mirror of https://github.com/leafspark/AutoGGUF
refactor: adapt gguf library to project
- remove comments - remove argparse help text
This commit is contained in:
BuildTools
parent
f7f9a457ea
commit
a7e8bf673e
|
|
@ -4403,83 +4403,81 @@ def __torch_function__(cls, func, types, args=(), kwargs=None):
|
|||
|
||||
|
||||
def parse_args() -> argparse.Namespace:
|
||||
parser = argparse.ArgumentParser(
|
||||
description="Convert a huggingface model to a GGML compatible file"
|
||||
)
|
||||
parser = argparse.ArgumentParser(description="")
|
||||
parser.add_argument(
|
||||
"--vocab-only",
|
||||
action="store_true",
|
||||
help="extract only the vocab",
|
||||
help="",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--outfile",
|
||||
type=Path,
|
||||
help="path to write to; default: based on input. {ftype} will be replaced by the outtype.",
|
||||
help="",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--outtype",
|
||||
type=str,
|
||||
choices=["f32", "f16", "bf16", "q8_0", "auto"],
|
||||
default="f16",
|
||||
help="output format - use f32 for float32, f16 for float16, bf16 for bfloat16, q8_0 for Q8_0, auto for the highest-fidelity 16-bit float type depending on the first loaded tensor type",
|
||||
help="",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--bigendian",
|
||||
action="store_true",
|
||||
help="model is executed on big endian machine",
|
||||
help="",
|
||||
)
|
||||
parser.add_argument(
|
||||
"model",
|
||||
type=Path,
|
||||
help="directory containing model file",
|
||||
help="",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--use-temp-file",
|
||||
action="store_true",
|
||||
help="use the tempfile library while processing (helpful when running out of memory, process killed)",
|
||||
help="",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--no-lazy",
|
||||
action="store_true",
|
||||
help="use more RAM by computing all outputs before writing (use in case lazy evaluation is broken)",
|
||||
help="",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--model-name",
|
||||
type=str,
|
||||
default=None,
|
||||
help="name of the model",
|
||||
help="",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--verbose",
|
||||
action="store_true",
|
||||
help="increase output verbosity",
|
||||
help="",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--split-max-tensors",
|
||||
type=int,
|
||||
default=0,
|
||||
help="max tensors in each split",
|
||||
help="",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--split-max-size",
|
||||
type=str,
|
||||
default="0",
|
||||
help="max size per split N(M|G)",
|
||||
help="",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--dry-run",
|
||||
action="store_true",
|
||||
help="only print out a split plan and exit, without writing any new files",
|
||||
help="",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--no-tensor-first-split",
|
||||
action="store_true",
|
||||
help="do not add tensors to the first split (disabled by default)",
|
||||
help="",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--metadata",
|
||||
type=Path,
|
||||
help="Specify the path for an authorship metadata override file",
|
||||
help="",
|
||||
)
|
||||
|
||||
return parser.parse_args()
|
||||
|
|
|
|||
File diff suppressed because it is too large
Load Diff
|
|
@ -4,471 +4,569 @@
|
|||
|
||||
from .constants import MODEL_ARCH, MODEL_TENSOR, MODEL_TENSORS, TENSOR_NAMES
|
||||
|
||||
|
||||
class TensorNameMap:
|
||||
mappings_cfg: dict[MODEL_TENSOR, tuple[str, ...]] = {
|
||||
# Token embeddings
|
||||
|
||||
MODEL_TENSOR.TOKEN_EMBD: (
|
||||
"gpt_neox.embed_in", # gptneox
|
||||
"transformer.wte", # gpt2 gpt-j mpt refact qwen dbrx jais
|
||||
"transformer.word_embeddings", # falcon
|
||||
"word_embeddings", # bloom
|
||||
"model.embed_tokens", # llama-hf
|
||||
"tok_embeddings", # llama-pth
|
||||
"embeddings.word_embeddings", # bert nomic-bert
|
||||
"language_model.embedding.word_embeddings", # persimmon
|
||||
"wte", # gpt2
|
||||
"transformer.embd.wte", # phi2
|
||||
"model.tok_embeddings", # internlm2
|
||||
"model.embedding", # mamba-qbert
|
||||
"backbone.embedding", # mamba
|
||||
"backbone.embeddings", # mamba-hf
|
||||
"transformer.in_out_embed", # Grok
|
||||
"embedding.word_embeddings", # chatglm
|
||||
"transformer.token_embeddings", # openelm
|
||||
"shared", # t5
|
||||
"gpt_neox.embed_in",
|
||||
"transformer.wte",
|
||||
"transformer.word_embeddings",
|
||||
"word_embeddings",
|
||||
"model.embed_tokens",
|
||||
"tok_embeddings",
|
||||
"embeddings.word_embeddings",
|
||||
"language_model.embedding.word_embeddings",
|
||||
"wte",
|
||||
"transformer.embd.wte",
|
||||
"model.tok_embeddings",
|
||||
"model.embedding",
|
||||
"backbone.embedding",
|
||||
"backbone.embeddings",
|
||||
"transformer.in_out_embed",
|
||||
"embedding.word_embeddings",
|
||||
"transformer.token_embeddings",
|
||||
"shared",
|
||||
),
|
||||
# Token type embeddings
|
||||
|
||||
MODEL_TENSOR.TOKEN_TYPES: (
|
||||
"embeddings.token_type_embeddings", # bert nomic-bert
|
||||
"embeddings.token_type_embeddings",
|
||||
),
|
||||
# Normalization of token embeddings
|
||||
|
||||
MODEL_TENSOR.TOKEN_EMBD_NORM: (
|
||||
"word_embeddings_layernorm", # bloom
|
||||
"embeddings.LayerNorm", # bert
|
||||
"emb_ln", # nomic-bert
|
||||
"transformer.norm", # openelm
|
||||
"word_embeddings_layernorm",
|
||||
"embeddings.LayerNorm",
|
||||
"emb_ln",
|
||||
"transformer.norm",
|
||||
),
|
||||
# Position embeddings
|
||||
|
||||
MODEL_TENSOR.POS_EMBD: (
|
||||
"transformer.wpe", # gpt2
|
||||
"embeddings.position_embeddings", # bert
|
||||
"wpe", # gpt2
|
||||
"transformer.wpe",
|
||||
"embeddings.position_embeddings",
|
||||
"wpe",
|
||||
),
|
||||
# Output
|
||||
|
||||
MODEL_TENSOR.OUTPUT: (
|
||||
"embed_out", # gptneox
|
||||
"lm_head", # gpt2 mpt falcon llama-hf baichuan qwen mamba dbrx jais
|
||||
"output", # llama-pth bloom internlm2
|
||||
"word_embeddings_for_head", # persimmon
|
||||
"lm_head.linear", # phi2
|
||||
"output_layer", # chatglm
|
||||
"embed_out",
|
||||
"lm_head",
|
||||
"output",
|
||||
"word_embeddings_for_head",
|
||||
"lm_head.linear",
|
||||
"output_layer",
|
||||
),
|
||||
# Output norm
|
||||
|
||||
MODEL_TENSOR.OUTPUT_NORM: (
|
||||
"gpt_neox.final_layer_norm", # gptneox
|
||||
"transformer.ln_f", # gpt2 gpt-j falcon jais
|
||||
"model.norm", # llama-hf baichuan internlm2
|
||||
"norm", # llama-pth
|
||||
"transformer.norm_f", # mpt dbrx
|
||||
"ln_f", # refact bloom qwen gpt2
|
||||
"language_model.encoder.final_layernorm", # persimmon
|
||||
"model.final_layernorm", # persimmon
|
||||
"lm_head.ln", # phi2
|
||||
"model.norm_f", # mamba-qbert
|
||||
"backbone.norm_f", # mamba
|
||||
"transformer.rms_norm", # Grok
|
||||
"encoder.final_layernorm", # chatglm
|
||||
"transformer.norm", # openelm
|
||||
"gpt_neox.final_layer_norm",
|
||||
"transformer.ln_f",
|
||||
"model.norm",
|
||||
"norm",
|
||||
"transformer.norm_f",
|
||||
"ln_f",
|
||||
"language_model.encoder.final_layernorm",
|
||||
"model.final_layernorm",
|
||||
"lm_head.ln",
|
||||
"model.norm_f",
|
||||
"backbone.norm_f",
|
||||
"transformer.rms_norm",
|
||||
"encoder.final_layernorm",
|
||||
"transformer.norm",
|
||||
"model.norm",
|
||||
),
|
||||
# Rope frequencies
|
||||
|
||||
MODEL_TENSOR.ROPE_FREQS: (
|
||||
"rope.freqs", # llama-pth
|
||||
"rotary_pos_emb.inv_freq", # chatglm
|
||||
"rope.freqs",
|
||||
"rotary_pos_emb.inv_freq",
|
||||
),
|
||||
}
|
||||
|
||||
block_mappings_cfg: dict[MODEL_TENSOR, tuple[str, ...]] = {
|
||||
# Attention norm
|
||||
|
||||
MODEL_TENSOR.ATTN_NORM: (
|
||||
"gpt_neox.layers.{bid}.input_layernorm", # gptneox
|
||||
"transformer.h.{bid}.ln_1", # gpt2 gpt-j refact qwen jais
|
||||
"transformer.blocks.{bid}.norm_1", # mpt
|
||||
"transformer.h.{bid}.input_layernorm", # falcon7b
|
||||
"h.{bid}.input_layernorm", # bloom
|
||||
"transformer.h.{bid}.ln_mlp", # falcon40b
|
||||
"model.layers.{bid}.input_layernorm", # llama-hf
|
||||
"layers.{bid}.attention_norm", # llama-pth
|
||||
"language_model.encoder.layers.{bid}.input_layernorm", # persimmon
|
||||
"model.layers.{bid}.ln1", # yi
|
||||
"h.{bid}.ln_1", # gpt2
|
||||
"transformer.h.{bid}.ln", # phi2
|
||||
"model.layers.layers.{bid}.norm", # plamo
|
||||
"model.layers.{bid}.attention_norm", # internlm2
|
||||
"model.layers.{bid}.norm", # mamba-qbert
|
||||
"backbone.layers.{bid}.norm", # mamba
|
||||
"transformer.decoder_layer.{bid}.rms_norm", # Grok
|
||||
"transformer.blocks.{bid}.norm_attn_norm.norm_1", # dbrx
|
||||
"encoder.layers.{bid}.input_layernorm", # chatglm
|
||||
"transformer.layers.{bid}.attn_norm", # openelm
|
||||
"gpt_neox.layers.{bid}.input_layernorm",
|
||||
"transformer.h.{bid}.ln_1",
|
||||
"transformer.blocks.{bid}.norm_1",
|
||||
"transformer.h.{bid}.input_layernorm",
|
||||
"h.{bid}.input_layernorm",
|
||||
"transformer.h.{bid}.ln_mlp",
|
||||
"model.layers.{bid}.input_layernorm",
|
||||
"layers.{bid}.attention_norm",
|
||||
"language_model.encoder.layers.{bid}.input_layernorm",
|
||||
"model.layers.{bid}.ln1",
|
||||
"h.{bid}.ln_1",
|
||||
"transformer.h.{bid}.ln",
|
||||
"model.layers.layers.{bid}.norm",
|
||||
"model.layers.{bid}.attention_norm",
|
||||
"model.layers.{bid}.norm",
|
||||
"backbone.layers.{bid}.norm",
|
||||
"transformer.decoder_layer.{bid}.rms_norm",
|
||||
"transformer.blocks.{bid}.norm_attn_norm.norm_1",
|
||||
"encoder.layers.{bid}.input_layernorm",
|
||||
"transformer.layers.{bid}.attn_norm",
|
||||
),
|
||||
# Attention norm 2
|
||||
|
||||
MODEL_TENSOR.ATTN_NORM_2: (
|
||||
"transformer.h.{bid}.ln_attn", # falcon40b
|
||||
"encoder.layer.{bid}.layer_norm_1", # jina-v2-code
|
||||
"transformer.h.{bid}.ln_attn",
|
||||
"encoder.layer.{bid}.layer_norm_1",
|
||||
),
|
||||
# Attention query-key-value
|
||||
|
||||
MODEL_TENSOR.ATTN_QKV: (
|
||||
"gpt_neox.layers.{bid}.attention.query_key_value", # gptneox
|
||||
"transformer.h.{bid}.attn.c_attn", # gpt2 qwen jais
|
||||
"transformer.blocks.{bid}.attn.Wqkv", # mpt
|
||||
"transformer.blocks.{bid}.norm_attn_norm.attn.Wqkv", # dbrx
|
||||
"transformer.h.{bid}.self_attention.query_key_value", # falcon
|
||||
"h.{bid}.self_attention.query_key_value", # bloom
|
||||
"language_model.encoder.layers.{bid}.self_attention.query_key_value", # persimmon
|
||||
"model.layers.{bid}.self_attn.query_key_value", # persimmon
|
||||
"h.{bid}.attn.c_attn", # gpt2
|
||||
"transformer.h.{bid}.mixer.Wqkv", # phi2
|
||||
"encoder.layers.{bid}.attn.Wqkv", # nomic-bert
|
||||
"model.layers.{bid}.self_attn.qkv_proj", # phi3
|
||||
"encoder.layers.{bid}.self_attention.query_key_value", # chatglm
|
||||
"transformer.layers.{bid}.attn.qkv_proj", # openelm
|
||||
"gpt_neox.layers.{bid}.attention.query_key_value",
|
||||
"transformer.h.{bid}.attn.c_attn",
|
||||
"transformer.blocks.{bid}.attn.Wqkv",
|
||||
"transformer.blocks.{bid}.norm_attn_norm.attn.Wqkv",
|
||||
"transformer.h.{bid}.self_attention.query_key_value",
|
||||
"h.{bid}.self_attention.query_key_value",
|
||||
"language_model.encoder.layers.{bid}.self_attention.query_key_value",
|
||||
"model.layers.{bid}.self_attn.query_key_value",
|
||||
"h.{bid}.attn.c_attn",
|
||||
"transformer.h.{bid}.mixer.Wqkv",
|
||||
"encoder.layers.{bid}.attn.Wqkv",
|
||||
"model.layers.{bid}.self_attn.qkv_proj",
|
||||
"encoder.layers.{bid}.self_attention.query_key_value",
|
||||
"transformer.layers.{bid}.attn.qkv_proj",
|
||||
),
|
||||
# Attention query
|
||||
|
||||
MODEL_TENSOR.ATTN_Q: (
|
||||
"model.layers.{bid}.self_attn.q_proj", # llama-hf
|
||||
"layers.{bid}.attention.wq", # llama-pth
|
||||
"encoder.layer.{bid}.attention.self.query", # bert
|
||||
"transformer.h.{bid}.attn.q_proj", # gpt-j
|
||||
"model.layers.layers.{bid}.self_attn.q_proj", # plamo
|
||||
"model.layers.{bid}.attention.wq", # internlm2
|
||||
"transformer.decoder_layer.{bid}.multi_head_attention.query", # Grok
|
||||
"model.layers.{bid}.self_attn.q_proj",
|
||||
"layers.{bid}.attention.wq",
|
||||
"encoder.layer.{bid}.attention.self.query",
|
||||
"transformer.h.{bid}.attn.q_proj",
|
||||
"model.layers.layers.{bid}.self_attn.q_proj",
|
||||
"model.layers.{bid}.attention.wq",
|
||||
"transformer.decoder_layer.{bid}.multi_head_attention.query",
|
||||
"transformer.h.{bid}.attn.attention.q_proj",
|
||||
),
|
||||
# Attention key
|
||||
|
||||
MODEL_TENSOR.ATTN_K: (
|
||||
"model.layers.{bid}.self_attn.k_proj", # llama-hf
|
||||
"layers.{bid}.attention.wk", # llama-pth
|
||||
"encoder.layer.{bid}.attention.self.key", # bert
|
||||
"transformer.h.{bid}.attn.k_proj", # gpt-j
|
||||
"transformer.h.{bid}.attn.k", # refact
|
||||
"model.layers.layers.{bid}.self_attn.k_proj", # plamo
|
||||
"model.layers.{bid}.attention.wk", # internlm2
|
||||
"transformer.decoder_layer.{bid}.multi_head_attention.key", # Grok
|
||||
"model.layers.{bid}.self_attn.k_proj",
|
||||
"layers.{bid}.attention.wk",
|
||||
"encoder.layer.{bid}.attention.self.key",
|
||||
"transformer.h.{bid}.attn.k_proj",
|
||||
"transformer.h.{bid}.attn.k",
|
||||
"model.layers.layers.{bid}.self_attn.k_proj",
|
||||
"model.layers.{bid}.attention.wk",
|
||||
"transformer.decoder_layer.{bid}.multi_head_attention.key",
|
||||
"transformer.h.{bid}.attn.attention.k_proj",
|
||||
),
|
||||
# Attention value
|
||||
|
||||
MODEL_TENSOR.ATTN_V: (
|
||||
"model.layers.{bid}.self_attn.v_proj", # llama-hf
|
||||
"layers.{bid}.attention.wv", # llama-pth
|
||||
"encoder.layer.{bid}.attention.self.value", # bert
|
||||
"transformer.h.{bid}.attn.v_proj", # gpt-j
|
||||
"transformer.h.{bid}.attn.v", # refact
|
||||
"model.layers.layers.{bid}.self_attn.v_proj", # plamo
|
||||
"model.layers.{bid}.attention.wv", # internlm2
|
||||
"transformer.decoder_layer.{bid}.multi_head_attention.value", # Grok
|
||||
"model.layers.{bid}.self_attn.v_proj",
|
||||
"layers.{bid}.attention.wv",
|
||||
"encoder.layer.{bid}.attention.self.value",
|
||||
"transformer.h.{bid}.attn.v_proj",
|
||||
"transformer.h.{bid}.attn.v",
|
||||
"model.layers.layers.{bid}.self_attn.v_proj",
|
||||
"model.layers.{bid}.attention.wv",
|
||||
"transformer.decoder_layer.{bid}.multi_head_attention.value",
|
||||
"transformer.h.{bid}.attn.attention.v_proj",
|
||||
),
|
||||
# Attention output
|
||||
|
||||
MODEL_TENSOR.ATTN_OUT: (
|
||||
"gpt_neox.layers.{bid}.attention.dense", # gptneox
|
||||
"transformer.h.{bid}.attn.c_proj", # gpt2 refact qwen jais
|
||||
"transformer.blocks.{bid}.attn.out_proj", # mpt
|
||||
"transformer.h.{bid}.self_attention.dense", # falcon
|
||||
"h.{bid}.self_attention.dense", # bloom
|
||||
"model.layers.{bid}.self_attn.o_proj", # llama-hf
|
||||
"layers.{bid}.attention.wo", # llama-pth
|
||||
"encoder.layer.{bid}.attention.output.dense", # bert
|
||||
"transformer.h.{bid}.attn.out_proj", # gpt-j
|
||||
"language_model.encoder.layers.{bid}.self_attention.dense", # persimmon
|
||||
"model.layers.{bid}.self_attn.dense", # persimmon
|
||||
"h.{bid}.attn.c_proj", # gpt2
|
||||
"transformer.h.{bid}.mixer.out_proj", # phi2
|
||||
"model.layers.layers.{bid}.self_attn.o_proj", # plamo
|
||||
"model.layers.{bid}.attention.wo", # internlm2
|
||||
"encoder.layers.{bid}.attn.out_proj", # nomic-bert
|
||||
"transformer.decoder_layer.{bid}.multi_head_attention.linear", # Grok
|
||||
"transformer.blocks.{bid}.norm_attn_norm.attn.out_proj", # dbrx
|
||||
"encoder.layers.{bid}.self_attention.dense", # chatglm
|
||||
"transformer.layers.{bid}.attn.out_proj", # openelm
|
||||
"gpt_neox.layers.{bid}.attention.dense",
|
||||
"transformer.h.{bid}.attn.c_proj",
|
||||
"transformer.blocks.{bid}.attn.out_proj",
|
||||
"transformer.h.{bid}.self_attention.dense",
|
||||
"h.{bid}.self_attention.dense",
|
||||
"model.layers.{bid}.self_attn.o_proj",
|
||||
"layers.{bid}.attention.wo",
|
||||
"encoder.layer.{bid}.attention.output.dense",
|
||||
"transformer.h.{bid}.attn.out_proj",
|
||||
"language_model.encoder.layers.{bid}.self_attention.dense",
|
||||
"model.layers.{bid}.self_attn.dense",
|
||||
"h.{bid}.attn.c_proj",
|
||||
"transformer.h.{bid}.mixer.out_proj",
|
||||
"model.layers.layers.{bid}.self_attn.o_proj",
|
||||
"model.layers.{bid}.attention.wo",
|
||||
"encoder.layers.{bid}.attn.out_proj",
|
||||
"transformer.decoder_layer.{bid}.multi_head_attention.linear",
|
||||
"transformer.blocks.{bid}.norm_attn_norm.attn.out_proj",
|
||||
"encoder.layers.{bid}.self_attention.dense",
|
||||
"transformer.layers.{bid}.attn.out_proj",
|
||||
"transformer.h.{bid}.attn.attention.out_proj",
|
||||
),
|
||||
# Attention output norm
|
||||
|
||||
MODEL_TENSOR.ATTN_OUT_NORM: (
|
||||
"encoder.layer.{bid}.attention.output.LayerNorm", # bert
|
||||
"encoder.layers.{bid}.norm1", # nomic-bert
|
||||
"transformer.decoder_layer.{bid}.rms_norm_1", # Grok
|
||||
"transformer.blocks.{bid}.norm_attn_norm.norm_2", # dbrx
|
||||
"encoder.layer.{bid}.attention.output.LayerNorm",
|
||||
"encoder.layers.{bid}.norm1",
|
||||
"transformer.decoder_layer.{bid}.rms_norm_1",
|
||||
"transformer.blocks.{bid}.norm_attn_norm.norm_2",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.ATTN_POST_NORM: (
|
||||
"model.layers.{bid}.post_attention_layernorm", # gemma2
|
||||
"model.layers.{bid}.post_attention_layernorm",
|
||||
),
|
||||
# Rotary embeddings
|
||||
|
||||
MODEL_TENSOR.ATTN_ROT_EMBD: (
|
||||
"model.layers.{bid}.self_attn.rotary_emb.inv_freq", # llama-hf
|
||||
"layers.{bid}.attention.inner_attention.rope.freqs", # llama-pth
|
||||
"model.layers.layers.{bid}.self_attn.rotary_emb.inv_freq", # plamo
|
||||
"transformer.h.{bid}.attn.rotary_emb.inv_freq", # codeshell
|
||||
"model.layers.{bid}.self_attn.rotary_emb.inv_freq",
|
||||
"layers.{bid}.attention.inner_attention.rope.freqs",
|
||||
"model.layers.layers.{bid}.self_attn.rotary_emb.inv_freq",
|
||||
"transformer.h.{bid}.attn.rotary_emb.inv_freq",
|
||||
),
|
||||
# Feed-forward norm
|
||||
|
||||
MODEL_TENSOR.FFN_NORM: (
|
||||
"gpt_neox.layers.{bid}.post_attention_layernorm", # gptneox
|
||||
"transformer.h.{bid}.ln_2", # gpt2 refact qwen jais
|
||||
"h.{bid}.post_attention_layernorm", # bloom
|
||||
"transformer.blocks.{bid}.norm_2", # mpt
|
||||
"model.layers.{bid}.post_attention_layernorm", # llama-hf
|
||||
"layers.{bid}.ffn_norm", # llama-pth
|
||||
"language_model.encoder.layers.{bid}.post_attention_layernorm", # persimmon
|
||||
"model.layers.{bid}.ln2", # yi
|
||||
"h.{bid}.ln_2", # gpt2
|
||||
"model.layers.{bid}.ffn_norm", # internlm2
|
||||
"transformer.decoder_layer.{bid}.rms_norm_2", # Grok
|
||||
"encoder.layers.{bid}.post_attention_layernorm", # chatglm
|
||||
"transformer.layers.{bid}.ffn_norm", # openelm
|
||||
"gpt_neox.layers.{bid}.post_attention_layernorm",
|
||||
"transformer.h.{bid}.ln_2",
|
||||
"h.{bid}.post_attention_layernorm",
|
||||
"transformer.blocks.{bid}.norm_2",
|
||||
"model.layers.{bid}.post_attention_layernorm",
|
||||
"layers.{bid}.ffn_norm",
|
||||
"language_model.encoder.layers.{bid}.post_attention_layernorm",
|
||||
"model.layers.{bid}.ln2",
|
||||
"h.{bid}.ln_2",
|
||||
"model.layers.{bid}.ffn_norm",
|
||||
"transformer.decoder_layer.{bid}.rms_norm_2",
|
||||
"encoder.layers.{bid}.post_attention_layernorm",
|
||||
"transformer.layers.{bid}.ffn_norm",
|
||||
),
|
||||
# Post feed-forward norm
|
||||
|
||||
MODEL_TENSOR.FFN_PRE_NORM: (
|
||||
"model.layers.{bid}.pre_feedforward_layernorm", # gemma2
|
||||
"model.layers.{bid}.pre_feedforward_layernorm",
|
||||
),
|
||||
# Post feed-forward norm
|
||||
|
||||
MODEL_TENSOR.FFN_POST_NORM: (
|
||||
"model.layers.{bid}.post_feedforward_layernorm", # gemma2
|
||||
"model.layers.{bid}.post_feedforward_layernorm",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.FFN_GATE_INP: (
|
||||
"layers.{bid}.feed_forward.gate", # mixtral
|
||||
"model.layers.{bid}.block_sparse_moe.gate", # mixtral
|
||||
"model.layers.{bid}.mlp.gate", # qwen2moe
|
||||
"transformer.decoder_layer.{bid}.router", # Grok
|
||||
"transformer.blocks.{bid}.ffn.router.layer", # dbrx
|
||||
"layers.{bid}.feed_forward.gate",
|
||||
"model.layers.{bid}.block_sparse_moe.gate",
|
||||
"model.layers.{bid}.mlp.gate",
|
||||
"transformer.decoder_layer.{bid}.router",
|
||||
"transformer.blocks.{bid}.ffn.router.layer",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.FFN_GATE_INP_SHEXP: (
|
||||
"model.layers.{bid}.mlp.shared_expert_gate", # qwen2moe
|
||||
"model.layers.{bid}.mlp.shared_expert_gate",
|
||||
),
|
||||
# Feed-forward up
|
||||
|
||||
MODEL_TENSOR.FFN_UP: (
|
||||
"gpt_neox.layers.{bid}.mlp.dense_h_to_4h", # gptneox
|
||||
"transformer.h.{bid}.mlp.c_fc", # gpt2 jais
|
||||
"transformer.blocks.{bid}.ffn.up_proj", # mpt
|
||||
"transformer.h.{bid}.mlp.dense_h_to_4h", # falcon
|
||||
"h.{bid}.mlp.dense_h_to_4h", # bloom
|
||||
"model.layers.{bid}.mlp.up_proj", # llama-hf refact
|
||||
"layers.{bid}.feed_forward.w3", # llama-pth
|
||||
"encoder.layer.{bid}.intermediate.dense", # bert
|
||||
"transformer.h.{bid}.mlp.fc_in", # gpt-j
|
||||
"transformer.h.{bid}.mlp.linear_3", # refact
|
||||
"language_model.encoder.layers.{bid}.mlp.dense_h_to_4h", # persimmon
|
||||
"model.layers.{bid}.mlp.dense_h_to_4h", # persimmon
|
||||
"transformer.h.{bid}.mlp.w1", # qwen
|
||||
"h.{bid}.mlp.c_fc", # gpt2
|
||||
"transformer.h.{bid}.mlp.fc1", # phi2
|
||||
"model.layers.{bid}.mlp.fc1", # phi2
|
||||
"model.layers.{bid}.mlp.gate_up_proj", # phi3
|
||||
"model.layers.layers.{bid}.mlp.up_proj", # plamo
|
||||
"model.layers.{bid}.feed_forward.w3", # internlm2
|
||||
"encoder.layers.{bid}.mlp.fc11", # nomic-bert
|
||||
"model.layers.{bid}.mlp.c_fc", # starcoder2
|
||||
"encoder.layer.{bid}.mlp.gated_layers_v", # jina-bert-v2
|
||||
"model.layers.{bid}.residual_mlp.w3", # arctic
|
||||
"encoder.layers.{bid}.mlp.dense_h_to_4h", # chatglm
|
||||
"gpt_neox.layers.{bid}.mlp.dense_h_to_4h",
|
||||
"transformer.h.{bid}.mlp.c_fc",
|
||||
"transformer.blocks.{bid}.ffn.up_proj",
|
||||
"transformer.h.{bid}.mlp.dense_h_to_4h",
|
||||
"h.{bid}.mlp.dense_h_to_4h",
|
||||
"model.layers.{bid}.mlp.up_proj",
|
||||
"layers.{bid}.feed_forward.w3",
|
||||
"encoder.layer.{bid}.intermediate.dense",
|
||||
"transformer.h.{bid}.mlp.fc_in",
|
||||
"transformer.h.{bid}.mlp.linear_3",
|
||||
"language_model.encoder.layers.{bid}.mlp.dense_h_to_4h",
|
||||
"model.layers.{bid}.mlp.dense_h_to_4h",
|
||||
"transformer.h.{bid}.mlp.w1",
|
||||
"h.{bid}.mlp.c_fc",
|
||||
"transformer.h.{bid}.mlp.fc1",
|
||||
"model.layers.{bid}.mlp.fc1",
|
||||
"model.layers.{bid}.mlp.gate_up_proj",
|
||||
"model.layers.layers.{bid}.mlp.up_proj",
|
||||
"model.layers.{bid}.feed_forward.w3",
|
||||
"encoder.layers.{bid}.mlp.fc11",
|
||||
"model.layers.{bid}.mlp.c_fc",
|
||||
"encoder.layer.{bid}.mlp.gated_layers_v",
|
||||
"model.layers.{bid}.residual_mlp.w3",
|
||||
"encoder.layers.{bid}.mlp.dense_h_to_4h",
|
||||
"transformer.h.{bid}.mlp.c_fc_1",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.FFN_UP_EXP: (
|
||||
"layers.{bid}.feed_forward.experts.w3", # mixtral (merged)
|
||||
"transformer.decoder_layer.{bid}.moe.linear_v", # Grok (merged)
|
||||
"transformer.blocks.{bid}.ffn.experts.mlp.v1", # dbrx
|
||||
"model.layers.{bid}.mlp.experts.up_proj", # qwen2moe (merged)
|
||||
"layers.{bid}.feed_forward.experts.w3",
|
||||
"transformer.decoder_layer.{bid}.moe.linear_v",
|
||||
"transformer.blocks.{bid}.ffn.experts.mlp.v1",
|
||||
"model.layers.{bid}.mlp.experts.up_proj",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.FFN_UP_SHEXP: (
|
||||
"model.layers.{bid}.mlp.shared_expert.up_proj", # qwen2moe
|
||||
"model.layers.{bid}.mlp.shared_experts.up_proj", # deepseek2
|
||||
"model.layers.{bid}.mlp.shared_expert.up_proj",
|
||||
"model.layers.{bid}.mlp.shared_experts.up_proj",
|
||||
),
|
||||
# AWQ-activation gate
|
||||
MODEL_TENSOR.FFN_ACT: ("transformer.blocks.{bid}.ffn.act",), # mpt
|
||||
# Feed-forward gate
|
||||
|
||||
MODEL_TENSOR.FFN_ACT: (
|
||||
"transformer.blocks.{bid}.ffn.act",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.FFN_GATE: (
|
||||
"model.layers.{bid}.mlp.gate_proj", # llama-hf refact
|
||||
"layers.{bid}.feed_forward.w1", # llama-pth
|
||||
"transformer.h.{bid}.mlp.w2", # qwen
|
||||
"transformer.h.{bid}.mlp.c_fc2", # jais
|
||||
"model.layers.layers.{bid}.mlp.gate_proj", # plamo
|
||||
"model.layers.{bid}.feed_forward.w1", # internlm2
|
||||
"encoder.layers.{bid}.mlp.fc12", # nomic-bert
|
||||
"encoder.layer.{bid}.mlp.gated_layers_w", # jina-bert-v2
|
||||
"transformer.h.{bid}.mlp.linear_1", # refact
|
||||
"model.layers.{bid}.residual_mlp.w1", # arctic
|
||||
"model.layers.{bid}.mlp.gate_proj",
|
||||
"layers.{bid}.feed_forward.w1",
|
||||
"transformer.h.{bid}.mlp.w2",
|
||||
"transformer.h.{bid}.mlp.c_fc2",
|
||||
"model.layers.layers.{bid}.mlp.gate_proj",
|
||||
"model.layers.{bid}.feed_forward.w1",
|
||||
"encoder.layers.{bid}.mlp.fc12",
|
||||
"encoder.layer.{bid}.mlp.gated_layers_w",
|
||||
"transformer.h.{bid}.mlp.linear_1",
|
||||
"model.layers.{bid}.residual_mlp.w1",
|
||||
"transformer.h.{bid}.mlp.c_fc_0",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.FFN_GATE_EXP: (
|
||||
"layers.{bid}.feed_forward.experts.w1", # mixtral (merged)
|
||||
"transformer.decoder_layer.{bid}.moe.linear", # Grok (merged)
|
||||
"transformer.blocks.{bid}.ffn.experts.mlp.w1", # dbrx
|
||||
"model.layers.{bid}.mlp.experts.gate_proj", # qwen2moe (merged)
|
||||
"layers.{bid}.feed_forward.experts.w1",
|
||||
"transformer.decoder_layer.{bid}.moe.linear",
|
||||
"transformer.blocks.{bid}.ffn.experts.mlp.w1",
|
||||
"model.layers.{bid}.mlp.experts.gate_proj",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.FFN_GATE_SHEXP: (
|
||||
"model.layers.{bid}.mlp.shared_expert.gate_proj", # qwen2moe
|
||||
"model.layers.{bid}.mlp.shared_experts.gate_proj", # deepseek2
|
||||
"model.layers.{bid}.mlp.shared_expert.gate_proj",
|
||||
"model.layers.{bid}.mlp.shared_experts.gate_proj",
|
||||
),
|
||||
# Feed-forward down
|
||||
|
||||
MODEL_TENSOR.FFN_DOWN: (
|
||||
"gpt_neox.layers.{bid}.mlp.dense_4h_to_h", # gptneox
|
||||
"transformer.h.{bid}.mlp.c_proj", # gpt2 refact qwen jais
|
||||
"transformer.blocks.{bid}.ffn.down_proj", # mpt
|
||||
"transformer.h.{bid}.mlp.dense_4h_to_h", # falcon
|
||||
"h.{bid}.mlp.dense_4h_to_h", # bloom
|
||||
"model.layers.{bid}.mlp.down_proj", # llama-hf
|
||||
"layers.{bid}.feed_forward.w2", # llama-pth
|
||||
"encoder.layer.{bid}.output.dense", # bert
|
||||
"transformer.h.{bid}.mlp.fc_out", # gpt-j
|
||||
"language_model.encoder.layers.{bid}.mlp.dense_4h_to_h", # persimmon
|
||||
"model.layers.{bid}.mlp.dense_4h_to_h", # persimmon
|
||||
"h.{bid}.mlp.c_proj", # gpt2
|
||||
"transformer.h.{bid}.mlp.fc2", # phi2
|
||||
"model.layers.{bid}.mlp.fc2", # phi2
|
||||
"model.layers.layers.{bid}.mlp.down_proj", # plamo
|
||||
"model.layers.{bid}.feed_forward.w2", # internlm2
|
||||
"encoder.layers.{bid}.mlp.fc2", # nomic-bert
|
||||
"model.layers.{bid}.mlp.c_proj", # starcoder2
|
||||
"encoder.layer.{bid}.mlp.wo", # jina-bert-v2
|
||||
"transformer.layers.{bid}.ffn.proj_2", # openelm
|
||||
"model.layers.{bid}.residual_mlp.w2", # arctic
|
||||
"encoder.layer.{bid}.mlp.down_layer", # jina-bert-v2
|
||||
"encoder.layers.{bid}.mlp.dense_4h_to_h", # chatglm
|
||||
"gpt_neox.layers.{bid}.mlp.dense_4h_to_h",
|
||||
"transformer.h.{bid}.mlp.c_proj",
|
||||
"transformer.blocks.{bid}.ffn.down_proj",
|
||||
"transformer.h.{bid}.mlp.dense_4h_to_h",
|
||||
"h.{bid}.mlp.dense_4h_to_h",
|
||||
"model.layers.{bid}.mlp.down_proj",
|
||||
"layers.{bid}.feed_forward.w2",
|
||||
"encoder.layer.{bid}.output.dense",
|
||||
"transformer.h.{bid}.mlp.fc_out",
|
||||
"language_model.encoder.layers.{bid}.mlp.dense_4h_to_h",
|
||||
"model.layers.{bid}.mlp.dense_4h_to_h",
|
||||
"h.{bid}.mlp.c_proj",
|
||||
"transformer.h.{bid}.mlp.fc2",
|
||||
"model.layers.{bid}.mlp.fc2",
|
||||
"model.layers.layers.{bid}.mlp.down_proj",
|
||||
"model.layers.{bid}.feed_forward.w2",
|
||||
"encoder.layers.{bid}.mlp.fc2",
|
||||
"model.layers.{bid}.mlp.c_proj",
|
||||
"encoder.layer.{bid}.mlp.wo",
|
||||
"transformer.layers.{bid}.ffn.proj_2",
|
||||
"model.layers.{bid}.residual_mlp.w2",
|
||||
"encoder.layer.{bid}.mlp.down_layer",
|
||||
"encoder.layers.{bid}.mlp.dense_4h_to_h",
|
||||
"model.layers.h.{bid}.mlp.c_proj",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.FFN_DOWN_EXP: (
|
||||
"layers.{bid}.feed_forward.experts.w2", # mixtral (merged)
|
||||
"transformer.decoder_layer.{bid}.moe.linear_1", # Grok (merged)
|
||||
"transformer.blocks.{bid}.ffn.experts.mlp.w2", # dbrx
|
||||
"model.layers.{bid}.mlp.experts.down_proj", # qwen2moe (merged)
|
||||
"layers.{bid}.feed_forward.experts.w2",
|
||||
"transformer.decoder_layer.{bid}.moe.linear_1",
|
||||
"transformer.blocks.{bid}.ffn.experts.mlp.w2",
|
||||
"model.layers.{bid}.mlp.experts.down_proj",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.FFN_DOWN_SHEXP: (
|
||||
"model.layers.{bid}.mlp.shared_expert.down_proj", # qwen2moe
|
||||
"model.layers.{bid}.mlp.shared_experts.down_proj", # deepseek2
|
||||
"model.layers.{bid}.mlp.shared_expert.down_proj",
|
||||
"model.layers.{bid}.mlp.shared_experts.down_proj",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.ATTN_Q_NORM: (
|
||||
"language_model.encoder.layers.{bid}.self_attention.q_layernorm",
|
||||
"model.layers.{bid}.self_attn.q_layernorm", # persimmon
|
||||
"model.layers.{bid}.self_attn.q_norm", # cohere
|
||||
"transformer.blocks.{bid}.attn.q_ln", # sea-lion
|
||||
"encoder.layer.{bid}.attention.self.layer_norm_q", # jina-bert-v2
|
||||
"transformer.layers.{bid}.attn.q_norm", # openelm
|
||||
"model.layers.{bid}.self_attn.q_layernorm",
|
||||
"model.layers.{bid}.self_attn.q_norm",
|
||||
"transformer.blocks.{bid}.attn.q_ln",
|
||||
"encoder.layer.{bid}.attention.self.layer_norm_q",
|
||||
"transformer.layers.{bid}.attn.q_norm",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.ATTN_K_NORM: (
|
||||
"language_model.encoder.layers.{bid}.self_attention.k_layernorm",
|
||||
"model.layers.{bid}.self_attn.k_layernorm", # persimmon
|
||||
"model.layers.{bid}.self_attn.k_norm", # cohere
|
||||
"transformer.blocks.{bid}.attn.k_ln", # sea-lion
|
||||
"encoder.layer.{bid}.attention.self.layer_norm_k", # jina-bert-v2
|
||||
"transformer.layers.{bid}.attn.k_norm", # openelm
|
||||
"model.layers.{bid}.self_attn.k_layernorm",
|
||||
"model.layers.{bid}.self_attn.k_norm",
|
||||
"transformer.blocks.{bid}.attn.k_ln",
|
||||
"encoder.layer.{bid}.attention.self.layer_norm_k",
|
||||
"transformer.layers.{bid}.attn.k_norm",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.ROPE_FREQS: (
|
||||
"language_model.encoder.layers.{bid}.self_attention.rotary_emb.inv_freq", # persimmon
|
||||
"language_model.encoder.layers.{bid}.self_attention.rotary_emb.inv_freq",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.LAYER_OUT_NORM: (
|
||||
"encoder.layer.{bid}.output.LayerNorm", # bert
|
||||
"encoder.layers.{bid}.norm2", # nomic-bert
|
||||
"transformer.decoder_layer.{bid}.rms_norm_3", # Grok
|
||||
"encoder.layer.{bid}.mlp.layernorm", # jina-bert-v2
|
||||
"encoder.layer.{bid}.layer_norm_2", # jina-v2-code
|
||||
"encoder.layer.{bid}.output.LayerNorm",
|
||||
"encoder.layers.{bid}.norm2",
|
||||
"transformer.decoder_layer.{bid}.rms_norm_3",
|
||||
"encoder.layer.{bid}.mlp.layernorm",
|
||||
"encoder.layer.{bid}.layer_norm_2"
|
||||
),
|
||||
|
||||
MODEL_TENSOR.SSM_IN: (
|
||||
"model.layers.{bid}.in_proj",
|
||||
"backbone.layers.{bid}.mixer.in_proj",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.SSM_CONV1D: (
|
||||
"model.layers.{bid}.conv1d",
|
||||
"backbone.layers.{bid}.mixer.conv1d",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.SSM_X: (
|
||||
"model.layers.{bid}.x_proj",
|
||||
"backbone.layers.{bid}.mixer.x_proj",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.SSM_DT: (
|
||||
"model.layers.{bid}.dt_proj",
|
||||
"backbone.layers.{bid}.mixer.dt_proj",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.SSM_A: (
|
||||
"model.layers.{bid}.A_log",
|
||||
"backbone.layers.{bid}.mixer.A_log",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.SSM_D: (
|
||||
"model.layers.{bid}.D",
|
||||
"backbone.layers.{bid}.mixer.D",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.SSM_OUT: (
|
||||
"model.layers.{bid}.out_proj",
|
||||
"backbone.layers.{bid}.mixer.out_proj",
|
||||
),
|
||||
MODEL_TENSOR.ATTN_Q_A: ("model.layers.{bid}.self_attn.q_a_proj",), # deepseek2
|
||||
MODEL_TENSOR.ATTN_Q_B: ("model.layers.{bid}.self_attn.q_b_proj",), # deepseek2
|
||||
|
||||
MODEL_TENSOR.ATTN_Q_A: (
|
||||
"model.layers.{bid}.self_attn.q_a_proj",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.ATTN_Q_B: (
|
||||
"model.layers.{bid}.self_attn.q_b_proj",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.ATTN_KV_A_MQA: (
|
||||
"model.layers.{bid}.self_attn.kv_a_proj_with_mqa", # deepseek2
|
||||
"model.layers.{bid}.self_attn.kv_a_proj_with_mqa",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.ATTN_KV_B: (
|
||||
"model.layers.{bid}.self_attn.kv_b_proj", # deepseek2
|
||||
"model.layers.{bid}.self_attn.kv_b_proj",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.ATTN_Q_A_NORM: (
|
||||
"model.layers.{bid}.self_attn.q_a_layernorm", # deepseek2
|
||||
"model.layers.{bid}.self_attn.q_a_layernorm",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.ATTN_KV_A_NORM: (
|
||||
"model.layers.{bid}.self_attn.kv_a_layernorm", # deepseek2
|
||||
"model.layers.{bid}.self_attn.kv_a_layernorm",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.ATTN_SUB_NORM: (
|
||||
"model.layers.{bid}.self_attn.inner_attn_ln", # bitnet
|
||||
"model.layers.{bid}.self_attn.inner_attn_ln",
|
||||
),
|
||||
MODEL_TENSOR.FFN_SUB_NORM: ("model.layers.{bid}.mlp.ffn_layernorm",), # bitnet
|
||||
MODEL_TENSOR.DEC_ATTN_NORM: ("decoder.block.{bid}.layer.0.layer_norm",), # t5
|
||||
MODEL_TENSOR.DEC_ATTN_Q: ("decoder.block.{bid}.layer.0.SelfAttention.q",), # t5
|
||||
MODEL_TENSOR.DEC_ATTN_K: ("decoder.block.{bid}.layer.0.SelfAttention.k",), # t5
|
||||
MODEL_TENSOR.DEC_ATTN_V: ("decoder.block.{bid}.layer.0.SelfAttention.v",), # t5
|
||||
|
||||
MODEL_TENSOR.FFN_SUB_NORM: (
|
||||
"model.layers.{bid}.mlp.ffn_layernorm",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.DEC_ATTN_NORM: (
|
||||
"decoder.block.{bid}.layer.0.layer_norm",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.DEC_ATTN_Q: (
|
||||
"decoder.block.{bid}.layer.0.SelfAttention.q",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.DEC_ATTN_K: (
|
||||
"decoder.block.{bid}.layer.0.SelfAttention.k",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.DEC_ATTN_V: (
|
||||
"decoder.block.{bid}.layer.0.SelfAttention.v",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.DEC_ATTN_OUT: (
|
||||
"decoder.block.{bid}.layer.0.SelfAttention.o", # t5
|
||||
"decoder.block.{bid}.layer.0.SelfAttention.o",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.DEC_ATTN_REL_B: (
|
||||
"decoder.block.{bid}.layer.0.SelfAttention.relative_attention_bias", # t5
|
||||
"decoder.block.{bid}.layer.0.SelfAttention.relative_attention_bias",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.DEC_CROSS_ATTN_NORM: (
|
||||
"decoder.block.{bid}.layer.1.layer_norm", # t5
|
||||
"decoder.block.{bid}.layer.1.layer_norm",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.DEC_CROSS_ATTN_Q: (
|
||||
"decoder.block.{bid}.layer.1.EncDecAttention.q", # t5
|
||||
"decoder.block.{bid}.layer.1.EncDecAttention.q",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.DEC_CROSS_ATTN_K: (
|
||||
"decoder.block.{bid}.layer.1.EncDecAttention.k", # t5
|
||||
"decoder.block.{bid}.layer.1.EncDecAttention.k",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.DEC_CROSS_ATTN_V: (
|
||||
"decoder.block.{bid}.layer.1.EncDecAttention.v", # t5
|
||||
"decoder.block.{bid}.layer.1.EncDecAttention.v",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.DEC_CROSS_ATTN_OUT: (
|
||||
"decoder.block.{bid}.layer.1.EncDecAttention.o", # t5
|
||||
"decoder.block.{bid}.layer.1.EncDecAttention.o",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.DEC_CROSS_ATTN_REL_B: (
|
||||
"decoder.block.{bid}.layer.1.EncDecAttention.relative_attention_bias", # t5
|
||||
"decoder.block.{bid}.layer.1.EncDecAttention.relative_attention_bias",
|
||||
),
|
||||
MODEL_TENSOR.DEC_FFN_NORM: ("decoder.block.{bid}.layer.2.layer_norm",), # t5
|
||||
|
||||
MODEL_TENSOR.DEC_FFN_NORM: (
|
||||
"decoder.block.{bid}.layer.2.layer_norm",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.DEC_FFN_GATE: (
|
||||
"decoder.block.{bid}.layer.2.DenseReluDense.wi_0", # flan-t5
|
||||
"decoder.block.{bid}.layer.2.DenseReluDense.wi_0",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.DEC_FFN_UP: (
|
||||
"decoder.block.{bid}.layer.2.DenseReluDense.wi", # t5
|
||||
"decoder.block.{bid}.layer.2.DenseReluDense.wi_1", # flan-t5
|
||||
"decoder.block.{bid}.layer.2.DenseReluDense.wi",
|
||||
"decoder.block.{bid}.layer.2.DenseReluDense.wi_1",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.DEC_FFN_DOWN: (
|
||||
"decoder.block.{bid}.layer.2.DenseReluDense.wo", # t5
|
||||
"decoder.block.{bid}.layer.2.DenseReluDense.wo",
|
||||
),
|
||||
MODEL_TENSOR.DEC_OUTPUT_NORM: ("decoder.final_layer_norm",), # t5
|
||||
MODEL_TENSOR.ENC_ATTN_NORM: ("encoder.block.{bid}.layer.0.layer_norm",), # t5
|
||||
MODEL_TENSOR.ENC_ATTN_Q: ("encoder.block.{bid}.layer.0.SelfAttention.q",), # t5
|
||||
MODEL_TENSOR.ENC_ATTN_K: ("encoder.block.{bid}.layer.0.SelfAttention.k",), # t5
|
||||
MODEL_TENSOR.ENC_ATTN_V: ("encoder.block.{bid}.layer.0.SelfAttention.v",), # t5
|
||||
|
||||
MODEL_TENSOR.DEC_OUTPUT_NORM: (
|
||||
"decoder.final_layer_norm",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.ENC_ATTN_NORM: (
|
||||
"encoder.block.{bid}.layer.0.layer_norm",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.ENC_ATTN_Q: (
|
||||
"encoder.block.{bid}.layer.0.SelfAttention.q",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.ENC_ATTN_K: (
|
||||
"encoder.block.{bid}.layer.0.SelfAttention.k",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.ENC_ATTN_V: (
|
||||
"encoder.block.{bid}.layer.0.SelfAttention.v",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.ENC_ATTN_OUT: (
|
||||
"encoder.block.{bid}.layer.0.SelfAttention.o", # t5
|
||||
"encoder.block.{bid}.layer.0.SelfAttention.o",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.ENC_ATTN_REL_B: (
|
||||
"encoder.block.{bid}.layer.0.SelfAttention.relative_attention_bias", # t5
|
||||
"encoder.block.{bid}.layer.0.SelfAttention.relative_attention_bias",
|
||||
),
|
||||
MODEL_TENSOR.ENC_FFN_NORM: ("encoder.block.{bid}.layer.1.layer_norm",), # t5
|
||||
|
||||
MODEL_TENSOR.ENC_FFN_NORM: (
|
||||
"encoder.block.{bid}.layer.1.layer_norm",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.ENC_FFN_GATE: (
|
||||
"encoder.block.{bid}.layer.1.DenseReluDense.wi_0", # flan-t5
|
||||
"encoder.block.{bid}.layer.1.DenseReluDense.wi_0",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.ENC_FFN_UP: (
|
||||
"encoder.block.{bid}.layer.1.DenseReluDense.wi", # t5
|
||||
"encoder.block.{bid}.layer.1.DenseReluDense.wi_1", # flan-t5
|
||||
"encoder.block.{bid}.layer.1.DenseReluDense.wi",
|
||||
"encoder.block.{bid}.layer.1.DenseReluDense.wi_1",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.ENC_FFN_DOWN: (
|
||||
"encoder.block.{bid}.layer.1.DenseReluDense.wo", # t5
|
||||
"encoder.block.{bid}.layer.1.DenseReluDense.wo",
|
||||
),
|
||||
|
||||
MODEL_TENSOR.ENC_OUTPUT_NORM: (
|
||||
"encoder.final_layer_norm",
|
||||
),
|
||||
MODEL_TENSOR.ENC_OUTPUT_NORM: ("encoder.final_layer_norm",), # t5
|
||||
}
|
||||
|
||||
# architecture-specific block mappings
|
||||
arch_block_mappings_cfg: dict[MODEL_ARCH, dict[MODEL_TENSOR, tuple[str, ...]]] = {
|
||||
MODEL_ARCH.ARCTIC: {
|
||||
MODEL_TENSOR.FFN_NORM: ("model.layers.{bid}.residual_layernorm",),
|
||||
MODEL_TENSOR.FFN_NORM_EXP: ("model.layers.{bid}.post_attention_layernorm",),
|
||||
MODEL_TENSOR.FFN_NORM: (
|
||||
"model.layers.{bid}.residual_layernorm",
|
||||
),
|
||||
MODEL_TENSOR.FFN_NORM_EXP: (
|
||||
"model.layers.{bid}.post_attention_layernorm",
|
||||
),
|
||||
},
|
||||
}
|
||||
|
||||
|
|
@ -490,35 +588,31 @@ def __init__(self, arch: MODEL_ARCH, n_blocks: int):
|
|||
if tensor not in MODEL_TENSORS[arch]:
|
||||
continue
|
||||
|
||||
tensor_name = TENSOR_NAMES[tensor].format(bid=bid)
|
||||
tensor_name = TENSOR_NAMES[tensor].format(bid = bid)
|
||||
self.mapping[tensor_name] = (tensor, tensor_name)
|
||||
for key in keys:
|
||||
key = key.format(bid=bid)
|
||||
key = key.format(bid = bid)
|
||||
self.mapping[key] = (tensor, tensor_name)
|
||||
|
||||
def get_type_and_name(
|
||||
self, key: str, try_suffixes: Sequence[str] = ()
|
||||
) -> tuple[MODEL_TENSOR, str] | None:
|
||||
def get_type_and_name(self, key: str, try_suffixes: Sequence[str] = ()) -> tuple[MODEL_TENSOR, str] | None:
|
||||
result = self.mapping.get(key)
|
||||
if result is not None:
|
||||
return result
|
||||
for suffix in try_suffixes:
|
||||
if key.endswith(suffix):
|
||||
result = self.mapping.get(key[: -len(suffix)])
|
||||
result = self.mapping.get(key[:-len(suffix)])
|
||||
if result is not None:
|
||||
return result[0], result[1] + suffix
|
||||
return None
|
||||
|
||||
def get_name(self, key: str, try_suffixes: Sequence[str] = ()) -> str | None:
|
||||
result = self.get_type_and_name(key, try_suffixes=try_suffixes)
|
||||
result = self.get_type_and_name(key, try_suffixes = try_suffixes)
|
||||
if result is None:
|
||||
return None
|
||||
return result[1]
|
||||
|
||||
def get_type(
|
||||
self, key: str, try_suffixes: Sequence[str] = ()
|
||||
) -> MODEL_TENSOR | None:
|
||||
result = self.get_type_and_name(key, try_suffixes=try_suffixes)
|
||||
def get_type(self, key: str, try_suffixes: Sequence[str] = ()) -> MODEL_TENSOR | None:
|
||||
result = self.get_type_and_name(key, try_suffixes = try_suffixes)
|
||||
if result is None:
|
||||
return None
|
||||
return result[0]
|
||||
|
|
@ -535,6 +629,5 @@ def __contains__(self, key: str) -> bool:
|
|||
def __repr__(self) -> str:
|
||||
return repr(self.mapping)
|
||||
|
||||
|
||||
def get_tensor_name_map(arch: MODEL_ARCH, n_blocks: int) -> TensorNameMap:
|
||||
return TensorNameMap(arch, n_blocks)
|
||||
return TensorNameMap(arch, n_blocks)
|
||||
Loading…
Reference in New Issue