fix(llama.cpp): Switch Granite param names to use _scale for consistency
0bdf04e7
fix(convert_hf_to_gguf/gguf-py): _multiplier -> _scale
65c5bb91
fix(llama.cpp): Use separate switch clause for granite in llm_load_hp…
5d054a42
Login to write a write a comment.
Login via GitHub