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

model: support arch DbrxForCausalLM (#6515)

Browse source 
* model: dbrx convert to gguf
#6344

* llama: support dbrx
#6344

* doc: dbrx: add the model as supported

* scripts: get-wikitext-2 add unzip

* llama: increase maximum experts allowed

* llama: factorize moe graph implementation between grok, mixtral and dbrx


---------

Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
This commit is contained in:
Pierrick Hymbert 2024-04-13 11:33:52 +02:00 • committed by GitHub
parent ab9a3240a9
commit 4bd0f93e4a
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
7 changed files with 428 additions and 148 deletions
Download patch file Download diff file Expand all files Collapse all files

15
gguf-py/gguf/constants.py
View file

@ -126,6 +126,7 @@ class MODEL_ARCH(IntEnum):
MAMBA = auto()
XVERSE = auto()
COMMAND_R = auto()
DBRX = auto()
class MODEL_TENSOR(IntEnum):
@ -195,6 +196,7 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
MODEL_ARCH.MAMBA: "mamba",
MODEL_ARCH.XVERSE: "xverse",
MODEL_ARCH.COMMAND_R: "command-r",
MODEL_ARCH.DBRX: "dbrx",
}
TENSOR_NAMES: dict[MODEL_TENSOR, str] = {
@ -642,6 +644,19 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
MODEL_TENSOR.ATTN_K_NORM,
MODEL_TENSOR.ATTN_Q_NORM,
],
MODEL_ARCH.DBRX: [
MODEL_TENSOR.TOKEN_EMBD,
MODEL_TENSOR.OUTPUT_NORM,
MODEL_TENSOR.OUTPUT,
MODEL_TENSOR.ATTN_NORM,
MODEL_TENSOR.ATTN_QKV,
MODEL_TENSOR.ATTN_OUT,
MODEL_TENSOR.ATTN_OUT_NORM,
MODEL_TENSOR.FFN_GATE_INP,
MODEL_TENSOR.FFN_GATE_EXP,
MODEL_TENSOR.FFN_DOWN_EXP,
MODEL_TENSOR.FFN_UP_EXP,
],
# TODO
}