From 4bd0f93e4ab4fe6682e7d0241c1bdec1397e954a Mon Sep 17 00:00:00 2001 From: Pierrick Hymbert Date: Sat, 13 Apr 2024 11:33:52 +0200 Subject: model: support arch `DbrxForCausalLM` (#6515) * 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> --- gguf-py/gguf/constants.py | 15 +++++++++++++++ 1 file changed, 15 insertions(+) (limited to 'gguf-py/gguf/constants.py') diff --git a/gguf-py/gguf/constants.py b/gguf-py/gguf/constants.py index a6454a10..2566b2fb 100644 --- a/gguf-py/gguf/constants.py +++ b/gguf-py/gguf/constants.py @@ -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 } -- cgit v1.2.3