diff options
| author | saood06 <saood05@gmail.com> | 2025-04-22 01:34:13 -0500 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2025-04-22 08:34:13 +0200 |
| commit | cc398007238cbbb064609cbdc9bc4aab03c658d7 (patch) | |
| tree | a414370696faa5ad8a89ac5aa7ec7cf2171a78a4 /gguf-py | |
| parent | 93cd77b65501246603061c7ee2801a992e3c6312 (diff) | |
Add support for bitnet2b_2501 model (#337)
* add support for bitnet2b_2501 model
* Fixes
* Support both model names
---------
Co-authored-by: potassiummmm <zhou.hansong@outlook.com>
Diffstat (limited to 'gguf-py')
| -rw-r--r-- | gguf-py/gguf/constants.py | 24 | ||||
| -rw-r--r-- | gguf-py/gguf/tensor_mapping.py | 5 |
2 files changed, 29 insertions, 0 deletions
diff --git a/gguf-py/gguf/constants.py b/gguf-py/gguf/constants.py index 93c614d6..22cde145 100644 --- a/gguf-py/gguf/constants.py +++ b/gguf-py/gguf/constants.py @@ -219,6 +219,7 @@ class MODEL_ARCH(IntEnum): DEEPSEEK2 = auto() CHATGLM = auto() BITNET = auto() + BITNET_25 = auto() T5 = auto() T5ENCODER = auto() JAIS = auto() @@ -351,6 +352,7 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = { MODEL_ARCH.DEEPSEEK2: "deepseek2", MODEL_ARCH.CHATGLM: "chatglm", MODEL_ARCH.BITNET: "bitnet", + MODEL_ARCH.BITNET_25: "bitnet-25", MODEL_ARCH.T5: "t5", MODEL_ARCH.T5ENCODER: "t5encoder", MODEL_ARCH.JAIS: "jais", @@ -1019,6 +1021,28 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = { MODEL_TENSOR.ATTN_SUB_NORM, MODEL_TENSOR.FFN_SUB_NORM, ], + MODEL_ARCH.BITNET_25: [ + MODEL_TENSOR.TOKEN_EMBD, + MODEL_TENSOR.OUTPUT_NORM, + MODEL_TENSOR.OUTPUT, + MODEL_TENSOR.ROPE_FREQS, + MODEL_TENSOR.ATTN_NORM, + MODEL_TENSOR.ATTN_Q, + MODEL_TENSOR.ATTN_K, + MODEL_TENSOR.ATTN_V, + MODEL_TENSOR.ATTN_OUT, + MODEL_TENSOR.ATTN_ROT_EMBD, + MODEL_TENSOR.FFN_GATE_INP, + MODEL_TENSOR.FFN_NORM, + MODEL_TENSOR.FFN_GATE, + MODEL_TENSOR.FFN_DOWN, + MODEL_TENSOR.FFN_UP, + MODEL_TENSOR.FFN_GATE_EXP, + MODEL_TENSOR.FFN_DOWN_EXP, + MODEL_TENSOR.FFN_UP_EXP, + MODEL_TENSOR.ATTN_SUB_NORM, + MODEL_TENSOR.FFN_SUB_NORM, + ], MODEL_ARCH.T5: [ MODEL_TENSOR.TOKEN_EMBD, MODEL_TENSOR.OUTPUT, diff --git a/gguf-py/gguf/tensor_mapping.py b/gguf-py/gguf/tensor_mapping.py index e8725426..1dea6a82 100644 --- a/gguf-py/gguf/tensor_mapping.py +++ b/gguf-py/gguf/tensor_mapping.py @@ -131,6 +131,7 @@ class TensorNameMap: "model.layers.{bid}.self_attn.qkv_proj", # phi3 "encoder.layers.{bid}.self_attention.query_key_value", # chatglm "transformer.layers.{bid}.attn.qkv_proj", # openelm + "layers.{bid}.attention.wqkv", ), # Attention query @@ -464,10 +465,14 @@ class TensorNameMap: MODEL_TENSOR.ATTN_SUB_NORM: ( "model.layers.{bid}.self_attn.inner_attn_ln", # bitnet + "layers.{bid}.attention.attn_sub_norm", # bitnet + "model.layers.{bid}.self_attn.attn_sub_norm", ), MODEL_TENSOR.FFN_SUB_NORM: ( "model.layers.{bid}.mlp.ffn_layernorm", # bitnet + "layers.{bid}.feed_forward.ffn_sub_norm", # bitnet + "model.layers.{bid}.mlp.ffn_sub_norm", ), MODEL_TENSOR.DEC_ATTN_NORM: ( |