File size: 2,031 Bytes
115147b |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 |
# Copyright (c) Meta Platforms, Inc. and affiliates.
# All rights reserved.
#
# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.
from transformers import PretrainedConfig
class MobileLLMConfig(PretrainedConfig):
model_type = "mobilellm"
keys_to_ignore_at_inference = ["past_key_values"]
def __init__(
self,
attention_bias=False,
bos_token_id=1,
eos_token_id=2,
hidden_act="silu",
hidden_size=1600,
initializer_range=0.02,
intermediate_size=4352,
num_hidden_layers=54,
num_attention_heads=25,
num_key_value_heads=5,
pretraining_tp=1,
rms_norm_eps=1e-5,
rope_scaling=None,
rope_theta=10000.0,
max_position_embeddings=2048,
tie_word_embeddings=False,
use_cache=True,
bf16=False,
fp16=True,
fp32=False,
vocab_size=32000,
share_embedding=True,
**kwargs,
):
self.attention_bias = attention_bias
self.bos_token_id = bos_token_id
self.eos_token_id = eos_token_id
self.hidden_act = hidden_act
self.hidden_size = hidden_size
self.initializer_range = initializer_range
self.intermediate_size = intermediate_size
self.num_hidden_layers = num_hidden_layers
self.num_attention_heads = num_attention_heads
self.num_key_value_heads = num_key_value_heads
self.pretraining_tp = pretraining_tp
self.rms_norm_eps = rms_norm_eps
self.rope_scaling = rope_scaling
self.rope_theta = rope_theta
self.max_position_embeddings = max_position_embeddings
self.use_cache = use_cache
self.bf16 = bf16
self.fp16 = fp16
self.fp32 = fp32
self.vocab_size = vocab_size
self.share_embedding = share_embedding
super().__init__(
tie_word_embeddings=tie_word_embeddings,
**kwargs
)
|