oopsie
Browse filesThis view is limited to 50 files because it contains too many changes.
See raw diff
- .gitattributes +0 -35
- README.md +0 -8
- __init__.py +0 -2
- config.json +0 -37
- configuration_grok.py +0 -151
- generation_config.json +0 -7
- model-00001-of-00019.safetensors +0 -3
- model-00002-of-00019.safetensors +0 -3
- model-00003-of-00019.safetensors +0 -3
- model-00004-of-00019.safetensors +0 -3
- model-00005-of-00019.safetensors +0 -3
- model-00006-of-00019.safetensors +0 -3
- model-00007-of-00019.safetensors +0 -3
- model-00008-of-00019.safetensors +0 -3
- model-00009-of-00019.safetensors +0 -3
- model-00010-of-00019.safetensors +0 -3
- model-00011-of-00019.safetensors +0 -3
- model-00012-of-00019.safetensors +0 -3
- model-00013-of-00019.safetensors +0 -3
- model-00014-of-00019.safetensors +0 -3
- model-00015-of-00019.safetensors +0 -3
- model-00016-of-00019.safetensors +0 -3
- model-00017-of-00019.safetensors +0 -3
- model-00018-of-00019.safetensors +0 -3
- model-00019-of-00019.safetensors +0 -3
- model.safetensors.index.json +0 -777
- modeling_grok.py +0 -838
- pytorch_model-00001-of-00019.bin +0 -3
- pytorch_model-00002-of-00019.bin +0 -3
- pytorch_model-00003-of-00019.bin +0 -3
- pytorch_model-00004-of-00019.bin +0 -3
- pytorch_model-00005-of-00019.bin +0 -3
- pytorch_model-00006-of-00019.bin +0 -3
- pytorch_model-00007-of-00019.bin +0 -3
- pytorch_model-00008-of-00019.bin +0 -3
- pytorch_model-00009-of-00019.bin +0 -3
- pytorch_model-00010-of-00019.bin +0 -3
- pytorch_model-00011-of-00019.bin +0 -3
- pytorch_model-00012-of-00019.bin +0 -3
- pytorch_model-00013-of-00019.bin +0 -3
- pytorch_model-00014-of-00019.bin +0 -3
- pytorch_model-00015-of-00019.bin +0 -3
- pytorch_model-00016-of-00019.bin +0 -3
- pytorch_model-00017-of-00019.bin +0 -3
- pytorch_model-00018-of-00019.bin +0 -3
- pytorch_model-00019-of-00019.bin +0 -3
- pytorch_model.bin.index.json +0 -0
- special_tokens_map.json +0 -23
- tokenizer.json +0 -0
- tokenizer.model +0 -3
.gitattributes
DELETED
|
@@ -1,35 +0,0 @@
|
|
| 1 |
-
*.7z filter=lfs diff=lfs merge=lfs -text
|
| 2 |
-
*.arrow filter=lfs diff=lfs merge=lfs -text
|
| 3 |
-
*.bin filter=lfs diff=lfs merge=lfs -text
|
| 4 |
-
*.bz2 filter=lfs diff=lfs merge=lfs -text
|
| 5 |
-
*.ckpt filter=lfs diff=lfs merge=lfs -text
|
| 6 |
-
*.ftz filter=lfs diff=lfs merge=lfs -text
|
| 7 |
-
*.gz filter=lfs diff=lfs merge=lfs -text
|
| 8 |
-
*.h5 filter=lfs diff=lfs merge=lfs -text
|
| 9 |
-
*.joblib filter=lfs diff=lfs merge=lfs -text
|
| 10 |
-
*.lfs.* filter=lfs diff=lfs merge=lfs -text
|
| 11 |
-
*.mlmodel filter=lfs diff=lfs merge=lfs -text
|
| 12 |
-
*.model filter=lfs diff=lfs merge=lfs -text
|
| 13 |
-
*.msgpack filter=lfs diff=lfs merge=lfs -text
|
| 14 |
-
*.npy filter=lfs diff=lfs merge=lfs -text
|
| 15 |
-
*.npz filter=lfs diff=lfs merge=lfs -text
|
| 16 |
-
*.onnx filter=lfs diff=lfs merge=lfs -text
|
| 17 |
-
*.ot filter=lfs diff=lfs merge=lfs -text
|
| 18 |
-
*.parquet filter=lfs diff=lfs merge=lfs -text
|
| 19 |
-
*.pb filter=lfs diff=lfs merge=lfs -text
|
| 20 |
-
*.pickle filter=lfs diff=lfs merge=lfs -text
|
| 21 |
-
*.pkl filter=lfs diff=lfs merge=lfs -text
|
| 22 |
-
*.pt filter=lfs diff=lfs merge=lfs -text
|
| 23 |
-
*.pth filter=lfs diff=lfs merge=lfs -text
|
| 24 |
-
*.rar filter=lfs diff=lfs merge=lfs -text
|
| 25 |
-
*.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 26 |
-
saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
| 27 |
-
*.tar.* filter=lfs diff=lfs merge=lfs -text
|
| 28 |
-
*.tar filter=lfs diff=lfs merge=lfs -text
|
| 29 |
-
*.tflite filter=lfs diff=lfs merge=lfs -text
|
| 30 |
-
*.tgz filter=lfs diff=lfs merge=lfs -text
|
| 31 |
-
*.wasm filter=lfs diff=lfs merge=lfs -text
|
| 32 |
-
*.xz filter=lfs diff=lfs merge=lfs -text
|
| 33 |
-
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
-
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
-
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
README.md
DELETED
|
@@ -1,8 +0,0 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: apache-2.0
|
| 3 |
-
library_name: transformers
|
| 4 |
-
---
|
| 5 |
-
|
| 6 |
-
Unofficial dequantized weight of [grok-1](https://huggingface.co/xai-org/grok-1) in HF Transformers format.
|
| 7 |
-
|
| 8 |
-
The weights are converted using the [script here](https://gist.github.com/chu-tianxiang/ec310e15d56949fd0f351cb5f65ee7a1) ran inside the [grok-1 repo](https://github.com/xai-org/grok-1). Since downloading the dequantized weight needs twice as much time, it's recommended to download the original weight and convert on your own.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
__init__.py
DELETED
|
@@ -1,2 +0,0 @@
|
|
| 1 |
-
from .configuration_grok import *
|
| 2 |
-
from .modeling_grok import *
|
|
|
|
|
|
|
|
|
config.json
DELETED
|
@@ -1,37 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"_name_or_path": "grok",
|
| 3 |
-
"architectures": [
|
| 4 |
-
"GrokForCausalLM"
|
| 5 |
-
],
|
| 6 |
-
"attention_dropout": 0.0,
|
| 7 |
-
"attn_output_multiplier": 0.08838834764831845,
|
| 8 |
-
"auto_map": {
|
| 9 |
-
"AutoConfig": "configuration_grok.GrokConfig",
|
| 10 |
-
"AutoModelForCausalLM": "modeling_grok.GrokForCausalLM"
|
| 11 |
-
},
|
| 12 |
-
"bos_token_id": 1,
|
| 13 |
-
"embedding_multiplier_scale": 78.38367176906169,
|
| 14 |
-
"eos_token_id": 2,
|
| 15 |
-
"hidden_act": "gelu_new",
|
| 16 |
-
"hidden_size": 6144,
|
| 17 |
-
"initializer_range": 0.02,
|
| 18 |
-
"intermediate_size": 32768,
|
| 19 |
-
"max_position_embeddings": 8192,
|
| 20 |
-
"model_type": "grok",
|
| 21 |
-
"num_attention_heads": 48,
|
| 22 |
-
"num_experts_per_tok": 2,
|
| 23 |
-
"num_hidden_layers": 64,
|
| 24 |
-
"num_key_value_heads": 8,
|
| 25 |
-
"num_local_experts": 8,
|
| 26 |
-
"output_multiplier_scale": 0.5773502691896257,
|
| 27 |
-
"output_router_logits": false,
|
| 28 |
-
"pad_token_id": 0,
|
| 29 |
-
"rms_norm_eps": 1e-05,
|
| 30 |
-
"rope_theta": 10000.0,
|
| 31 |
-
"router_aux_loss_coef": 0.02,
|
| 32 |
-
"sliding_window": null,
|
| 33 |
-
"torch_dtype": "bfloat16",
|
| 34 |
-
"transformers_version": "4.38.2",
|
| 35 |
-
"use_cache": true,
|
| 36 |
-
"vocab_size": 131072
|
| 37 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
configuration_grok.py
DELETED
|
@@ -1,151 +0,0 @@
|
|
| 1 |
-
# coding=utf-8
|
| 2 |
-
# Copyright 2023 Mixtral AI and the HuggingFace Inc. team. All rights reserved.
|
| 3 |
-
#
|
| 4 |
-
# Licensed under the Apache License, Version 2.0 (the "License");
|
| 5 |
-
# you may not use this file except in compliance with the License.
|
| 6 |
-
# You may obtain a copy of the License at
|
| 7 |
-
#
|
| 8 |
-
# http://www.apache.org/licenses/LICENSE-2.0
|
| 9 |
-
#
|
| 10 |
-
# Unless required by applicable law or agreed to in writing, software
|
| 11 |
-
# distributed under the License is distributed on an "AS IS" BASIS,
|
| 12 |
-
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
| 13 |
-
# See the License for the specific language governing permissions and
|
| 14 |
-
# limitations under the License.
|
| 15 |
-
""" Grok model configuration"""
|
| 16 |
-
|
| 17 |
-
from transformers.configuration_utils import PretrainedConfig
|
| 18 |
-
from transformers.utils import logging
|
| 19 |
-
|
| 20 |
-
|
| 21 |
-
logger = logging.get_logger(__name__)
|
| 22 |
-
|
| 23 |
-
|
| 24 |
-
class GrokConfig(PretrainedConfig):
|
| 25 |
-
r"""
|
| 26 |
-
This is the configuration class to store the configuration of a [`GrokModel`].
|
| 27 |
-
|
| 28 |
-
Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
|
| 29 |
-
documentation from [`PretrainedConfig`] for more information.
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
Args:
|
| 33 |
-
vocab_size (`int`, *optional*, defaults to 32000):
|
| 34 |
-
Vocabulary size of the Mixtral model. Defines the number of different tokens that can be represented by the
|
| 35 |
-
`inputs_ids` passed when calling [`MixtralModel`]
|
| 36 |
-
hidden_size (`int`, *optional*, defaults to 4096):
|
| 37 |
-
Dimension of the hidden representations.
|
| 38 |
-
intermediate_size (`int`, *optional*, defaults to 14336):
|
| 39 |
-
Dimension of the MLP representations.
|
| 40 |
-
num_hidden_layers (`int`, *optional*, defaults to 32):
|
| 41 |
-
Number of hidden layers in the Transformer encoder.
|
| 42 |
-
num_attention_heads (`int`, *optional*, defaults to 32):
|
| 43 |
-
Number of attention heads for each attention layer in the Transformer encoder.
|
| 44 |
-
num_key_value_heads (`int`, *optional*, defaults to 8):
|
| 45 |
-
This is the number of key_value heads that should be used to implement Grouped Query Attention. If
|
| 46 |
-
`num_key_value_heads=num_attention_heads`, the model will use Multi Head Attention (MHA), if
|
| 47 |
-
`num_key_value_heads=1 the model will use Multi Query Attention (MQA) otherwise GQA is used. When
|
| 48 |
-
converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed
|
| 49 |
-
by meanpooling all the original heads within that group. For more details checkout [this
|
| 50 |
-
paper](https://arxiv.org/pdf/2305.13245.pdf). If it is not specified, will default to `8`.
|
| 51 |
-
hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
|
| 52 |
-
The non-linear activation function (function or string) in the decoder.
|
| 53 |
-
max_position_embeddings (`int`, *optional*, defaults to `4096*32`):
|
| 54 |
-
The maximum sequence length that this model might ever be used with. Mixtral's sliding window attention
|
| 55 |
-
allows sequence of up to 4096*32 tokens.
|
| 56 |
-
initializer_range (`float`, *optional*, defaults to 0.02):
|
| 57 |
-
The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
|
| 58 |
-
rms_norm_eps (`float`, *optional*, defaults to 1e-05):
|
| 59 |
-
The epsilon used by the rms normalization layers.
|
| 60 |
-
use_cache (`bool`, *optional*, defaults to `True`):
|
| 61 |
-
Whether or not the model should return the last key/values attentions (not used by all models). Only
|
| 62 |
-
relevant if `config.is_decoder=True`.
|
| 63 |
-
pad_token_id (`int`, *optional*):
|
| 64 |
-
The id of the padding token.
|
| 65 |
-
bos_token_id (`int`, *optional*, defaults to 1):
|
| 66 |
-
The id of the "beginning-of-sequence" token.
|
| 67 |
-
eos_token_id (`int`, *optional*, defaults to 2):
|
| 68 |
-
The id of the "end-of-sequence" token.
|
| 69 |
-
tie_word_embeddings (`bool`, *optional*, defaults to `True`):
|
| 70 |
-
Whether the model's input and output word embeddings should be tied.
|
| 71 |
-
rope_theta (`float`, *optional*, defaults to 100000.0):
|
| 72 |
-
The base period of the RoPE embeddings.
|
| 73 |
-
attention_dropout (`float`, *optional*, defaults to 0.0):
|
| 74 |
-
The dropout ratio for the attention probabilities.
|
| 75 |
-
num_experts_per_tok (`int`, *optional*, defaults to 2):
|
| 76 |
-
The number of experts to root per-token, can be also interpreted as the `top-p` routing
|
| 77 |
-
parameter
|
| 78 |
-
num_local_experts (`int`, *optional*, defaults to 8):
|
| 79 |
-
Number of experts per Sparse MLP layer.
|
| 80 |
-
output_router_logits (`bool`, *optional*, defaults to `False`):
|
| 81 |
-
Whether or not the router logits should be returned by the model. Enabeling this will also
|
| 82 |
-
allow the model to output the auxiliary loss. See [here]() for more details
|
| 83 |
-
router_aux_loss_coef (`float`, *optional*, defaults to 0.001):
|
| 84 |
-
The aux loss factor for the total loss.
|
| 85 |
-
|
| 86 |
-
"""
|
| 87 |
-
|
| 88 |
-
model_type = "grok"
|
| 89 |
-
keys_to_ignore_at_inference = ["past_key_values"]
|
| 90 |
-
|
| 91 |
-
def __init__(
|
| 92 |
-
self,
|
| 93 |
-
vocab_size=131072,
|
| 94 |
-
hidden_size=6144,
|
| 95 |
-
intermediate_size=32768,
|
| 96 |
-
num_hidden_layers=64,
|
| 97 |
-
num_attention_heads=48,
|
| 98 |
-
num_key_value_heads=8,
|
| 99 |
-
hidden_act="silu",
|
| 100 |
-
max_position_embeddings=4096,
|
| 101 |
-
initializer_range=0.02,
|
| 102 |
-
rms_norm_eps=1e-5,
|
| 103 |
-
use_cache=True,
|
| 104 |
-
pad_token_id=0,
|
| 105 |
-
bos_token_id=1,
|
| 106 |
-
eos_token_id=2,
|
| 107 |
-
tie_word_embeddings=True,
|
| 108 |
-
rope_theta=1e5,
|
| 109 |
-
attention_dropout=0.0,
|
| 110 |
-
num_experts_per_tok=2,
|
| 111 |
-
num_local_experts=8,
|
| 112 |
-
output_router_logits=False,
|
| 113 |
-
router_aux_loss_coef=0.001,
|
| 114 |
-
output_multiplier_scale=0.5773502691896257,
|
| 115 |
-
embedding_multiplier_scale=78.38367176906169,
|
| 116 |
-
attn_output_multiplier=0.08838834764831845,
|
| 117 |
-
**kwargs,
|
| 118 |
-
):
|
| 119 |
-
self.vocab_size = vocab_size
|
| 120 |
-
self.max_position_embeddings = max_position_embeddings
|
| 121 |
-
self.hidden_size = hidden_size
|
| 122 |
-
self.intermediate_size = intermediate_size
|
| 123 |
-
self.num_hidden_layers = num_hidden_layers
|
| 124 |
-
self.num_attention_heads = num_attention_heads
|
| 125 |
-
|
| 126 |
-
# for backward compatibility
|
| 127 |
-
if num_key_value_heads is None:
|
| 128 |
-
num_key_value_heads = num_attention_heads
|
| 129 |
-
|
| 130 |
-
self.num_key_value_heads = num_key_value_heads
|
| 131 |
-
self.hidden_act = hidden_act
|
| 132 |
-
self.initializer_range = initializer_range
|
| 133 |
-
self.rms_norm_eps = rms_norm_eps
|
| 134 |
-
self.use_cache = use_cache
|
| 135 |
-
self.rope_theta = rope_theta
|
| 136 |
-
self.attention_dropout = attention_dropout
|
| 137 |
-
|
| 138 |
-
self.num_experts_per_tok = num_experts_per_tok
|
| 139 |
-
self.num_local_experts = num_local_experts
|
| 140 |
-
self.output_router_logits = output_router_logits
|
| 141 |
-
self.router_aux_loss_coef = router_aux_loss_coef
|
| 142 |
-
self.output_multiplier_scale = output_multiplier_scale
|
| 143 |
-
self.embedding_multiplier_scale = embedding_multiplier_scale
|
| 144 |
-
self.attn_output_multiplier = attn_output_multiplier
|
| 145 |
-
super().__init__(
|
| 146 |
-
pad_token_id=pad_token_id,
|
| 147 |
-
bos_token_id=bos_token_id,
|
| 148 |
-
eos_token_id=eos_token_id,
|
| 149 |
-
tie_word_embeddings=tie_word_embeddings,
|
| 150 |
-
**kwargs,
|
| 151 |
-
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
generation_config.json
DELETED
|
@@ -1,7 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"_from_model_config": true,
|
| 3 |
-
"bos_token_id": 1,
|
| 4 |
-
"eos_token_id": 2,
|
| 5 |
-
"pad_token_id": 0,
|
| 6 |
-
"transformers_version": "4.38.2"
|
| 7 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
model-00001-of-00019.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:fae4638b4fc0299dcafe947f1cdf76bf7a69d4e9fb69c193f67ab89aa34c376c
|
| 3 |
-
size 4958457672
|
|
|
|
|
|
|
|
|
|
|
|
model-00002-of-00019.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:eac30089a1f3cd214a210a67d795a748ff00b12d73cddce7cfb7e8a04cacc2b1
|
| 3 |
-
size 4958655552
|
|
|
|
|
|
|
|
|
|
|
|
model-00003-of-00019.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:c6d74b4180040b1a209adbf3f695370f465bcaee39e3f11d51096c5d10e202fb
|
| 3 |
-
size 4732261544
|
|
|
|
|
|
|
|
|
|
|
|
model-00004-of-00019.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:a3058039cb929b9a468fd22d13255e595307e413144b421b793b81ccdab93d58
|
| 3 |
-
size 4958655592
|
|
|
|
|
|
|
|
|
|
|
|
model-00005-of-00019.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:28e64f1177eef0b0bc86593fb6dc4458a5b49d0d5ef5d4fafb66b567c2c7287b
|
| 3 |
-
size 4732261592
|
|
|
|
|
|
|
|
|
|
|
|
model-00006-of-00019.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:c56aef5249666b49c1343228ddbf1279c2ab4422cb98cb53b3e5b0387d27676e
|
| 3 |
-
size 4958655600
|
|
|
|
|
|
|
|
|
|
|
|
model-00007-of-00019.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:478526e8598450fd4231dfe84832fb9865eb62c231273349ecbb2593f4b7cf72
|
| 3 |
-
size 4732261592
|
|
|
|
|
|
|
|
|
|
|
|
model-00008-of-00019.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:0b4648e9c1e92a9d7344fb2e8d2dd4896a4489cc309b64f11f34110db374aa41
|
| 3 |
-
size 4958655600
|
|
|
|
|
|
|
|
|
|
|
|
model-00009-of-00019.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:d45d38b0ced5ef5f6470f47b4df1d5f74e6482a903bba04fd63f4441eaeb035c
|
| 3 |
-
size 4732261592
|
|
|
|
|
|
|
|
|
|
|
|
model-00010-of-00019.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:e2a53115b58fa9544a0b0e00696cda3ece771dd77edf5dc0e9c320fcfbf782f5
|
| 3 |
-
size 4958655600
|
|
|
|
|
|
|
|
|
|
|
|
model-00011-of-00019.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:f5e6b630792a33357c2993309479d172d79fedcbe9be540480e3ca582ff65a58
|
| 3 |
-
size 4732261592
|
|
|
|
|
|
|
|
|
|
|
|
model-00012-of-00019.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:96b2f0a0c985921a9c15b55f134594812e6e53f76bc91890421d32070d16f048
|
| 3 |
-
size 4958655600
|
|
|
|
|
|
|
|
|
|
|
|
model-00013-of-00019.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:96ed70e3f45808ee0ece374045275be2d0d1cee75bab3a3a863e461e5a66c1d7
|
| 3 |
-
size 4732261592
|
|
|
|
|
|
|
|
|
|
|
|
model-00014-of-00019.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:bdc9513b9dd70b7974e136f843a7613653b0aeca76ece604995d6d21e93720e8
|
| 3 |
-
size 4958655600
|
|
|
|
|
|
|
|
|
|
|
|
model-00015-of-00019.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:51379d67ac3b5525ce49eda9ee30edd2e98d27697b3d888e8b2c9e2d87080510
|
| 3 |
-
size 4732261592
|
|
|
|
|
|
|
|
|
|
|
|
model-00016-of-00019.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:bc38dde94a0ec295b7c57d725d3c048502aae2868de10ae96c8d623fb4606722
|
| 3 |
-
size 4958655600
|
|
|
|
|
|
|
|
|
|
|
|
model-00017-of-00019.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:d55f1848da09769f88fdc1f4933909a29a7d0686b830e4f711d903e46592cf42
|
| 3 |
-
size 4732261592
|
|
|
|
|
|
|
|
|
|
|
|
model-00018-of-00019.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:10a6ab4893eb048f142e8900e3e30bbd3c0aaeaa00704dad666c38b3ed0e63f5
|
| 3 |
-
size 4958655600
|
|
|
|
|
|
|
|
|
|
|
|
model-00019-of-00019.safetensors
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:3b054303f82fc96d8343b969c52174e7d3423fcf3694d0eda6c50ebb9471813c
|
| 3 |
-
size 2768858160
|
|
|
|
|
|
|
|
|
|
|
|
model.safetensors.index.json
DELETED
|
@@ -1,777 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"metadata": {
|
| 3 |
-
"total_size": 90213212160
|
| 4 |
-
},
|
| 5 |
-
"weight_map": {
|
| 6 |
-
"transformer.decoder_layer.0.moe.0.linear.weight": "model-00001-of-00019.safetensors",
|
| 7 |
-
"transformer.decoder_layer.0.moe.0.linear_1.weight": "model-00001-of-00019.safetensors",
|
| 8 |
-
"transformer.decoder_layer.0.moe.0.linear_v.weight": "model-00001-of-00019.safetensors",
|
| 9 |
-
"transformer.decoder_layer.0.multi_head_attention.key.weight": "model-00001-of-00019.safetensors",
|
| 10 |
-
"transformer.decoder_layer.0.multi_head_attention.linear.weight": "model-00001-of-00019.safetensors",
|
| 11 |
-
"transformer.decoder_layer.0.multi_head_attention.query.weight": "model-00001-of-00019.safetensors",
|
| 12 |
-
"transformer.decoder_layer.0.multi_head_attention.value.weight": "model-00001-of-00019.safetensors",
|
| 13 |
-
"transformer.decoder_layer.0.rms_norm.weight": "model-00001-of-00019.safetensors",
|
| 14 |
-
"transformer.decoder_layer.0.rms_norm_1.weight": "model-00001-of-00019.safetensors",
|
| 15 |
-
"transformer.decoder_layer.0.rms_norm_2.weight": "model-00001-of-00019.safetensors",
|
| 16 |
-
"transformer.decoder_layer.0.rms_norm_3.weight": "model-00001-of-00019.safetensors",
|
| 17 |
-
"transformer.decoder_layer.0.router.weight": "model-00001-of-00019.safetensors",
|
| 18 |
-
"transformer.decoder_layer.1.moe.0.linear.weight": "model-00001-of-00019.safetensors",
|
| 19 |
-
"transformer.decoder_layer.1.moe.0.linear_1.weight": "model-00001-of-00019.safetensors",
|
| 20 |
-
"transformer.decoder_layer.1.moe.0.linear_v.weight": "model-00001-of-00019.safetensors",
|
| 21 |
-
"transformer.decoder_layer.1.multi_head_attention.key.weight": "model-00001-of-00019.safetensors",
|
| 22 |
-
"transformer.decoder_layer.1.multi_head_attention.linear.weight": "model-00001-of-00019.safetensors",
|
| 23 |
-
"transformer.decoder_layer.1.multi_head_attention.query.weight": "model-00001-of-00019.safetensors",
|
| 24 |
-
"transformer.decoder_layer.1.multi_head_attention.value.weight": "model-00001-of-00019.safetensors",
|
| 25 |
-
"transformer.decoder_layer.1.rms_norm.weight": "model-00001-of-00019.safetensors",
|
| 26 |
-
"transformer.decoder_layer.1.rms_norm_1.weight": "model-00001-of-00019.safetensors",
|
| 27 |
-
"transformer.decoder_layer.1.rms_norm_2.weight": "model-00001-of-00019.safetensors",
|
| 28 |
-
"transformer.decoder_layer.1.rms_norm_3.weight": "model-00001-of-00019.safetensors",
|
| 29 |
-
"transformer.decoder_layer.1.router.weight": "model-00001-of-00019.safetensors",
|
| 30 |
-
"transformer.decoder_layer.10.moe.0.linear.weight": "model-00004-of-00019.safetensors",
|
| 31 |
-
"transformer.decoder_layer.10.moe.0.linear_1.weight": "model-00004-of-00019.safetensors",
|
| 32 |
-
"transformer.decoder_layer.10.moe.0.linear_v.weight": "model-00004-of-00019.safetensors",
|
| 33 |
-
"transformer.decoder_layer.10.multi_head_attention.key.weight": "model-00004-of-00019.safetensors",
|
| 34 |
-
"transformer.decoder_layer.10.multi_head_attention.linear.weight": "model-00004-of-00019.safetensors",
|
| 35 |
-
"transformer.decoder_layer.10.multi_head_attention.query.weight": "model-00004-of-00019.safetensors",
|
| 36 |
-
"transformer.decoder_layer.10.multi_head_attention.value.weight": "model-00004-of-00019.safetensors",
|
| 37 |
-
"transformer.decoder_layer.10.rms_norm.weight": "model-00004-of-00019.safetensors",
|
| 38 |
-
"transformer.decoder_layer.10.rms_norm_1.weight": "model-00004-of-00019.safetensors",
|
| 39 |
-
"transformer.decoder_layer.10.rms_norm_2.weight": "model-00004-of-00019.safetensors",
|
| 40 |
-
"transformer.decoder_layer.10.rms_norm_3.weight": "model-00004-of-00019.safetensors",
|
| 41 |
-
"transformer.decoder_layer.10.router.weight": "model-00004-of-00019.safetensors",
|
| 42 |
-
"transformer.decoder_layer.11.moe.0.linear.weight": "model-00004-of-00019.safetensors",
|
| 43 |
-
"transformer.decoder_layer.11.moe.0.linear_1.weight": "model-00004-of-00019.safetensors",
|
| 44 |
-
"transformer.decoder_layer.11.moe.0.linear_v.weight": "model-00004-of-00019.safetensors",
|
| 45 |
-
"transformer.decoder_layer.11.multi_head_attention.key.weight": "model-00004-of-00019.safetensors",
|
| 46 |
-
"transformer.decoder_layer.11.multi_head_attention.linear.weight": "model-00004-of-00019.safetensors",
|
| 47 |
-
"transformer.decoder_layer.11.multi_head_attention.query.weight": "model-00004-of-00019.safetensors",
|
| 48 |
-
"transformer.decoder_layer.11.multi_head_attention.value.weight": "model-00004-of-00019.safetensors",
|
| 49 |
-
"transformer.decoder_layer.11.rms_norm.weight": "model-00004-of-00019.safetensors",
|
| 50 |
-
"transformer.decoder_layer.11.rms_norm_1.weight": "model-00004-of-00019.safetensors",
|
| 51 |
-
"transformer.decoder_layer.11.rms_norm_2.weight": "model-00004-of-00019.safetensors",
|
| 52 |
-
"transformer.decoder_layer.11.rms_norm_3.weight": "model-00004-of-00019.safetensors",
|
| 53 |
-
"transformer.decoder_layer.11.router.weight": "model-00004-of-00019.safetensors",
|
| 54 |
-
"transformer.decoder_layer.12.moe.0.linear.weight": "model-00004-of-00019.safetensors",
|
| 55 |
-
"transformer.decoder_layer.12.moe.0.linear_1.weight": "model-00004-of-00019.safetensors",
|
| 56 |
-
"transformer.decoder_layer.12.moe.0.linear_v.weight": "model-00004-of-00019.safetensors",
|
| 57 |
-
"transformer.decoder_layer.12.multi_head_attention.key.weight": "model-00004-of-00019.safetensors",
|
| 58 |
-
"transformer.decoder_layer.12.multi_head_attention.linear.weight": "model-00004-of-00019.safetensors",
|
| 59 |
-
"transformer.decoder_layer.12.multi_head_attention.query.weight": "model-00004-of-00019.safetensors",
|
| 60 |
-
"transformer.decoder_layer.12.multi_head_attention.value.weight": "model-00004-of-00019.safetensors",
|
| 61 |
-
"transformer.decoder_layer.12.rms_norm.weight": "model-00004-of-00019.safetensors",
|
| 62 |
-
"transformer.decoder_layer.12.rms_norm_1.weight": "model-00004-of-00019.safetensors",
|
| 63 |
-
"transformer.decoder_layer.12.rms_norm_2.weight": "model-00004-of-00019.safetensors",
|
| 64 |
-
"transformer.decoder_layer.12.rms_norm_3.weight": "model-00004-of-00019.safetensors",
|
| 65 |
-
"transformer.decoder_layer.12.router.weight": "model-00004-of-00019.safetensors",
|
| 66 |
-
"transformer.decoder_layer.13.moe.0.linear.weight": "model-00005-of-00019.safetensors",
|
| 67 |
-
"transformer.decoder_layer.13.moe.0.linear_1.weight": "model-00005-of-00019.safetensors",
|
| 68 |
-
"transformer.decoder_layer.13.moe.0.linear_v.weight": "model-00005-of-00019.safetensors",
|
| 69 |
-
"transformer.decoder_layer.13.multi_head_attention.key.weight": "model-00005-of-00019.safetensors",
|
| 70 |
-
"transformer.decoder_layer.13.multi_head_attention.linear.weight": "model-00005-of-00019.safetensors",
|
| 71 |
-
"transformer.decoder_layer.13.multi_head_attention.query.weight": "model-00005-of-00019.safetensors",
|
| 72 |
-
"transformer.decoder_layer.13.multi_head_attention.value.weight": "model-00005-of-00019.safetensors",
|
| 73 |
-
"transformer.decoder_layer.13.rms_norm.weight": "model-00005-of-00019.safetensors",
|
| 74 |
-
"transformer.decoder_layer.13.rms_norm_1.weight": "model-00005-of-00019.safetensors",
|
| 75 |
-
"transformer.decoder_layer.13.rms_norm_2.weight": "model-00005-of-00019.safetensors",
|
| 76 |
-
"transformer.decoder_layer.13.rms_norm_3.weight": "model-00005-of-00019.safetensors",
|
| 77 |
-
"transformer.decoder_layer.13.router.weight": "model-00005-of-00019.safetensors",
|
| 78 |
-
"transformer.decoder_layer.14.moe.0.linear.weight": "model-00005-of-00019.safetensors",
|
| 79 |
-
"transformer.decoder_layer.14.moe.0.linear_1.weight": "model-00005-of-00019.safetensors",
|
| 80 |
-
"transformer.decoder_layer.14.moe.0.linear_v.weight": "model-00005-of-00019.safetensors",
|
| 81 |
-
"transformer.decoder_layer.14.multi_head_attention.key.weight": "model-00005-of-00019.safetensors",
|
| 82 |
-
"transformer.decoder_layer.14.multi_head_attention.linear.weight": "model-00005-of-00019.safetensors",
|
| 83 |
-
"transformer.decoder_layer.14.multi_head_attention.query.weight": "model-00005-of-00019.safetensors",
|
| 84 |
-
"transformer.decoder_layer.14.multi_head_attention.value.weight": "model-00005-of-00019.safetensors",
|
| 85 |
-
"transformer.decoder_layer.14.rms_norm.weight": "model-00005-of-00019.safetensors",
|
| 86 |
-
"transformer.decoder_layer.14.rms_norm_1.weight": "model-00005-of-00019.safetensors",
|
| 87 |
-
"transformer.decoder_layer.14.rms_norm_2.weight": "model-00005-of-00019.safetensors",
|
| 88 |
-
"transformer.decoder_layer.14.rms_norm_3.weight": "model-00005-of-00019.safetensors",
|
| 89 |
-
"transformer.decoder_layer.14.router.weight": "model-00005-of-00019.safetensors",
|
| 90 |
-
"transformer.decoder_layer.15.moe.0.linear.weight": "model-00005-of-00019.safetensors",
|
| 91 |
-
"transformer.decoder_layer.15.moe.0.linear_1.weight": "model-00005-of-00019.safetensors",
|
| 92 |
-
"transformer.decoder_layer.15.moe.0.linear_v.weight": "model-00005-of-00019.safetensors",
|
| 93 |
-
"transformer.decoder_layer.15.multi_head_attention.key.weight": "model-00005-of-00019.safetensors",
|
| 94 |
-
"transformer.decoder_layer.15.multi_head_attention.linear.weight": "model-00005-of-00019.safetensors",
|
| 95 |
-
"transformer.decoder_layer.15.multi_head_attention.query.weight": "model-00005-of-00019.safetensors",
|
| 96 |
-
"transformer.decoder_layer.15.multi_head_attention.value.weight": "model-00005-of-00019.safetensors",
|
| 97 |
-
"transformer.decoder_layer.15.rms_norm.weight": "model-00005-of-00019.safetensors",
|
| 98 |
-
"transformer.decoder_layer.15.rms_norm_1.weight": "model-00005-of-00019.safetensors",
|
| 99 |
-
"transformer.decoder_layer.15.rms_norm_2.weight": "model-00005-of-00019.safetensors",
|
| 100 |
-
"transformer.decoder_layer.15.rms_norm_3.weight": "model-00005-of-00019.safetensors",
|
| 101 |
-
"transformer.decoder_layer.15.router.weight": "model-00005-of-00019.safetensors",
|
| 102 |
-
"transformer.decoder_layer.16.moe.0.linear.weight": "model-00006-of-00019.safetensors",
|
| 103 |
-
"transformer.decoder_layer.16.moe.0.linear_1.weight": "model-00006-of-00019.safetensors",
|
| 104 |
-
"transformer.decoder_layer.16.moe.0.linear_v.weight": "model-00005-of-00019.safetensors",
|
| 105 |
-
"transformer.decoder_layer.16.multi_head_attention.key.weight": "model-00005-of-00019.safetensors",
|
| 106 |
-
"transformer.decoder_layer.16.multi_head_attention.linear.weight": "model-00005-of-00019.safetensors",
|
| 107 |
-
"transformer.decoder_layer.16.multi_head_attention.query.weight": "model-00005-of-00019.safetensors",
|
| 108 |
-
"transformer.decoder_layer.16.multi_head_attention.value.weight": "model-00005-of-00019.safetensors",
|
| 109 |
-
"transformer.decoder_layer.16.rms_norm.weight": "model-00006-of-00019.safetensors",
|
| 110 |
-
"transformer.decoder_layer.16.rms_norm_1.weight": "model-00006-of-00019.safetensors",
|
| 111 |
-
"transformer.decoder_layer.16.rms_norm_2.weight": "model-00006-of-00019.safetensors",
|
| 112 |
-
"transformer.decoder_layer.16.rms_norm_3.weight": "model-00006-of-00019.safetensors",
|
| 113 |
-
"transformer.decoder_layer.16.router.weight": "model-00005-of-00019.safetensors",
|
| 114 |
-
"transformer.decoder_layer.17.moe.0.linear.weight": "model-00006-of-00019.safetensors",
|
| 115 |
-
"transformer.decoder_layer.17.moe.0.linear_1.weight": "model-00006-of-00019.safetensors",
|
| 116 |
-
"transformer.decoder_layer.17.moe.0.linear_v.weight": "model-00006-of-00019.safetensors",
|
| 117 |
-
"transformer.decoder_layer.17.multi_head_attention.key.weight": "model-00006-of-00019.safetensors",
|
| 118 |
-
"transformer.decoder_layer.17.multi_head_attention.linear.weight": "model-00006-of-00019.safetensors",
|
| 119 |
-
"transformer.decoder_layer.17.multi_head_attention.query.weight": "model-00006-of-00019.safetensors",
|
| 120 |
-
"transformer.decoder_layer.17.multi_head_attention.value.weight": "model-00006-of-00019.safetensors",
|
| 121 |
-
"transformer.decoder_layer.17.rms_norm.weight": "model-00006-of-00019.safetensors",
|
| 122 |
-
"transformer.decoder_layer.17.rms_norm_1.weight": "model-00006-of-00019.safetensors",
|
| 123 |
-
"transformer.decoder_layer.17.rms_norm_2.weight": "model-00006-of-00019.safetensors",
|
| 124 |
-
"transformer.decoder_layer.17.rms_norm_3.weight": "model-00006-of-00019.safetensors",
|
| 125 |
-
"transformer.decoder_layer.17.router.weight": "model-00006-of-00019.safetensors",
|
| 126 |
-
"transformer.decoder_layer.18.moe.0.linear.weight": "model-00006-of-00019.safetensors",
|
| 127 |
-
"transformer.decoder_layer.18.moe.0.linear_1.weight": "model-00006-of-00019.safetensors",
|
| 128 |
-
"transformer.decoder_layer.18.moe.0.linear_v.weight": "model-00006-of-00019.safetensors",
|
| 129 |
-
"transformer.decoder_layer.18.multi_head_attention.key.weight": "model-00006-of-00019.safetensors",
|
| 130 |
-
"transformer.decoder_layer.18.multi_head_attention.linear.weight": "model-00006-of-00019.safetensors",
|
| 131 |
-
"transformer.decoder_layer.18.multi_head_attention.query.weight": "model-00006-of-00019.safetensors",
|
| 132 |
-
"transformer.decoder_layer.18.multi_head_attention.value.weight": "model-00006-of-00019.safetensors",
|
| 133 |
-
"transformer.decoder_layer.18.rms_norm.weight": "model-00006-of-00019.safetensors",
|
| 134 |
-
"transformer.decoder_layer.18.rms_norm_1.weight": "model-00006-of-00019.safetensors",
|
| 135 |
-
"transformer.decoder_layer.18.rms_norm_2.weight": "model-00006-of-00019.safetensors",
|
| 136 |
-
"transformer.decoder_layer.18.rms_norm_3.weight": "model-00006-of-00019.safetensors",
|
| 137 |
-
"transformer.decoder_layer.18.router.weight": "model-00006-of-00019.safetensors",
|
| 138 |
-
"transformer.decoder_layer.19.moe.0.linear.weight": "model-00006-of-00019.safetensors",
|
| 139 |
-
"transformer.decoder_layer.19.moe.0.linear_1.weight": "model-00006-of-00019.safetensors",
|
| 140 |
-
"transformer.decoder_layer.19.moe.0.linear_v.weight": "model-00006-of-00019.safetensors",
|
| 141 |
-
"transformer.decoder_layer.19.multi_head_attention.key.weight": "model-00006-of-00019.safetensors",
|
| 142 |
-
"transformer.decoder_layer.19.multi_head_attention.linear.weight": "model-00006-of-00019.safetensors",
|
| 143 |
-
"transformer.decoder_layer.19.multi_head_attention.query.weight": "model-00006-of-00019.safetensors",
|
| 144 |
-
"transformer.decoder_layer.19.multi_head_attention.value.weight": "model-00006-of-00019.safetensors",
|
| 145 |
-
"transformer.decoder_layer.19.rms_norm.weight": "model-00006-of-00019.safetensors",
|
| 146 |
-
"transformer.decoder_layer.19.rms_norm_1.weight": "model-00006-of-00019.safetensors",
|
| 147 |
-
"transformer.decoder_layer.19.rms_norm_2.weight": "model-00006-of-00019.safetensors",
|
| 148 |
-
"transformer.decoder_layer.19.rms_norm_3.weight": "model-00006-of-00019.safetensors",
|
| 149 |
-
"transformer.decoder_layer.19.router.weight": "model-00006-of-00019.safetensors",
|
| 150 |
-
"transformer.decoder_layer.2.moe.0.linear.weight": "model-00002-of-00019.safetensors",
|
| 151 |
-
"transformer.decoder_layer.2.moe.0.linear_1.weight": "model-00002-of-00019.safetensors",
|
| 152 |
-
"transformer.decoder_layer.2.moe.0.linear_v.weight": "model-00001-of-00019.safetensors",
|
| 153 |
-
"transformer.decoder_layer.2.multi_head_attention.key.weight": "model-00001-of-00019.safetensors",
|
| 154 |
-
"transformer.decoder_layer.2.multi_head_attention.linear.weight": "model-00001-of-00019.safetensors",
|
| 155 |
-
"transformer.decoder_layer.2.multi_head_attention.query.weight": "model-00001-of-00019.safetensors",
|
| 156 |
-
"transformer.decoder_layer.2.multi_head_attention.value.weight": "model-00001-of-00019.safetensors",
|
| 157 |
-
"transformer.decoder_layer.2.rms_norm.weight": "model-00002-of-00019.safetensors",
|
| 158 |
-
"transformer.decoder_layer.2.rms_norm_1.weight": "model-00002-of-00019.safetensors",
|
| 159 |
-
"transformer.decoder_layer.2.rms_norm_2.weight": "model-00002-of-00019.safetensors",
|
| 160 |
-
"transformer.decoder_layer.2.rms_norm_3.weight": "model-00002-of-00019.safetensors",
|
| 161 |
-
"transformer.decoder_layer.2.router.weight": "model-00001-of-00019.safetensors",
|
| 162 |
-
"transformer.decoder_layer.20.moe.0.linear.weight": "model-00007-of-00019.safetensors",
|
| 163 |
-
"transformer.decoder_layer.20.moe.0.linear_1.weight": "model-00007-of-00019.safetensors",
|
| 164 |
-
"transformer.decoder_layer.20.moe.0.linear_v.weight": "model-00007-of-00019.safetensors",
|
| 165 |
-
"transformer.decoder_layer.20.multi_head_attention.key.weight": "model-00007-of-00019.safetensors",
|
| 166 |
-
"transformer.decoder_layer.20.multi_head_attention.linear.weight": "model-00007-of-00019.safetensors",
|
| 167 |
-
"transformer.decoder_layer.20.multi_head_attention.query.weight": "model-00007-of-00019.safetensors",
|
| 168 |
-
"transformer.decoder_layer.20.multi_head_attention.value.weight": "model-00007-of-00019.safetensors",
|
| 169 |
-
"transformer.decoder_layer.20.rms_norm.weight": "model-00007-of-00019.safetensors",
|
| 170 |
-
"transformer.decoder_layer.20.rms_norm_1.weight": "model-00007-of-00019.safetensors",
|
| 171 |
-
"transformer.decoder_layer.20.rms_norm_2.weight": "model-00007-of-00019.safetensors",
|
| 172 |
-
"transformer.decoder_layer.20.rms_norm_3.weight": "model-00007-of-00019.safetensors",
|
| 173 |
-
"transformer.decoder_layer.20.router.weight": "model-00007-of-00019.safetensors",
|
| 174 |
-
"transformer.decoder_layer.21.moe.0.linear.weight": "model-00007-of-00019.safetensors",
|
| 175 |
-
"transformer.decoder_layer.21.moe.0.linear_1.weight": "model-00007-of-00019.safetensors",
|
| 176 |
-
"transformer.decoder_layer.21.moe.0.linear_v.weight": "model-00007-of-00019.safetensors",
|
| 177 |
-
"transformer.decoder_layer.21.multi_head_attention.key.weight": "model-00007-of-00019.safetensors",
|
| 178 |
-
"transformer.decoder_layer.21.multi_head_attention.linear.weight": "model-00007-of-00019.safetensors",
|
| 179 |
-
"transformer.decoder_layer.21.multi_head_attention.query.weight": "model-00007-of-00019.safetensors",
|
| 180 |
-
"transformer.decoder_layer.21.multi_head_attention.value.weight": "model-00007-of-00019.safetensors",
|
| 181 |
-
"transformer.decoder_layer.21.rms_norm.weight": "model-00007-of-00019.safetensors",
|
| 182 |
-
"transformer.decoder_layer.21.rms_norm_1.weight": "model-00007-of-00019.safetensors",
|
| 183 |
-
"transformer.decoder_layer.21.rms_norm_2.weight": "model-00007-of-00019.safetensors",
|
| 184 |
-
"transformer.decoder_layer.21.rms_norm_3.weight": "model-00007-of-00019.safetensors",
|
| 185 |
-
"transformer.decoder_layer.21.router.weight": "model-00007-of-00019.safetensors",
|
| 186 |
-
"transformer.decoder_layer.22.moe.0.linear.weight": "model-00007-of-00019.safetensors",
|
| 187 |
-
"transformer.decoder_layer.22.moe.0.linear_1.weight": "model-00007-of-00019.safetensors",
|
| 188 |
-
"transformer.decoder_layer.22.moe.0.linear_v.weight": "model-00007-of-00019.safetensors",
|
| 189 |
-
"transformer.decoder_layer.22.multi_head_attention.key.weight": "model-00007-of-00019.safetensors",
|
| 190 |
-
"transformer.decoder_layer.22.multi_head_attention.linear.weight": "model-00007-of-00019.safetensors",
|
| 191 |
-
"transformer.decoder_layer.22.multi_head_attention.query.weight": "model-00007-of-00019.safetensors",
|
| 192 |
-
"transformer.decoder_layer.22.multi_head_attention.value.weight": "model-00007-of-00019.safetensors",
|
| 193 |
-
"transformer.decoder_layer.22.rms_norm.weight": "model-00007-of-00019.safetensors",
|
| 194 |
-
"transformer.decoder_layer.22.rms_norm_1.weight": "model-00007-of-00019.safetensors",
|
| 195 |
-
"transformer.decoder_layer.22.rms_norm_2.weight": "model-00007-of-00019.safetensors",
|
| 196 |
-
"transformer.decoder_layer.22.rms_norm_3.weight": "model-00007-of-00019.safetensors",
|
| 197 |
-
"transformer.decoder_layer.22.router.weight": "model-00007-of-00019.safetensors",
|
| 198 |
-
"transformer.decoder_layer.23.moe.0.linear.weight": "model-00008-of-00019.safetensors",
|
| 199 |
-
"transformer.decoder_layer.23.moe.0.linear_1.weight": "model-00008-of-00019.safetensors",
|
| 200 |
-
"transformer.decoder_layer.23.moe.0.linear_v.weight": "model-00007-of-00019.safetensors",
|
| 201 |
-
"transformer.decoder_layer.23.multi_head_attention.key.weight": "model-00007-of-00019.safetensors",
|
| 202 |
-
"transformer.decoder_layer.23.multi_head_attention.linear.weight": "model-00007-of-00019.safetensors",
|
| 203 |
-
"transformer.decoder_layer.23.multi_head_attention.query.weight": "model-00007-of-00019.safetensors",
|
| 204 |
-
"transformer.decoder_layer.23.multi_head_attention.value.weight": "model-00007-of-00019.safetensors",
|
| 205 |
-
"transformer.decoder_layer.23.rms_norm.weight": "model-00008-of-00019.safetensors",
|
| 206 |
-
"transformer.decoder_layer.23.rms_norm_1.weight": "model-00008-of-00019.safetensors",
|
| 207 |
-
"transformer.decoder_layer.23.rms_norm_2.weight": "model-00008-of-00019.safetensors",
|
| 208 |
-
"transformer.decoder_layer.23.rms_norm_3.weight": "model-00008-of-00019.safetensors",
|
| 209 |
-
"transformer.decoder_layer.23.router.weight": "model-00007-of-00019.safetensors",
|
| 210 |
-
"transformer.decoder_layer.24.moe.0.linear.weight": "model-00008-of-00019.safetensors",
|
| 211 |
-
"transformer.decoder_layer.24.moe.0.linear_1.weight": "model-00008-of-00019.safetensors",
|
| 212 |
-
"transformer.decoder_layer.24.moe.0.linear_v.weight": "model-00008-of-00019.safetensors",
|
| 213 |
-
"transformer.decoder_layer.24.multi_head_attention.key.weight": "model-00008-of-00019.safetensors",
|
| 214 |
-
"transformer.decoder_layer.24.multi_head_attention.linear.weight": "model-00008-of-00019.safetensors",
|
| 215 |
-
"transformer.decoder_layer.24.multi_head_attention.query.weight": "model-00008-of-00019.safetensors",
|
| 216 |
-
"transformer.decoder_layer.24.multi_head_attention.value.weight": "model-00008-of-00019.safetensors",
|
| 217 |
-
"transformer.decoder_layer.24.rms_norm.weight": "model-00008-of-00019.safetensors",
|
| 218 |
-
"transformer.decoder_layer.24.rms_norm_1.weight": "model-00008-of-00019.safetensors",
|
| 219 |
-
"transformer.decoder_layer.24.rms_norm_2.weight": "model-00008-of-00019.safetensors",
|
| 220 |
-
"transformer.decoder_layer.24.rms_norm_3.weight": "model-00008-of-00019.safetensors",
|
| 221 |
-
"transformer.decoder_layer.24.router.weight": "model-00008-of-00019.safetensors",
|
| 222 |
-
"transformer.decoder_layer.25.moe.0.linear.weight": "model-00008-of-00019.safetensors",
|
| 223 |
-
"transformer.decoder_layer.25.moe.0.linear_1.weight": "model-00008-of-00019.safetensors",
|
| 224 |
-
"transformer.decoder_layer.25.moe.0.linear_v.weight": "model-00008-of-00019.safetensors",
|
| 225 |
-
"transformer.decoder_layer.25.multi_head_attention.key.weight": "model-00008-of-00019.safetensors",
|
| 226 |
-
"transformer.decoder_layer.25.multi_head_attention.linear.weight": "model-00008-of-00019.safetensors",
|
| 227 |
-
"transformer.decoder_layer.25.multi_head_attention.query.weight": "model-00008-of-00019.safetensors",
|
| 228 |
-
"transformer.decoder_layer.25.multi_head_attention.value.weight": "model-00008-of-00019.safetensors",
|
| 229 |
-
"transformer.decoder_layer.25.rms_norm.weight": "model-00008-of-00019.safetensors",
|
| 230 |
-
"transformer.decoder_layer.25.rms_norm_1.weight": "model-00008-of-00019.safetensors",
|
| 231 |
-
"transformer.decoder_layer.25.rms_norm_2.weight": "model-00008-of-00019.safetensors",
|
| 232 |
-
"transformer.decoder_layer.25.rms_norm_3.weight": "model-00008-of-00019.safetensors",
|
| 233 |
-
"transformer.decoder_layer.25.router.weight": "model-00008-of-00019.safetensors",
|
| 234 |
-
"transformer.decoder_layer.26.moe.0.linear.weight": "model-00008-of-00019.safetensors",
|
| 235 |
-
"transformer.decoder_layer.26.moe.0.linear_1.weight": "model-00008-of-00019.safetensors",
|
| 236 |
-
"transformer.decoder_layer.26.moe.0.linear_v.weight": "model-00008-of-00019.safetensors",
|
| 237 |
-
"transformer.decoder_layer.26.multi_head_attention.key.weight": "model-00008-of-00019.safetensors",
|
| 238 |
-
"transformer.decoder_layer.26.multi_head_attention.linear.weight": "model-00008-of-00019.safetensors",
|
| 239 |
-
"transformer.decoder_layer.26.multi_head_attention.query.weight": "model-00008-of-00019.safetensors",
|
| 240 |
-
"transformer.decoder_layer.26.multi_head_attention.value.weight": "model-00008-of-00019.safetensors",
|
| 241 |
-
"transformer.decoder_layer.26.rms_norm.weight": "model-00008-of-00019.safetensors",
|
| 242 |
-
"transformer.decoder_layer.26.rms_norm_1.weight": "model-00008-of-00019.safetensors",
|
| 243 |
-
"transformer.decoder_layer.26.rms_norm_2.weight": "model-00008-of-00019.safetensors",
|
| 244 |
-
"transformer.decoder_layer.26.rms_norm_3.weight": "model-00008-of-00019.safetensors",
|
| 245 |
-
"transformer.decoder_layer.26.router.weight": "model-00008-of-00019.safetensors",
|
| 246 |
-
"transformer.decoder_layer.27.moe.0.linear.weight": "model-00009-of-00019.safetensors",
|
| 247 |
-
"transformer.decoder_layer.27.moe.0.linear_1.weight": "model-00009-of-00019.safetensors",
|
| 248 |
-
"transformer.decoder_layer.27.moe.0.linear_v.weight": "model-00009-of-00019.safetensors",
|
| 249 |
-
"transformer.decoder_layer.27.multi_head_attention.key.weight": "model-00009-of-00019.safetensors",
|
| 250 |
-
"transformer.decoder_layer.27.multi_head_attention.linear.weight": "model-00009-of-00019.safetensors",
|
| 251 |
-
"transformer.decoder_layer.27.multi_head_attention.query.weight": "model-00009-of-00019.safetensors",
|
| 252 |
-
"transformer.decoder_layer.27.multi_head_attention.value.weight": "model-00009-of-00019.safetensors",
|
| 253 |
-
"transformer.decoder_layer.27.rms_norm.weight": "model-00009-of-00019.safetensors",
|
| 254 |
-
"transformer.decoder_layer.27.rms_norm_1.weight": "model-00009-of-00019.safetensors",
|
| 255 |
-
"transformer.decoder_layer.27.rms_norm_2.weight": "model-00009-of-00019.safetensors",
|
| 256 |
-
"transformer.decoder_layer.27.rms_norm_3.weight": "model-00009-of-00019.safetensors",
|
| 257 |
-
"transformer.decoder_layer.27.router.weight": "model-00009-of-00019.safetensors",
|
| 258 |
-
"transformer.decoder_layer.28.moe.0.linear.weight": "model-00009-of-00019.safetensors",
|
| 259 |
-
"transformer.decoder_layer.28.moe.0.linear_1.weight": "model-00009-of-00019.safetensors",
|
| 260 |
-
"transformer.decoder_layer.28.moe.0.linear_v.weight": "model-00009-of-00019.safetensors",
|
| 261 |
-
"transformer.decoder_layer.28.multi_head_attention.key.weight": "model-00009-of-00019.safetensors",
|
| 262 |
-
"transformer.decoder_layer.28.multi_head_attention.linear.weight": "model-00009-of-00019.safetensors",
|
| 263 |
-
"transformer.decoder_layer.28.multi_head_attention.query.weight": "model-00009-of-00019.safetensors",
|
| 264 |
-
"transformer.decoder_layer.28.multi_head_attention.value.weight": "model-00009-of-00019.safetensors",
|
| 265 |
-
"transformer.decoder_layer.28.rms_norm.weight": "model-00009-of-00019.safetensors",
|
| 266 |
-
"transformer.decoder_layer.28.rms_norm_1.weight": "model-00009-of-00019.safetensors",
|
| 267 |
-
"transformer.decoder_layer.28.rms_norm_2.weight": "model-00009-of-00019.safetensors",
|
| 268 |
-
"transformer.decoder_layer.28.rms_norm_3.weight": "model-00009-of-00019.safetensors",
|
| 269 |
-
"transformer.decoder_layer.28.router.weight": "model-00009-of-00019.safetensors",
|
| 270 |
-
"transformer.decoder_layer.29.moe.0.linear.weight": "model-00009-of-00019.safetensors",
|
| 271 |
-
"transformer.decoder_layer.29.moe.0.linear_1.weight": "model-00009-of-00019.safetensors",
|
| 272 |
-
"transformer.decoder_layer.29.moe.0.linear_v.weight": "model-00009-of-00019.safetensors",
|
| 273 |
-
"transformer.decoder_layer.29.multi_head_attention.key.weight": "model-00009-of-00019.safetensors",
|
| 274 |
-
"transformer.decoder_layer.29.multi_head_attention.linear.weight": "model-00009-of-00019.safetensors",
|
| 275 |
-
"transformer.decoder_layer.29.multi_head_attention.query.weight": "model-00009-of-00019.safetensors",
|
| 276 |
-
"transformer.decoder_layer.29.multi_head_attention.value.weight": "model-00009-of-00019.safetensors",
|
| 277 |
-
"transformer.decoder_layer.29.rms_norm.weight": "model-00009-of-00019.safetensors",
|
| 278 |
-
"transformer.decoder_layer.29.rms_norm_1.weight": "model-00009-of-00019.safetensors",
|
| 279 |
-
"transformer.decoder_layer.29.rms_norm_2.weight": "model-00009-of-00019.safetensors",
|
| 280 |
-
"transformer.decoder_layer.29.rms_norm_3.weight": "model-00009-of-00019.safetensors",
|
| 281 |
-
"transformer.decoder_layer.29.router.weight": "model-00009-of-00019.safetensors",
|
| 282 |
-
"transformer.decoder_layer.3.moe.0.linear.weight": "model-00002-of-00019.safetensors",
|
| 283 |
-
"transformer.decoder_layer.3.moe.0.linear_1.weight": "model-00002-of-00019.safetensors",
|
| 284 |
-
"transformer.decoder_layer.3.moe.0.linear_v.weight": "model-00002-of-00019.safetensors",
|
| 285 |
-
"transformer.decoder_layer.3.multi_head_attention.key.weight": "model-00002-of-00019.safetensors",
|
| 286 |
-
"transformer.decoder_layer.3.multi_head_attention.linear.weight": "model-00002-of-00019.safetensors",
|
| 287 |
-
"transformer.decoder_layer.3.multi_head_attention.query.weight": "model-00002-of-00019.safetensors",
|
| 288 |
-
"transformer.decoder_layer.3.multi_head_attention.value.weight": "model-00002-of-00019.safetensors",
|
| 289 |
-
"transformer.decoder_layer.3.rms_norm.weight": "model-00002-of-00019.safetensors",
|
| 290 |
-
"transformer.decoder_layer.3.rms_norm_1.weight": "model-00002-of-00019.safetensors",
|
| 291 |
-
"transformer.decoder_layer.3.rms_norm_2.weight": "model-00002-of-00019.safetensors",
|
| 292 |
-
"transformer.decoder_layer.3.rms_norm_3.weight": "model-00002-of-00019.safetensors",
|
| 293 |
-
"transformer.decoder_layer.3.router.weight": "model-00002-of-00019.safetensors",
|
| 294 |
-
"transformer.decoder_layer.30.moe.0.linear.weight": "model-00010-of-00019.safetensors",
|
| 295 |
-
"transformer.decoder_layer.30.moe.0.linear_1.weight": "model-00010-of-00019.safetensors",
|
| 296 |
-
"transformer.decoder_layer.30.moe.0.linear_v.weight": "model-00009-of-00019.safetensors",
|
| 297 |
-
"transformer.decoder_layer.30.multi_head_attention.key.weight": "model-00009-of-00019.safetensors",
|
| 298 |
-
"transformer.decoder_layer.30.multi_head_attention.linear.weight": "model-00009-of-00019.safetensors",
|
| 299 |
-
"transformer.decoder_layer.30.multi_head_attention.query.weight": "model-00009-of-00019.safetensors",
|
| 300 |
-
"transformer.decoder_layer.30.multi_head_attention.value.weight": "model-00009-of-00019.safetensors",
|
| 301 |
-
"transformer.decoder_layer.30.rms_norm.weight": "model-00010-of-00019.safetensors",
|
| 302 |
-
"transformer.decoder_layer.30.rms_norm_1.weight": "model-00010-of-00019.safetensors",
|
| 303 |
-
"transformer.decoder_layer.30.rms_norm_2.weight": "model-00010-of-00019.safetensors",
|
| 304 |
-
"transformer.decoder_layer.30.rms_norm_3.weight": "model-00010-of-00019.safetensors",
|
| 305 |
-
"transformer.decoder_layer.30.router.weight": "model-00009-of-00019.safetensors",
|
| 306 |
-
"transformer.decoder_layer.31.moe.0.linear.weight": "model-00010-of-00019.safetensors",
|
| 307 |
-
"transformer.decoder_layer.31.moe.0.linear_1.weight": "model-00010-of-00019.safetensors",
|
| 308 |
-
"transformer.decoder_layer.31.moe.0.linear_v.weight": "model-00010-of-00019.safetensors",
|
| 309 |
-
"transformer.decoder_layer.31.multi_head_attention.key.weight": "model-00010-of-00019.safetensors",
|
| 310 |
-
"transformer.decoder_layer.31.multi_head_attention.linear.weight": "model-00010-of-00019.safetensors",
|
| 311 |
-
"transformer.decoder_layer.31.multi_head_attention.query.weight": "model-00010-of-00019.safetensors",
|
| 312 |
-
"transformer.decoder_layer.31.multi_head_attention.value.weight": "model-00010-of-00019.safetensors",
|
| 313 |
-
"transformer.decoder_layer.31.rms_norm.weight": "model-00010-of-00019.safetensors",
|
| 314 |
-
"transformer.decoder_layer.31.rms_norm_1.weight": "model-00010-of-00019.safetensors",
|
| 315 |
-
"transformer.decoder_layer.31.rms_norm_2.weight": "model-00010-of-00019.safetensors",
|
| 316 |
-
"transformer.decoder_layer.31.rms_norm_3.weight": "model-00010-of-00019.safetensors",
|
| 317 |
-
"transformer.decoder_layer.31.router.weight": "model-00010-of-00019.safetensors",
|
| 318 |
-
"transformer.decoder_layer.32.moe.0.linear.weight": "model-00010-of-00019.safetensors",
|
| 319 |
-
"transformer.decoder_layer.32.moe.0.linear_1.weight": "model-00010-of-00019.safetensors",
|
| 320 |
-
"transformer.decoder_layer.32.moe.0.linear_v.weight": "model-00010-of-00019.safetensors",
|
| 321 |
-
"transformer.decoder_layer.32.multi_head_attention.key.weight": "model-00010-of-00019.safetensors",
|
| 322 |
-
"transformer.decoder_layer.32.multi_head_attention.linear.weight": "model-00010-of-00019.safetensors",
|
| 323 |
-
"transformer.decoder_layer.32.multi_head_attention.query.weight": "model-00010-of-00019.safetensors",
|
| 324 |
-
"transformer.decoder_layer.32.multi_head_attention.value.weight": "model-00010-of-00019.safetensors",
|
| 325 |
-
"transformer.decoder_layer.32.rms_norm.weight": "model-00010-of-00019.safetensors",
|
| 326 |
-
"transformer.decoder_layer.32.rms_norm_1.weight": "model-00010-of-00019.safetensors",
|
| 327 |
-
"transformer.decoder_layer.32.rms_norm_2.weight": "model-00010-of-00019.safetensors",
|
| 328 |
-
"transformer.decoder_layer.32.rms_norm_3.weight": "model-00010-of-00019.safetensors",
|
| 329 |
-
"transformer.decoder_layer.32.router.weight": "model-00010-of-00019.safetensors",
|
| 330 |
-
"transformer.decoder_layer.33.moe.0.linear.weight": "model-00010-of-00019.safetensors",
|
| 331 |
-
"transformer.decoder_layer.33.moe.0.linear_1.weight": "model-00010-of-00019.safetensors",
|
| 332 |
-
"transformer.decoder_layer.33.moe.0.linear_v.weight": "model-00010-of-00019.safetensors",
|
| 333 |
-
"transformer.decoder_layer.33.multi_head_attention.key.weight": "model-00010-of-00019.safetensors",
|
| 334 |
-
"transformer.decoder_layer.33.multi_head_attention.linear.weight": "model-00010-of-00019.safetensors",
|
| 335 |
-
"transformer.decoder_layer.33.multi_head_attention.query.weight": "model-00010-of-00019.safetensors",
|
| 336 |
-
"transformer.decoder_layer.33.multi_head_attention.value.weight": "model-00010-of-00019.safetensors",
|
| 337 |
-
"transformer.decoder_layer.33.rms_norm.weight": "model-00010-of-00019.safetensors",
|
| 338 |
-
"transformer.decoder_layer.33.rms_norm_1.weight": "model-00010-of-00019.safetensors",
|
| 339 |
-
"transformer.decoder_layer.33.rms_norm_2.weight": "model-00010-of-00019.safetensors",
|
| 340 |
-
"transformer.decoder_layer.33.rms_norm_3.weight": "model-00010-of-00019.safetensors",
|
| 341 |
-
"transformer.decoder_layer.33.router.weight": "model-00010-of-00019.safetensors",
|
| 342 |
-
"transformer.decoder_layer.34.moe.0.linear.weight": "model-00011-of-00019.safetensors",
|
| 343 |
-
"transformer.decoder_layer.34.moe.0.linear_1.weight": "model-00011-of-00019.safetensors",
|
| 344 |
-
"transformer.decoder_layer.34.moe.0.linear_v.weight": "model-00011-of-00019.safetensors",
|
| 345 |
-
"transformer.decoder_layer.34.multi_head_attention.key.weight": "model-00011-of-00019.safetensors",
|
| 346 |
-
"transformer.decoder_layer.34.multi_head_attention.linear.weight": "model-00011-of-00019.safetensors",
|
| 347 |
-
"transformer.decoder_layer.34.multi_head_attention.query.weight": "model-00011-of-00019.safetensors",
|
| 348 |
-
"transformer.decoder_layer.34.multi_head_attention.value.weight": "model-00011-of-00019.safetensors",
|
| 349 |
-
"transformer.decoder_layer.34.rms_norm.weight": "model-00011-of-00019.safetensors",
|
| 350 |
-
"transformer.decoder_layer.34.rms_norm_1.weight": "model-00011-of-00019.safetensors",
|
| 351 |
-
"transformer.decoder_layer.34.rms_norm_2.weight": "model-00011-of-00019.safetensors",
|
| 352 |
-
"transformer.decoder_layer.34.rms_norm_3.weight": "model-00011-of-00019.safetensors",
|
| 353 |
-
"transformer.decoder_layer.34.router.weight": "model-00011-of-00019.safetensors",
|
| 354 |
-
"transformer.decoder_layer.35.moe.0.linear.weight": "model-00011-of-00019.safetensors",
|
| 355 |
-
"transformer.decoder_layer.35.moe.0.linear_1.weight": "model-00011-of-00019.safetensors",
|
| 356 |
-
"transformer.decoder_layer.35.moe.0.linear_v.weight": "model-00011-of-00019.safetensors",
|
| 357 |
-
"transformer.decoder_layer.35.multi_head_attention.key.weight": "model-00011-of-00019.safetensors",
|
| 358 |
-
"transformer.decoder_layer.35.multi_head_attention.linear.weight": "model-00011-of-00019.safetensors",
|
| 359 |
-
"transformer.decoder_layer.35.multi_head_attention.query.weight": "model-00011-of-00019.safetensors",
|
| 360 |
-
"transformer.decoder_layer.35.multi_head_attention.value.weight": "model-00011-of-00019.safetensors",
|
| 361 |
-
"transformer.decoder_layer.35.rms_norm.weight": "model-00011-of-00019.safetensors",
|
| 362 |
-
"transformer.decoder_layer.35.rms_norm_1.weight": "model-00011-of-00019.safetensors",
|
| 363 |
-
"transformer.decoder_layer.35.rms_norm_2.weight": "model-00011-of-00019.safetensors",
|
| 364 |
-
"transformer.decoder_layer.35.rms_norm_3.weight": "model-00011-of-00019.safetensors",
|
| 365 |
-
"transformer.decoder_layer.35.router.weight": "model-00011-of-00019.safetensors",
|
| 366 |
-
"transformer.decoder_layer.36.moe.0.linear.weight": "model-00011-of-00019.safetensors",
|
| 367 |
-
"transformer.decoder_layer.36.moe.0.linear_1.weight": "model-00011-of-00019.safetensors",
|
| 368 |
-
"transformer.decoder_layer.36.moe.0.linear_v.weight": "model-00011-of-00019.safetensors",
|
| 369 |
-
"transformer.decoder_layer.36.multi_head_attention.key.weight": "model-00011-of-00019.safetensors",
|
| 370 |
-
"transformer.decoder_layer.36.multi_head_attention.linear.weight": "model-00011-of-00019.safetensors",
|
| 371 |
-
"transformer.decoder_layer.36.multi_head_attention.query.weight": "model-00011-of-00019.safetensors",
|
| 372 |
-
"transformer.decoder_layer.36.multi_head_attention.value.weight": "model-00011-of-00019.safetensors",
|
| 373 |
-
"transformer.decoder_layer.36.rms_norm.weight": "model-00011-of-00019.safetensors",
|
| 374 |
-
"transformer.decoder_layer.36.rms_norm_1.weight": "model-00011-of-00019.safetensors",
|
| 375 |
-
"transformer.decoder_layer.36.rms_norm_2.weight": "model-00011-of-00019.safetensors",
|
| 376 |
-
"transformer.decoder_layer.36.rms_norm_3.weight": "model-00011-of-00019.safetensors",
|
| 377 |
-
"transformer.decoder_layer.36.router.weight": "model-00011-of-00019.safetensors",
|
| 378 |
-
"transformer.decoder_layer.37.moe.0.linear.weight": "model-00012-of-00019.safetensors",
|
| 379 |
-
"transformer.decoder_layer.37.moe.0.linear_1.weight": "model-00012-of-00019.safetensors",
|
| 380 |
-
"transformer.decoder_layer.37.moe.0.linear_v.weight": "model-00011-of-00019.safetensors",
|
| 381 |
-
"transformer.decoder_layer.37.multi_head_attention.key.weight": "model-00011-of-00019.safetensors",
|
| 382 |
-
"transformer.decoder_layer.37.multi_head_attention.linear.weight": "model-00011-of-00019.safetensors",
|
| 383 |
-
"transformer.decoder_layer.37.multi_head_attention.query.weight": "model-00011-of-00019.safetensors",
|
| 384 |
-
"transformer.decoder_layer.37.multi_head_attention.value.weight": "model-00011-of-00019.safetensors",
|
| 385 |
-
"transformer.decoder_layer.37.rms_norm.weight": "model-00012-of-00019.safetensors",
|
| 386 |
-
"transformer.decoder_layer.37.rms_norm_1.weight": "model-00012-of-00019.safetensors",
|
| 387 |
-
"transformer.decoder_layer.37.rms_norm_2.weight": "model-00012-of-00019.safetensors",
|
| 388 |
-
"transformer.decoder_layer.37.rms_norm_3.weight": "model-00012-of-00019.safetensors",
|
| 389 |
-
"transformer.decoder_layer.37.router.weight": "model-00011-of-00019.safetensors",
|
| 390 |
-
"transformer.decoder_layer.38.moe.0.linear.weight": "model-00012-of-00019.safetensors",
|
| 391 |
-
"transformer.decoder_layer.38.moe.0.linear_1.weight": "model-00012-of-00019.safetensors",
|
| 392 |
-
"transformer.decoder_layer.38.moe.0.linear_v.weight": "model-00012-of-00019.safetensors",
|
| 393 |
-
"transformer.decoder_layer.38.multi_head_attention.key.weight": "model-00012-of-00019.safetensors",
|
| 394 |
-
"transformer.decoder_layer.38.multi_head_attention.linear.weight": "model-00012-of-00019.safetensors",
|
| 395 |
-
"transformer.decoder_layer.38.multi_head_attention.query.weight": "model-00012-of-00019.safetensors",
|
| 396 |
-
"transformer.decoder_layer.38.multi_head_attention.value.weight": "model-00012-of-00019.safetensors",
|
| 397 |
-
"transformer.decoder_layer.38.rms_norm.weight": "model-00012-of-00019.safetensors",
|
| 398 |
-
"transformer.decoder_layer.38.rms_norm_1.weight": "model-00012-of-00019.safetensors",
|
| 399 |
-
"transformer.decoder_layer.38.rms_norm_2.weight": "model-00012-of-00019.safetensors",
|
| 400 |
-
"transformer.decoder_layer.38.rms_norm_3.weight": "model-00012-of-00019.safetensors",
|
| 401 |
-
"transformer.decoder_layer.38.router.weight": "model-00012-of-00019.safetensors",
|
| 402 |
-
"transformer.decoder_layer.39.moe.0.linear.weight": "model-00012-of-00019.safetensors",
|
| 403 |
-
"transformer.decoder_layer.39.moe.0.linear_1.weight": "model-00012-of-00019.safetensors",
|
| 404 |
-
"transformer.decoder_layer.39.moe.0.linear_v.weight": "model-00012-of-00019.safetensors",
|
| 405 |
-
"transformer.decoder_layer.39.multi_head_attention.key.weight": "model-00012-of-00019.safetensors",
|
| 406 |
-
"transformer.decoder_layer.39.multi_head_attention.linear.weight": "model-00012-of-00019.safetensors",
|
| 407 |
-
"transformer.decoder_layer.39.multi_head_attention.query.weight": "model-00012-of-00019.safetensors",
|
| 408 |
-
"transformer.decoder_layer.39.multi_head_attention.value.weight": "model-00012-of-00019.safetensors",
|
| 409 |
-
"transformer.decoder_layer.39.rms_norm.weight": "model-00012-of-00019.safetensors",
|
| 410 |
-
"transformer.decoder_layer.39.rms_norm_1.weight": "model-00012-of-00019.safetensors",
|
| 411 |
-
"transformer.decoder_layer.39.rms_norm_2.weight": "model-00012-of-00019.safetensors",
|
| 412 |
-
"transformer.decoder_layer.39.rms_norm_3.weight": "model-00012-of-00019.safetensors",
|
| 413 |
-
"transformer.decoder_layer.39.router.weight": "model-00012-of-00019.safetensors",
|
| 414 |
-
"transformer.decoder_layer.4.moe.0.linear.weight": "model-00002-of-00019.safetensors",
|
| 415 |
-
"transformer.decoder_layer.4.moe.0.linear_1.weight": "model-00002-of-00019.safetensors",
|
| 416 |
-
"transformer.decoder_layer.4.moe.0.linear_v.weight": "model-00002-of-00019.safetensors",
|
| 417 |
-
"transformer.decoder_layer.4.multi_head_attention.key.weight": "model-00002-of-00019.safetensors",
|
| 418 |
-
"transformer.decoder_layer.4.multi_head_attention.linear.weight": "model-00002-of-00019.safetensors",
|
| 419 |
-
"transformer.decoder_layer.4.multi_head_attention.query.weight": "model-00002-of-00019.safetensors",
|
| 420 |
-
"transformer.decoder_layer.4.multi_head_attention.value.weight": "model-00002-of-00019.safetensors",
|
| 421 |
-
"transformer.decoder_layer.4.rms_norm.weight": "model-00002-of-00019.safetensors",
|
| 422 |
-
"transformer.decoder_layer.4.rms_norm_1.weight": "model-00002-of-00019.safetensors",
|
| 423 |
-
"transformer.decoder_layer.4.rms_norm_2.weight": "model-00002-of-00019.safetensors",
|
| 424 |
-
"transformer.decoder_layer.4.rms_norm_3.weight": "model-00002-of-00019.safetensors",
|
| 425 |
-
"transformer.decoder_layer.4.router.weight": "model-00002-of-00019.safetensors",
|
| 426 |
-
"transformer.decoder_layer.40.moe.0.linear.weight": "model-00012-of-00019.safetensors",
|
| 427 |
-
"transformer.decoder_layer.40.moe.0.linear_1.weight": "model-00012-of-00019.safetensors",
|
| 428 |
-
"transformer.decoder_layer.40.moe.0.linear_v.weight": "model-00012-of-00019.safetensors",
|
| 429 |
-
"transformer.decoder_layer.40.multi_head_attention.key.weight": "model-00012-of-00019.safetensors",
|
| 430 |
-
"transformer.decoder_layer.40.multi_head_attention.linear.weight": "model-00012-of-00019.safetensors",
|
| 431 |
-
"transformer.decoder_layer.40.multi_head_attention.query.weight": "model-00012-of-00019.safetensors",
|
| 432 |
-
"transformer.decoder_layer.40.multi_head_attention.value.weight": "model-00012-of-00019.safetensors",
|
| 433 |
-
"transformer.decoder_layer.40.rms_norm.weight": "model-00012-of-00019.safetensors",
|
| 434 |
-
"transformer.decoder_layer.40.rms_norm_1.weight": "model-00012-of-00019.safetensors",
|
| 435 |
-
"transformer.decoder_layer.40.rms_norm_2.weight": "model-00012-of-00019.safetensors",
|
| 436 |
-
"transformer.decoder_layer.40.rms_norm_3.weight": "model-00012-of-00019.safetensors",
|
| 437 |
-
"transformer.decoder_layer.40.router.weight": "model-00012-of-00019.safetensors",
|
| 438 |
-
"transformer.decoder_layer.41.moe.0.linear.weight": "model-00013-of-00019.safetensors",
|
| 439 |
-
"transformer.decoder_layer.41.moe.0.linear_1.weight": "model-00013-of-00019.safetensors",
|
| 440 |
-
"transformer.decoder_layer.41.moe.0.linear_v.weight": "model-00013-of-00019.safetensors",
|
| 441 |
-
"transformer.decoder_layer.41.multi_head_attention.key.weight": "model-00013-of-00019.safetensors",
|
| 442 |
-
"transformer.decoder_layer.41.multi_head_attention.linear.weight": "model-00013-of-00019.safetensors",
|
| 443 |
-
"transformer.decoder_layer.41.multi_head_attention.query.weight": "model-00013-of-00019.safetensors",
|
| 444 |
-
"transformer.decoder_layer.41.multi_head_attention.value.weight": "model-00013-of-00019.safetensors",
|
| 445 |
-
"transformer.decoder_layer.41.rms_norm.weight": "model-00013-of-00019.safetensors",
|
| 446 |
-
"transformer.decoder_layer.41.rms_norm_1.weight": "model-00013-of-00019.safetensors",
|
| 447 |
-
"transformer.decoder_layer.41.rms_norm_2.weight": "model-00013-of-00019.safetensors",
|
| 448 |
-
"transformer.decoder_layer.41.rms_norm_3.weight": "model-00013-of-00019.safetensors",
|
| 449 |
-
"transformer.decoder_layer.41.router.weight": "model-00013-of-00019.safetensors",
|
| 450 |
-
"transformer.decoder_layer.42.moe.0.linear.weight": "model-00013-of-00019.safetensors",
|
| 451 |
-
"transformer.decoder_layer.42.moe.0.linear_1.weight": "model-00013-of-00019.safetensors",
|
| 452 |
-
"transformer.decoder_layer.42.moe.0.linear_v.weight": "model-00013-of-00019.safetensors",
|
| 453 |
-
"transformer.decoder_layer.42.multi_head_attention.key.weight": "model-00013-of-00019.safetensors",
|
| 454 |
-
"transformer.decoder_layer.42.multi_head_attention.linear.weight": "model-00013-of-00019.safetensors",
|
| 455 |
-
"transformer.decoder_layer.42.multi_head_attention.query.weight": "model-00013-of-00019.safetensors",
|
| 456 |
-
"transformer.decoder_layer.42.multi_head_attention.value.weight": "model-00013-of-00019.safetensors",
|
| 457 |
-
"transformer.decoder_layer.42.rms_norm.weight": "model-00013-of-00019.safetensors",
|
| 458 |
-
"transformer.decoder_layer.42.rms_norm_1.weight": "model-00013-of-00019.safetensors",
|
| 459 |
-
"transformer.decoder_layer.42.rms_norm_2.weight": "model-00013-of-00019.safetensors",
|
| 460 |
-
"transformer.decoder_layer.42.rms_norm_3.weight": "model-00013-of-00019.safetensors",
|
| 461 |
-
"transformer.decoder_layer.42.router.weight": "model-00013-of-00019.safetensors",
|
| 462 |
-
"transformer.decoder_layer.43.moe.0.linear.weight": "model-00013-of-00019.safetensors",
|
| 463 |
-
"transformer.decoder_layer.43.moe.0.linear_1.weight": "model-00013-of-00019.safetensors",
|
| 464 |
-
"transformer.decoder_layer.43.moe.0.linear_v.weight": "model-00013-of-00019.safetensors",
|
| 465 |
-
"transformer.decoder_layer.43.multi_head_attention.key.weight": "model-00013-of-00019.safetensors",
|
| 466 |
-
"transformer.decoder_layer.43.multi_head_attention.linear.weight": "model-00013-of-00019.safetensors",
|
| 467 |
-
"transformer.decoder_layer.43.multi_head_attention.query.weight": "model-00013-of-00019.safetensors",
|
| 468 |
-
"transformer.decoder_layer.43.multi_head_attention.value.weight": "model-00013-of-00019.safetensors",
|
| 469 |
-
"transformer.decoder_layer.43.rms_norm.weight": "model-00013-of-00019.safetensors",
|
| 470 |
-
"transformer.decoder_layer.43.rms_norm_1.weight": "model-00013-of-00019.safetensors",
|
| 471 |
-
"transformer.decoder_layer.43.rms_norm_2.weight": "model-00013-of-00019.safetensors",
|
| 472 |
-
"transformer.decoder_layer.43.rms_norm_3.weight": "model-00013-of-00019.safetensors",
|
| 473 |
-
"transformer.decoder_layer.43.router.weight": "model-00013-of-00019.safetensors",
|
| 474 |
-
"transformer.decoder_layer.44.moe.0.linear.weight": "model-00014-of-00019.safetensors",
|
| 475 |
-
"transformer.decoder_layer.44.moe.0.linear_1.weight": "model-00014-of-00019.safetensors",
|
| 476 |
-
"transformer.decoder_layer.44.moe.0.linear_v.weight": "model-00013-of-00019.safetensors",
|
| 477 |
-
"transformer.decoder_layer.44.multi_head_attention.key.weight": "model-00013-of-00019.safetensors",
|
| 478 |
-
"transformer.decoder_layer.44.multi_head_attention.linear.weight": "model-00013-of-00019.safetensors",
|
| 479 |
-
"transformer.decoder_layer.44.multi_head_attention.query.weight": "model-00013-of-00019.safetensors",
|
| 480 |
-
"transformer.decoder_layer.44.multi_head_attention.value.weight": "model-00013-of-00019.safetensors",
|
| 481 |
-
"transformer.decoder_layer.44.rms_norm.weight": "model-00014-of-00019.safetensors",
|
| 482 |
-
"transformer.decoder_layer.44.rms_norm_1.weight": "model-00014-of-00019.safetensors",
|
| 483 |
-
"transformer.decoder_layer.44.rms_norm_2.weight": "model-00014-of-00019.safetensors",
|
| 484 |
-
"transformer.decoder_layer.44.rms_norm_3.weight": "model-00014-of-00019.safetensors",
|
| 485 |
-
"transformer.decoder_layer.44.router.weight": "model-00013-of-00019.safetensors",
|
| 486 |
-
"transformer.decoder_layer.45.moe.0.linear.weight": "model-00014-of-00019.safetensors",
|
| 487 |
-
"transformer.decoder_layer.45.moe.0.linear_1.weight": "model-00014-of-00019.safetensors",
|
| 488 |
-
"transformer.decoder_layer.45.moe.0.linear_v.weight": "model-00014-of-00019.safetensors",
|
| 489 |
-
"transformer.decoder_layer.45.multi_head_attention.key.weight": "model-00014-of-00019.safetensors",
|
| 490 |
-
"transformer.decoder_layer.45.multi_head_attention.linear.weight": "model-00014-of-00019.safetensors",
|
| 491 |
-
"transformer.decoder_layer.45.multi_head_attention.query.weight": "model-00014-of-00019.safetensors",
|
| 492 |
-
"transformer.decoder_layer.45.multi_head_attention.value.weight": "model-00014-of-00019.safetensors",
|
| 493 |
-
"transformer.decoder_layer.45.rms_norm.weight": "model-00014-of-00019.safetensors",
|
| 494 |
-
"transformer.decoder_layer.45.rms_norm_1.weight": "model-00014-of-00019.safetensors",
|
| 495 |
-
"transformer.decoder_layer.45.rms_norm_2.weight": "model-00014-of-00019.safetensors",
|
| 496 |
-
"transformer.decoder_layer.45.rms_norm_3.weight": "model-00014-of-00019.safetensors",
|
| 497 |
-
"transformer.decoder_layer.45.router.weight": "model-00014-of-00019.safetensors",
|
| 498 |
-
"transformer.decoder_layer.46.moe.0.linear.weight": "model-00014-of-00019.safetensors",
|
| 499 |
-
"transformer.decoder_layer.46.moe.0.linear_1.weight": "model-00014-of-00019.safetensors",
|
| 500 |
-
"transformer.decoder_layer.46.moe.0.linear_v.weight": "model-00014-of-00019.safetensors",
|
| 501 |
-
"transformer.decoder_layer.46.multi_head_attention.key.weight": "model-00014-of-00019.safetensors",
|
| 502 |
-
"transformer.decoder_layer.46.multi_head_attention.linear.weight": "model-00014-of-00019.safetensors",
|
| 503 |
-
"transformer.decoder_layer.46.multi_head_attention.query.weight": "model-00014-of-00019.safetensors",
|
| 504 |
-
"transformer.decoder_layer.46.multi_head_attention.value.weight": "model-00014-of-00019.safetensors",
|
| 505 |
-
"transformer.decoder_layer.46.rms_norm.weight": "model-00014-of-00019.safetensors",
|
| 506 |
-
"transformer.decoder_layer.46.rms_norm_1.weight": "model-00014-of-00019.safetensors",
|
| 507 |
-
"transformer.decoder_layer.46.rms_norm_2.weight": "model-00014-of-00019.safetensors",
|
| 508 |
-
"transformer.decoder_layer.46.rms_norm_3.weight": "model-00014-of-00019.safetensors",
|
| 509 |
-
"transformer.decoder_layer.46.router.weight": "model-00014-of-00019.safetensors",
|
| 510 |
-
"transformer.decoder_layer.47.moe.0.linear.weight": "model-00014-of-00019.safetensors",
|
| 511 |
-
"transformer.decoder_layer.47.moe.0.linear_1.weight": "model-00014-of-00019.safetensors",
|
| 512 |
-
"transformer.decoder_layer.47.moe.0.linear_v.weight": "model-00014-of-00019.safetensors",
|
| 513 |
-
"transformer.decoder_layer.47.multi_head_attention.key.weight": "model-00014-of-00019.safetensors",
|
| 514 |
-
"transformer.decoder_layer.47.multi_head_attention.linear.weight": "model-00014-of-00019.safetensors",
|
| 515 |
-
"transformer.decoder_layer.47.multi_head_attention.query.weight": "model-00014-of-00019.safetensors",
|
| 516 |
-
"transformer.decoder_layer.47.multi_head_attention.value.weight": "model-00014-of-00019.safetensors",
|
| 517 |
-
"transformer.decoder_layer.47.rms_norm.weight": "model-00014-of-00019.safetensors",
|
| 518 |
-
"transformer.decoder_layer.47.rms_norm_1.weight": "model-00014-of-00019.safetensors",
|
| 519 |
-
"transformer.decoder_layer.47.rms_norm_2.weight": "model-00014-of-00019.safetensors",
|
| 520 |
-
"transformer.decoder_layer.47.rms_norm_3.weight": "model-00014-of-00019.safetensors",
|
| 521 |
-
"transformer.decoder_layer.47.router.weight": "model-00014-of-00019.safetensors",
|
| 522 |
-
"transformer.decoder_layer.48.moe.0.linear.weight": "model-00015-of-00019.safetensors",
|
| 523 |
-
"transformer.decoder_layer.48.moe.0.linear_1.weight": "model-00015-of-00019.safetensors",
|
| 524 |
-
"transformer.decoder_layer.48.moe.0.linear_v.weight": "model-00015-of-00019.safetensors",
|
| 525 |
-
"transformer.decoder_layer.48.multi_head_attention.key.weight": "model-00015-of-00019.safetensors",
|
| 526 |
-
"transformer.decoder_layer.48.multi_head_attention.linear.weight": "model-00015-of-00019.safetensors",
|
| 527 |
-
"transformer.decoder_layer.48.multi_head_attention.query.weight": "model-00015-of-00019.safetensors",
|
| 528 |
-
"transformer.decoder_layer.48.multi_head_attention.value.weight": "model-00015-of-00019.safetensors",
|
| 529 |
-
"transformer.decoder_layer.48.rms_norm.weight": "model-00015-of-00019.safetensors",
|
| 530 |
-
"transformer.decoder_layer.48.rms_norm_1.weight": "model-00015-of-00019.safetensors",
|
| 531 |
-
"transformer.decoder_layer.48.rms_norm_2.weight": "model-00015-of-00019.safetensors",
|
| 532 |
-
"transformer.decoder_layer.48.rms_norm_3.weight": "model-00015-of-00019.safetensors",
|
| 533 |
-
"transformer.decoder_layer.48.router.weight": "model-00015-of-00019.safetensors",
|
| 534 |
-
"transformer.decoder_layer.49.moe.0.linear.weight": "model-00015-of-00019.safetensors",
|
| 535 |
-
"transformer.decoder_layer.49.moe.0.linear_1.weight": "model-00015-of-00019.safetensors",
|
| 536 |
-
"transformer.decoder_layer.49.moe.0.linear_v.weight": "model-00015-of-00019.safetensors",
|
| 537 |
-
"transformer.decoder_layer.49.multi_head_attention.key.weight": "model-00015-of-00019.safetensors",
|
| 538 |
-
"transformer.decoder_layer.49.multi_head_attention.linear.weight": "model-00015-of-00019.safetensors",
|
| 539 |
-
"transformer.decoder_layer.49.multi_head_attention.query.weight": "model-00015-of-00019.safetensors",
|
| 540 |
-
"transformer.decoder_layer.49.multi_head_attention.value.weight": "model-00015-of-00019.safetensors",
|
| 541 |
-
"transformer.decoder_layer.49.rms_norm.weight": "model-00015-of-00019.safetensors",
|
| 542 |
-
"transformer.decoder_layer.49.rms_norm_1.weight": "model-00015-of-00019.safetensors",
|
| 543 |
-
"transformer.decoder_layer.49.rms_norm_2.weight": "model-00015-of-00019.safetensors",
|
| 544 |
-
"transformer.decoder_layer.49.rms_norm_3.weight": "model-00015-of-00019.safetensors",
|
| 545 |
-
"transformer.decoder_layer.49.router.weight": "model-00015-of-00019.safetensors",
|
| 546 |
-
"transformer.decoder_layer.5.moe.0.linear.weight": "model-00002-of-00019.safetensors",
|
| 547 |
-
"transformer.decoder_layer.5.moe.0.linear_1.weight": "model-00002-of-00019.safetensors",
|
| 548 |
-
"transformer.decoder_layer.5.moe.0.linear_v.weight": "model-00002-of-00019.safetensors",
|
| 549 |
-
"transformer.decoder_layer.5.multi_head_attention.key.weight": "model-00002-of-00019.safetensors",
|
| 550 |
-
"transformer.decoder_layer.5.multi_head_attention.linear.weight": "model-00002-of-00019.safetensors",
|
| 551 |
-
"transformer.decoder_layer.5.multi_head_attention.query.weight": "model-00002-of-00019.safetensors",
|
| 552 |
-
"transformer.decoder_layer.5.multi_head_attention.value.weight": "model-00002-of-00019.safetensors",
|
| 553 |
-
"transformer.decoder_layer.5.rms_norm.weight": "model-00002-of-00019.safetensors",
|
| 554 |
-
"transformer.decoder_layer.5.rms_norm_1.weight": "model-00002-of-00019.safetensors",
|
| 555 |
-
"transformer.decoder_layer.5.rms_norm_2.weight": "model-00002-of-00019.safetensors",
|
| 556 |
-
"transformer.decoder_layer.5.rms_norm_3.weight": "model-00002-of-00019.safetensors",
|
| 557 |
-
"transformer.decoder_layer.5.router.weight": "model-00002-of-00019.safetensors",
|
| 558 |
-
"transformer.decoder_layer.50.moe.0.linear.weight": "model-00015-of-00019.safetensors",
|
| 559 |
-
"transformer.decoder_layer.50.moe.0.linear_1.weight": "model-00015-of-00019.safetensors",
|
| 560 |
-
"transformer.decoder_layer.50.moe.0.linear_v.weight": "model-00015-of-00019.safetensors",
|
| 561 |
-
"transformer.decoder_layer.50.multi_head_attention.key.weight": "model-00015-of-00019.safetensors",
|
| 562 |
-
"transformer.decoder_layer.50.multi_head_attention.linear.weight": "model-00015-of-00019.safetensors",
|
| 563 |
-
"transformer.decoder_layer.50.multi_head_attention.query.weight": "model-00015-of-00019.safetensors",
|
| 564 |
-
"transformer.decoder_layer.50.multi_head_attention.value.weight": "model-00015-of-00019.safetensors",
|
| 565 |
-
"transformer.decoder_layer.50.rms_norm.weight": "model-00015-of-00019.safetensors",
|
| 566 |
-
"transformer.decoder_layer.50.rms_norm_1.weight": "model-00015-of-00019.safetensors",
|
| 567 |
-
"transformer.decoder_layer.50.rms_norm_2.weight": "model-00015-of-00019.safetensors",
|
| 568 |
-
"transformer.decoder_layer.50.rms_norm_3.weight": "model-00015-of-00019.safetensors",
|
| 569 |
-
"transformer.decoder_layer.50.router.weight": "model-00015-of-00019.safetensors",
|
| 570 |
-
"transformer.decoder_layer.51.moe.0.linear.weight": "model-00016-of-00019.safetensors",
|
| 571 |
-
"transformer.decoder_layer.51.moe.0.linear_1.weight": "model-00016-of-00019.safetensors",
|
| 572 |
-
"transformer.decoder_layer.51.moe.0.linear_v.weight": "model-00015-of-00019.safetensors",
|
| 573 |
-
"transformer.decoder_layer.51.multi_head_attention.key.weight": "model-00015-of-00019.safetensors",
|
| 574 |
-
"transformer.decoder_layer.51.multi_head_attention.linear.weight": "model-00015-of-00019.safetensors",
|
| 575 |
-
"transformer.decoder_layer.51.multi_head_attention.query.weight": "model-00015-of-00019.safetensors",
|
| 576 |
-
"transformer.decoder_layer.51.multi_head_attention.value.weight": "model-00015-of-00019.safetensors",
|
| 577 |
-
"transformer.decoder_layer.51.rms_norm.weight": "model-00016-of-00019.safetensors",
|
| 578 |
-
"transformer.decoder_layer.51.rms_norm_1.weight": "model-00016-of-00019.safetensors",
|
| 579 |
-
"transformer.decoder_layer.51.rms_norm_2.weight": "model-00016-of-00019.safetensors",
|
| 580 |
-
"transformer.decoder_layer.51.rms_norm_3.weight": "model-00016-of-00019.safetensors",
|
| 581 |
-
"transformer.decoder_layer.51.router.weight": "model-00015-of-00019.safetensors",
|
| 582 |
-
"transformer.decoder_layer.52.moe.0.linear.weight": "model-00016-of-00019.safetensors",
|
| 583 |
-
"transformer.decoder_layer.52.moe.0.linear_1.weight": "model-00016-of-00019.safetensors",
|
| 584 |
-
"transformer.decoder_layer.52.moe.0.linear_v.weight": "model-00016-of-00019.safetensors",
|
| 585 |
-
"transformer.decoder_layer.52.multi_head_attention.key.weight": "model-00016-of-00019.safetensors",
|
| 586 |
-
"transformer.decoder_layer.52.multi_head_attention.linear.weight": "model-00016-of-00019.safetensors",
|
| 587 |
-
"transformer.decoder_layer.52.multi_head_attention.query.weight": "model-00016-of-00019.safetensors",
|
| 588 |
-
"transformer.decoder_layer.52.multi_head_attention.value.weight": "model-00016-of-00019.safetensors",
|
| 589 |
-
"transformer.decoder_layer.52.rms_norm.weight": "model-00016-of-00019.safetensors",
|
| 590 |
-
"transformer.decoder_layer.52.rms_norm_1.weight": "model-00016-of-00019.safetensors",
|
| 591 |
-
"transformer.decoder_layer.52.rms_norm_2.weight": "model-00016-of-00019.safetensors",
|
| 592 |
-
"transformer.decoder_layer.52.rms_norm_3.weight": "model-00016-of-00019.safetensors",
|
| 593 |
-
"transformer.decoder_layer.52.router.weight": "model-00016-of-00019.safetensors",
|
| 594 |
-
"transformer.decoder_layer.53.moe.0.linear.weight": "model-00016-of-00019.safetensors",
|
| 595 |
-
"transformer.decoder_layer.53.moe.0.linear_1.weight": "model-00016-of-00019.safetensors",
|
| 596 |
-
"transformer.decoder_layer.53.moe.0.linear_v.weight": "model-00016-of-00019.safetensors",
|
| 597 |
-
"transformer.decoder_layer.53.multi_head_attention.key.weight": "model-00016-of-00019.safetensors",
|
| 598 |
-
"transformer.decoder_layer.53.multi_head_attention.linear.weight": "model-00016-of-00019.safetensors",
|
| 599 |
-
"transformer.decoder_layer.53.multi_head_attention.query.weight": "model-00016-of-00019.safetensors",
|
| 600 |
-
"transformer.decoder_layer.53.multi_head_attention.value.weight": "model-00016-of-00019.safetensors",
|
| 601 |
-
"transformer.decoder_layer.53.rms_norm.weight": "model-00016-of-00019.safetensors",
|
| 602 |
-
"transformer.decoder_layer.53.rms_norm_1.weight": "model-00016-of-00019.safetensors",
|
| 603 |
-
"transformer.decoder_layer.53.rms_norm_2.weight": "model-00016-of-00019.safetensors",
|
| 604 |
-
"transformer.decoder_layer.53.rms_norm_3.weight": "model-00016-of-00019.safetensors",
|
| 605 |
-
"transformer.decoder_layer.53.router.weight": "model-00016-of-00019.safetensors",
|
| 606 |
-
"transformer.decoder_layer.54.moe.0.linear.weight": "model-00016-of-00019.safetensors",
|
| 607 |
-
"transformer.decoder_layer.54.moe.0.linear_1.weight": "model-00016-of-00019.safetensors",
|
| 608 |
-
"transformer.decoder_layer.54.moe.0.linear_v.weight": "model-00016-of-00019.safetensors",
|
| 609 |
-
"transformer.decoder_layer.54.multi_head_attention.key.weight": "model-00016-of-00019.safetensors",
|
| 610 |
-
"transformer.decoder_layer.54.multi_head_attention.linear.weight": "model-00016-of-00019.safetensors",
|
| 611 |
-
"transformer.decoder_layer.54.multi_head_attention.query.weight": "model-00016-of-00019.safetensors",
|
| 612 |
-
"transformer.decoder_layer.54.multi_head_attention.value.weight": "model-00016-of-00019.safetensors",
|
| 613 |
-
"transformer.decoder_layer.54.rms_norm.weight": "model-00016-of-00019.safetensors",
|
| 614 |
-
"transformer.decoder_layer.54.rms_norm_1.weight": "model-00016-of-00019.safetensors",
|
| 615 |
-
"transformer.decoder_layer.54.rms_norm_2.weight": "model-00016-of-00019.safetensors",
|
| 616 |
-
"transformer.decoder_layer.54.rms_norm_3.weight": "model-00016-of-00019.safetensors",
|
| 617 |
-
"transformer.decoder_layer.54.router.weight": "model-00016-of-00019.safetensors",
|
| 618 |
-
"transformer.decoder_layer.55.moe.0.linear.weight": "model-00017-of-00019.safetensors",
|
| 619 |
-
"transformer.decoder_layer.55.moe.0.linear_1.weight": "model-00017-of-00019.safetensors",
|
| 620 |
-
"transformer.decoder_layer.55.moe.0.linear_v.weight": "model-00017-of-00019.safetensors",
|
| 621 |
-
"transformer.decoder_layer.55.multi_head_attention.key.weight": "model-00017-of-00019.safetensors",
|
| 622 |
-
"transformer.decoder_layer.55.multi_head_attention.linear.weight": "model-00017-of-00019.safetensors",
|
| 623 |
-
"transformer.decoder_layer.55.multi_head_attention.query.weight": "model-00017-of-00019.safetensors",
|
| 624 |
-
"transformer.decoder_layer.55.multi_head_attention.value.weight": "model-00017-of-00019.safetensors",
|
| 625 |
-
"transformer.decoder_layer.55.rms_norm.weight": "model-00017-of-00019.safetensors",
|
| 626 |
-
"transformer.decoder_layer.55.rms_norm_1.weight": "model-00017-of-00019.safetensors",
|
| 627 |
-
"transformer.decoder_layer.55.rms_norm_2.weight": "model-00017-of-00019.safetensors",
|
| 628 |
-
"transformer.decoder_layer.55.rms_norm_3.weight": "model-00017-of-00019.safetensors",
|
| 629 |
-
"transformer.decoder_layer.55.router.weight": "model-00017-of-00019.safetensors",
|
| 630 |
-
"transformer.decoder_layer.56.moe.0.linear.weight": "model-00017-of-00019.safetensors",
|
| 631 |
-
"transformer.decoder_layer.56.moe.0.linear_1.weight": "model-00017-of-00019.safetensors",
|
| 632 |
-
"transformer.decoder_layer.56.moe.0.linear_v.weight": "model-00017-of-00019.safetensors",
|
| 633 |
-
"transformer.decoder_layer.56.multi_head_attention.key.weight": "model-00017-of-00019.safetensors",
|
| 634 |
-
"transformer.decoder_layer.56.multi_head_attention.linear.weight": "model-00017-of-00019.safetensors",
|
| 635 |
-
"transformer.decoder_layer.56.multi_head_attention.query.weight": "model-00017-of-00019.safetensors",
|
| 636 |
-
"transformer.decoder_layer.56.multi_head_attention.value.weight": "model-00017-of-00019.safetensors",
|
| 637 |
-
"transformer.decoder_layer.56.rms_norm.weight": "model-00017-of-00019.safetensors",
|
| 638 |
-
"transformer.decoder_layer.56.rms_norm_1.weight": "model-00017-of-00019.safetensors",
|
| 639 |
-
"transformer.decoder_layer.56.rms_norm_2.weight": "model-00017-of-00019.safetensors",
|
| 640 |
-
"transformer.decoder_layer.56.rms_norm_3.weight": "model-00017-of-00019.safetensors",
|
| 641 |
-
"transformer.decoder_layer.56.router.weight": "model-00017-of-00019.safetensors",
|
| 642 |
-
"transformer.decoder_layer.57.moe.0.linear.weight": "model-00017-of-00019.safetensors",
|
| 643 |
-
"transformer.decoder_layer.57.moe.0.linear_1.weight": "model-00017-of-00019.safetensors",
|
| 644 |
-
"transformer.decoder_layer.57.moe.0.linear_v.weight": "model-00017-of-00019.safetensors",
|
| 645 |
-
"transformer.decoder_layer.57.multi_head_attention.key.weight": "model-00017-of-00019.safetensors",
|
| 646 |
-
"transformer.decoder_layer.57.multi_head_attention.linear.weight": "model-00017-of-00019.safetensors",
|
| 647 |
-
"transformer.decoder_layer.57.multi_head_attention.query.weight": "model-00017-of-00019.safetensors",
|
| 648 |
-
"transformer.decoder_layer.57.multi_head_attention.value.weight": "model-00017-of-00019.safetensors",
|
| 649 |
-
"transformer.decoder_layer.57.rms_norm.weight": "model-00017-of-00019.safetensors",
|
| 650 |
-
"transformer.decoder_layer.57.rms_norm_1.weight": "model-00017-of-00019.safetensors",
|
| 651 |
-
"transformer.decoder_layer.57.rms_norm_2.weight": "model-00017-of-00019.safetensors",
|
| 652 |
-
"transformer.decoder_layer.57.rms_norm_3.weight": "model-00017-of-00019.safetensors",
|
| 653 |
-
"transformer.decoder_layer.57.router.weight": "model-00017-of-00019.safetensors",
|
| 654 |
-
"transformer.decoder_layer.58.moe.0.linear.weight": "model-00018-of-00019.safetensors",
|
| 655 |
-
"transformer.decoder_layer.58.moe.0.linear_1.weight": "model-00018-of-00019.safetensors",
|
| 656 |
-
"transformer.decoder_layer.58.moe.0.linear_v.weight": "model-00017-of-00019.safetensors",
|
| 657 |
-
"transformer.decoder_layer.58.multi_head_attention.key.weight": "model-00017-of-00019.safetensors",
|
| 658 |
-
"transformer.decoder_layer.58.multi_head_attention.linear.weight": "model-00017-of-00019.safetensors",
|
| 659 |
-
"transformer.decoder_layer.58.multi_head_attention.query.weight": "model-00017-of-00019.safetensors",
|
| 660 |
-
"transformer.decoder_layer.58.multi_head_attention.value.weight": "model-00017-of-00019.safetensors",
|
| 661 |
-
"transformer.decoder_layer.58.rms_norm.weight": "model-00018-of-00019.safetensors",
|
| 662 |
-
"transformer.decoder_layer.58.rms_norm_1.weight": "model-00018-of-00019.safetensors",
|
| 663 |
-
"transformer.decoder_layer.58.rms_norm_2.weight": "model-00018-of-00019.safetensors",
|
| 664 |
-
"transformer.decoder_layer.58.rms_norm_3.weight": "model-00018-of-00019.safetensors",
|
| 665 |
-
"transformer.decoder_layer.58.router.weight": "model-00017-of-00019.safetensors",
|
| 666 |
-
"transformer.decoder_layer.59.moe.0.linear.weight": "model-00018-of-00019.safetensors",
|
| 667 |
-
"transformer.decoder_layer.59.moe.0.linear_1.weight": "model-00018-of-00019.safetensors",
|
| 668 |
-
"transformer.decoder_layer.59.moe.0.linear_v.weight": "model-00018-of-00019.safetensors",
|
| 669 |
-
"transformer.decoder_layer.59.multi_head_attention.key.weight": "model-00018-of-00019.safetensors",
|
| 670 |
-
"transformer.decoder_layer.59.multi_head_attention.linear.weight": "model-00018-of-00019.safetensors",
|
| 671 |
-
"transformer.decoder_layer.59.multi_head_attention.query.weight": "model-00018-of-00019.safetensors",
|
| 672 |
-
"transformer.decoder_layer.59.multi_head_attention.value.weight": "model-00018-of-00019.safetensors",
|
| 673 |
-
"transformer.decoder_layer.59.rms_norm.weight": "model-00018-of-00019.safetensors",
|
| 674 |
-
"transformer.decoder_layer.59.rms_norm_1.weight": "model-00018-of-00019.safetensors",
|
| 675 |
-
"transformer.decoder_layer.59.rms_norm_2.weight": "model-00018-of-00019.safetensors",
|
| 676 |
-
"transformer.decoder_layer.59.rms_norm_3.weight": "model-00018-of-00019.safetensors",
|
| 677 |
-
"transformer.decoder_layer.59.router.weight": "model-00018-of-00019.safetensors",
|
| 678 |
-
"transformer.decoder_layer.6.moe.0.linear.weight": "model-00003-of-00019.safetensors",
|
| 679 |
-
"transformer.decoder_layer.6.moe.0.linear_1.weight": "model-00003-of-00019.safetensors",
|
| 680 |
-
"transformer.decoder_layer.6.moe.0.linear_v.weight": "model-00003-of-00019.safetensors",
|
| 681 |
-
"transformer.decoder_layer.6.multi_head_attention.key.weight": "model-00003-of-00019.safetensors",
|
| 682 |
-
"transformer.decoder_layer.6.multi_head_attention.linear.weight": "model-00003-of-00019.safetensors",
|
| 683 |
-
"transformer.decoder_layer.6.multi_head_attention.query.weight": "model-00003-of-00019.safetensors",
|
| 684 |
-
"transformer.decoder_layer.6.multi_head_attention.value.weight": "model-00003-of-00019.safetensors",
|
| 685 |
-
"transformer.decoder_layer.6.rms_norm.weight": "model-00003-of-00019.safetensors",
|
| 686 |
-
"transformer.decoder_layer.6.rms_norm_1.weight": "model-00003-of-00019.safetensors",
|
| 687 |
-
"transformer.decoder_layer.6.rms_norm_2.weight": "model-00003-of-00019.safetensors",
|
| 688 |
-
"transformer.decoder_layer.6.rms_norm_3.weight": "model-00003-of-00019.safetensors",
|
| 689 |
-
"transformer.decoder_layer.6.router.weight": "model-00003-of-00019.safetensors",
|
| 690 |
-
"transformer.decoder_layer.60.moe.0.linear.weight": "model-00018-of-00019.safetensors",
|
| 691 |
-
"transformer.decoder_layer.60.moe.0.linear_1.weight": "model-00018-of-00019.safetensors",
|
| 692 |
-
"transformer.decoder_layer.60.moe.0.linear_v.weight": "model-00018-of-00019.safetensors",
|
| 693 |
-
"transformer.decoder_layer.60.multi_head_attention.key.weight": "model-00018-of-00019.safetensors",
|
| 694 |
-
"transformer.decoder_layer.60.multi_head_attention.linear.weight": "model-00018-of-00019.safetensors",
|
| 695 |
-
"transformer.decoder_layer.60.multi_head_attention.query.weight": "model-00018-of-00019.safetensors",
|
| 696 |
-
"transformer.decoder_layer.60.multi_head_attention.value.weight": "model-00018-of-00019.safetensors",
|
| 697 |
-
"transformer.decoder_layer.60.rms_norm.weight": "model-00018-of-00019.safetensors",
|
| 698 |
-
"transformer.decoder_layer.60.rms_norm_1.weight": "model-00018-of-00019.safetensors",
|
| 699 |
-
"transformer.decoder_layer.60.rms_norm_2.weight": "model-00018-of-00019.safetensors",
|
| 700 |
-
"transformer.decoder_layer.60.rms_norm_3.weight": "model-00018-of-00019.safetensors",
|
| 701 |
-
"transformer.decoder_layer.60.router.weight": "model-00018-of-00019.safetensors",
|
| 702 |
-
"transformer.decoder_layer.61.moe.0.linear.weight": "model-00018-of-00019.safetensors",
|
| 703 |
-
"transformer.decoder_layer.61.moe.0.linear_1.weight": "model-00018-of-00019.safetensors",
|
| 704 |
-
"transformer.decoder_layer.61.moe.0.linear_v.weight": "model-00018-of-00019.safetensors",
|
| 705 |
-
"transformer.decoder_layer.61.multi_head_attention.key.weight": "model-00018-of-00019.safetensors",
|
| 706 |
-
"transformer.decoder_layer.61.multi_head_attention.linear.weight": "model-00018-of-00019.safetensors",
|
| 707 |
-
"transformer.decoder_layer.61.multi_head_attention.query.weight": "model-00018-of-00019.safetensors",
|
| 708 |
-
"transformer.decoder_layer.61.multi_head_attention.value.weight": "model-00018-of-00019.safetensors",
|
| 709 |
-
"transformer.decoder_layer.61.rms_norm.weight": "model-00018-of-00019.safetensors",
|
| 710 |
-
"transformer.decoder_layer.61.rms_norm_1.weight": "model-00018-of-00019.safetensors",
|
| 711 |
-
"transformer.decoder_layer.61.rms_norm_2.weight": "model-00018-of-00019.safetensors",
|
| 712 |
-
"transformer.decoder_layer.61.rms_norm_3.weight": "model-00018-of-00019.safetensors",
|
| 713 |
-
"transformer.decoder_layer.61.router.weight": "model-00018-of-00019.safetensors",
|
| 714 |
-
"transformer.decoder_layer.62.moe.0.linear.weight": "model-00019-of-00019.safetensors",
|
| 715 |
-
"transformer.decoder_layer.62.moe.0.linear_1.weight": "model-00019-of-00019.safetensors",
|
| 716 |
-
"transformer.decoder_layer.62.moe.0.linear_v.weight": "model-00019-of-00019.safetensors",
|
| 717 |
-
"transformer.decoder_layer.62.multi_head_attention.key.weight": "model-00019-of-00019.safetensors",
|
| 718 |
-
"transformer.decoder_layer.62.multi_head_attention.linear.weight": "model-00019-of-00019.safetensors",
|
| 719 |
-
"transformer.decoder_layer.62.multi_head_attention.query.weight": "model-00019-of-00019.safetensors",
|
| 720 |
-
"transformer.decoder_layer.62.multi_head_attention.value.weight": "model-00019-of-00019.safetensors",
|
| 721 |
-
"transformer.decoder_layer.62.rms_norm.weight": "model-00019-of-00019.safetensors",
|
| 722 |
-
"transformer.decoder_layer.62.rms_norm_1.weight": "model-00019-of-00019.safetensors",
|
| 723 |
-
"transformer.decoder_layer.62.rms_norm_2.weight": "model-00019-of-00019.safetensors",
|
| 724 |
-
"transformer.decoder_layer.62.rms_norm_3.weight": "model-00019-of-00019.safetensors",
|
| 725 |
-
"transformer.decoder_layer.62.router.weight": "model-00019-of-00019.safetensors",
|
| 726 |
-
"transformer.decoder_layer.63.moe.0.linear.weight": "model-00019-of-00019.safetensors",
|
| 727 |
-
"transformer.decoder_layer.63.moe.0.linear_1.weight": "model-00019-of-00019.safetensors",
|
| 728 |
-
"transformer.decoder_layer.63.moe.0.linear_v.weight": "model-00019-of-00019.safetensors",
|
| 729 |
-
"transformer.decoder_layer.63.multi_head_attention.key.weight": "model-00019-of-00019.safetensors",
|
| 730 |
-
"transformer.decoder_layer.63.multi_head_attention.linear.weight": "model-00019-of-00019.safetensors",
|
| 731 |
-
"transformer.decoder_layer.63.multi_head_attention.query.weight": "model-00019-of-00019.safetensors",
|
| 732 |
-
"transformer.decoder_layer.63.multi_head_attention.value.weight": "model-00019-of-00019.safetensors",
|
| 733 |
-
"transformer.decoder_layer.63.rms_norm.weight": "model-00019-of-00019.safetensors",
|
| 734 |
-
"transformer.decoder_layer.63.rms_norm_1.weight": "model-00019-of-00019.safetensors",
|
| 735 |
-
"transformer.decoder_layer.63.rms_norm_2.weight": "model-00019-of-00019.safetensors",
|
| 736 |
-
"transformer.decoder_layer.63.rms_norm_3.weight": "model-00019-of-00019.safetensors",
|
| 737 |
-
"transformer.decoder_layer.63.router.weight": "model-00019-of-00019.safetensors",
|
| 738 |
-
"transformer.decoder_layer.7.moe.0.linear.weight": "model-00003-of-00019.safetensors",
|
| 739 |
-
"transformer.decoder_layer.7.moe.0.linear_1.weight": "model-00003-of-00019.safetensors",
|
| 740 |
-
"transformer.decoder_layer.7.moe.0.linear_v.weight": "model-00003-of-00019.safetensors",
|
| 741 |
-
"transformer.decoder_layer.7.multi_head_attention.key.weight": "model-00003-of-00019.safetensors",
|
| 742 |
-
"transformer.decoder_layer.7.multi_head_attention.linear.weight": "model-00003-of-00019.safetensors",
|
| 743 |
-
"transformer.decoder_layer.7.multi_head_attention.query.weight": "model-00003-of-00019.safetensors",
|
| 744 |
-
"transformer.decoder_layer.7.multi_head_attention.value.weight": "model-00003-of-00019.safetensors",
|
| 745 |
-
"transformer.decoder_layer.7.rms_norm.weight": "model-00003-of-00019.safetensors",
|
| 746 |
-
"transformer.decoder_layer.7.rms_norm_1.weight": "model-00003-of-00019.safetensors",
|
| 747 |
-
"transformer.decoder_layer.7.rms_norm_2.weight": "model-00003-of-00019.safetensors",
|
| 748 |
-
"transformer.decoder_layer.7.rms_norm_3.weight": "model-00003-of-00019.safetensors",
|
| 749 |
-
"transformer.decoder_layer.7.router.weight": "model-00003-of-00019.safetensors",
|
| 750 |
-
"transformer.decoder_layer.8.moe.0.linear.weight": "model-00003-of-00019.safetensors",
|
| 751 |
-
"transformer.decoder_layer.8.moe.0.linear_1.weight": "model-00003-of-00019.safetensors",
|
| 752 |
-
"transformer.decoder_layer.8.moe.0.linear_v.weight": "model-00003-of-00019.safetensors",
|
| 753 |
-
"transformer.decoder_layer.8.multi_head_attention.key.weight": "model-00003-of-00019.safetensors",
|
| 754 |
-
"transformer.decoder_layer.8.multi_head_attention.linear.weight": "model-00003-of-00019.safetensors",
|
| 755 |
-
"transformer.decoder_layer.8.multi_head_attention.query.weight": "model-00003-of-00019.safetensors",
|
| 756 |
-
"transformer.decoder_layer.8.multi_head_attention.value.weight": "model-00003-of-00019.safetensors",
|
| 757 |
-
"transformer.decoder_layer.8.rms_norm.weight": "model-00003-of-00019.safetensors",
|
| 758 |
-
"transformer.decoder_layer.8.rms_norm_1.weight": "model-00003-of-00019.safetensors",
|
| 759 |
-
"transformer.decoder_layer.8.rms_norm_2.weight": "model-00003-of-00019.safetensors",
|
| 760 |
-
"transformer.decoder_layer.8.rms_norm_3.weight": "model-00003-of-00019.safetensors",
|
| 761 |
-
"transformer.decoder_layer.8.router.weight": "model-00003-of-00019.safetensors",
|
| 762 |
-
"transformer.decoder_layer.9.moe.0.linear.weight": "model-00004-of-00019.safetensors",
|
| 763 |
-
"transformer.decoder_layer.9.moe.0.linear_1.weight": "model-00004-of-00019.safetensors",
|
| 764 |
-
"transformer.decoder_layer.9.moe.0.linear_v.weight": "model-00003-of-00019.safetensors",
|
| 765 |
-
"transformer.decoder_layer.9.multi_head_attention.key.weight": "model-00003-of-00019.safetensors",
|
| 766 |
-
"transformer.decoder_layer.9.multi_head_attention.linear.weight": "model-00003-of-00019.safetensors",
|
| 767 |
-
"transformer.decoder_layer.9.multi_head_attention.query.weight": "model-00003-of-00019.safetensors",
|
| 768 |
-
"transformer.decoder_layer.9.multi_head_attention.value.weight": "model-00003-of-00019.safetensors",
|
| 769 |
-
"transformer.decoder_layer.9.rms_norm.weight": "model-00004-of-00019.safetensors",
|
| 770 |
-
"transformer.decoder_layer.9.rms_norm_1.weight": "model-00004-of-00019.safetensors",
|
| 771 |
-
"transformer.decoder_layer.9.rms_norm_2.weight": "model-00004-of-00019.safetensors",
|
| 772 |
-
"transformer.decoder_layer.9.rms_norm_3.weight": "model-00004-of-00019.safetensors",
|
| 773 |
-
"transformer.decoder_layer.9.router.weight": "model-00003-of-00019.safetensors",
|
| 774 |
-
"transformer.in_out_embed.weight": "model-00001-of-00019.safetensors",
|
| 775 |
-
"transformer.rms_norm.weight": "model-00019-of-00019.safetensors"
|
| 776 |
-
}
|
| 777 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
modeling_grok.py
DELETED
|
@@ -1,838 +0,0 @@
|
|
| 1 |
-
# coding=utf-8
|
| 2 |
-
# Modified from https://raw.githubusercontent.com/huggingface/transformers/v4.38.2/src/transformers/models/mixtral/modeling_mixtral.py
|
| 3 |
-
# Copyright 2023 Mistral AI and the HuggingFace Inc. team. All rights reserved.
|
| 4 |
-
#
|
| 5 |
-
# This code is based on EleutherAI's GPT-NeoX library and the GPT-NeoX
|
| 6 |
-
# and OPT implementations in this library. It has been modified from its
|
| 7 |
-
# original forms to accommodate minor architectural differences compared
|
| 8 |
-
# to GPT-NeoX and OPT used by the Meta AI team that trained the model.
|
| 9 |
-
#
|
| 10 |
-
# Licensed under the Apache License, Version 2.0 (the "License");
|
| 11 |
-
# you may not use this file except in compliance with the License.
|
| 12 |
-
# You may obtain a copy of the License at
|
| 13 |
-
#
|
| 14 |
-
# http://www.apache.org/licenses/LICENSE-2.0
|
| 15 |
-
#
|
| 16 |
-
# Unless required by applicable law or agreed to in writing, software
|
| 17 |
-
# distributed under the License is distributed on an "AS IS" BASIS,
|
| 18 |
-
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
| 19 |
-
# See the License for the specific language governing permissions and
|
| 20 |
-
# limitations under the License.
|
| 21 |
-
""" PyTorch Grok-1 model."""
|
| 22 |
-
import inspect
|
| 23 |
-
import math
|
| 24 |
-
import warnings
|
| 25 |
-
from typing import List, Optional, Tuple, Union
|
| 26 |
-
|
| 27 |
-
import torch
|
| 28 |
-
import torch.nn.functional as F
|
| 29 |
-
import torch.utils.checkpoint
|
| 30 |
-
from torch import nn
|
| 31 |
-
from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
|
| 32 |
-
|
| 33 |
-
from transformers.activations import ACT2FN
|
| 34 |
-
from transformers.cache_utils import Cache, DynamicCache
|
| 35 |
-
from transformers.modeling_attn_mask_utils import (
|
| 36 |
-
_prepare_4d_causal_attention_mask,
|
| 37 |
-
)
|
| 38 |
-
from transformers.modeling_outputs import (
|
| 39 |
-
MoeCausalLMOutputWithPast,
|
| 40 |
-
MoeModelOutputWithPast,
|
| 41 |
-
SequenceClassifierOutputWithPast,
|
| 42 |
-
)
|
| 43 |
-
from transformers.modeling_utils import PreTrainedModel
|
| 44 |
-
from transformers.pytorch_utils import is_torch_greater_or_equal_than_1_13
|
| 45 |
-
from transformers.utils import (
|
| 46 |
-
add_start_docstrings,
|
| 47 |
-
add_start_docstrings_to_model_forward,
|
| 48 |
-
logging,
|
| 49 |
-
replace_return_docstrings,
|
| 50 |
-
)
|
| 51 |
-
from transformers.utils.import_utils import is_torch_fx_available
|
| 52 |
-
from .configuration_grok import GrokConfig
|
| 53 |
-
|
| 54 |
-
|
| 55 |
-
# This makes `_prepare_4d_causal_attention_mask` a leaf function in the FX graph.
|
| 56 |
-
# It means that the function will not be traced through and simply appear as a node in the graph.
|
| 57 |
-
if is_torch_fx_available():
|
| 58 |
-
if not is_torch_greater_or_equal_than_1_13:
|
| 59 |
-
import torch.fx
|
| 60 |
-
|
| 61 |
-
_prepare_4d_causal_attention_mask = torch.fx.wrap(_prepare_4d_causal_attention_mask)
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
logger = logging.get_logger(__name__)
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
# Copied from transformers.models.llama.modeling_llama._get_unpad_data
|
| 68 |
-
def _get_unpad_data(attention_mask):
|
| 69 |
-
seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
|
| 70 |
-
indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
|
| 71 |
-
max_seqlen_in_batch = seqlens_in_batch.max().item()
|
| 72 |
-
cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.int32), (1, 0))
|
| 73 |
-
return (
|
| 74 |
-
indices,
|
| 75 |
-
cu_seqlens,
|
| 76 |
-
max_seqlen_in_batch,
|
| 77 |
-
)
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
# Copied from transformers.models.llama.modeling_llama.LlamaRMSNorm with Llama->Grok
|
| 81 |
-
class GrokRMSNorm(nn.Module):
|
| 82 |
-
def __init__(self, hidden_size, eps=1e-6):
|
| 83 |
-
"""
|
| 84 |
-
GrokRMSNorm is equivalent to T5LayerNorm
|
| 85 |
-
"""
|
| 86 |
-
super().__init__()
|
| 87 |
-
self.weight = nn.Parameter(torch.ones(hidden_size, dtype=torch.float32))
|
| 88 |
-
self.variance_epsilon = eps
|
| 89 |
-
|
| 90 |
-
def forward(self, hidden_states):
|
| 91 |
-
input_dtype = hidden_states.dtype
|
| 92 |
-
hidden_states = hidden_states.to(torch.float32)
|
| 93 |
-
variance = hidden_states.pow(2).mean(-1, keepdim=True)
|
| 94 |
-
hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)
|
| 95 |
-
return (self.weight * hidden_states).to(input_dtype)
|
| 96 |
-
|
| 97 |
-
|
| 98 |
-
# Copied from transformers.models.mistral.modeling_mistral.MistralRotaryEmbedding with Mistral->Grok
|
| 99 |
-
class GrokRotaryEmbedding(nn.Module):
|
| 100 |
-
def __init__(self, dim, max_position_embeddings=2048, base=10000, device=None):
|
| 101 |
-
super().__init__()
|
| 102 |
-
|
| 103 |
-
self.dim = dim
|
| 104 |
-
self.max_position_embeddings = max_position_embeddings
|
| 105 |
-
self.base = base
|
| 106 |
-
inv_freq = 1.0 / (self.base ** (torch.arange(0, self.dim, 2, dtype=torch.int64).float().to(device) / self.dim))
|
| 107 |
-
self.register_buffer("inv_freq", inv_freq, persistent=False)
|
| 108 |
-
|
| 109 |
-
# Build here to make `torch.jit.trace` work.
|
| 110 |
-
self._set_cos_sin_cache(
|
| 111 |
-
seq_len=max_position_embeddings, device=self.inv_freq.device, dtype=torch.get_default_dtype()
|
| 112 |
-
)
|
| 113 |
-
|
| 114 |
-
def _set_cos_sin_cache(self, seq_len, device, dtype):
|
| 115 |
-
self.max_seq_len_cached = seq_len
|
| 116 |
-
t = torch.arange(self.max_seq_len_cached, device=device, dtype=torch.int64).type_as(self.inv_freq)
|
| 117 |
-
|
| 118 |
-
freqs = torch.outer(t, self.inv_freq)
|
| 119 |
-
# Different from paper, but it uses a different permutation in order to obtain the same calculation
|
| 120 |
-
emb = torch.cat((freqs, freqs), dim=-1)
|
| 121 |
-
self.register_buffer("cos_cached", emb.cos().to(dtype), persistent=False)
|
| 122 |
-
self.register_buffer("sin_cached", emb.sin().to(dtype), persistent=False)
|
| 123 |
-
|
| 124 |
-
def forward(self, x, seq_len=None):
|
| 125 |
-
# x: [bs, num_attention_heads, seq_len, head_size]
|
| 126 |
-
if seq_len > self.max_seq_len_cached:
|
| 127 |
-
self._set_cos_sin_cache(seq_len=seq_len, device=x.device, dtype=x.dtype)
|
| 128 |
-
|
| 129 |
-
return (
|
| 130 |
-
self.cos_cached[:seq_len].to(dtype=x.dtype),
|
| 131 |
-
self.sin_cached[:seq_len].to(dtype=x.dtype),
|
| 132 |
-
)
|
| 133 |
-
|
| 134 |
-
|
| 135 |
-
# Copied from transformers.models.llama.modeling_llama.rotate_half
|
| 136 |
-
def rotate_half(x):
|
| 137 |
-
"""Rotates half the hidden dims of the input."""
|
| 138 |
-
x1 = x[..., : x.shape[-1] // 2]
|
| 139 |
-
x2 = x[..., x.shape[-1] // 2 :]
|
| 140 |
-
return torch.cat((-x2, x1), dim=-1)
|
| 141 |
-
|
| 142 |
-
|
| 143 |
-
# Copied from transformers.models.mistral.modeling_mistral.apply_rotary_pos_emb
|
| 144 |
-
def apply_rotary_pos_emb(q, k, cos, sin, position_ids, unsqueeze_dim=1):
|
| 145 |
-
"""Applies Rotary Position Embedding to the query and key tensors.
|
| 146 |
-
|
| 147 |
-
Args:
|
| 148 |
-
q (`torch.Tensor`): The query tensor.
|
| 149 |
-
k (`torch.Tensor`): The key tensor.
|
| 150 |
-
cos (`torch.Tensor`): The cosine part of the rotary embedding.
|
| 151 |
-
sin (`torch.Tensor`): The sine part of the rotary embedding.
|
| 152 |
-
position_ids (`torch.Tensor`):
|
| 153 |
-
The position indices of the tokens corresponding to the query and key tensors. For example, this can be
|
| 154 |
-
used to pass offsetted position ids when working with a KV-cache.
|
| 155 |
-
unsqueeze_dim (`int`, *optional*, defaults to 1):
|
| 156 |
-
The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and
|
| 157 |
-
sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note
|
| 158 |
-
that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and
|
| 159 |
-
k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes
|
| 160 |
-
cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have
|
| 161 |
-
the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.
|
| 162 |
-
Returns:
|
| 163 |
-
`tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding.
|
| 164 |
-
"""
|
| 165 |
-
cos = cos[position_ids].unsqueeze(unsqueeze_dim)
|
| 166 |
-
sin = sin[position_ids].unsqueeze(unsqueeze_dim)
|
| 167 |
-
q_embed = (q * cos) + (rotate_half(q) * sin)
|
| 168 |
-
k_embed = (k * cos) + (rotate_half(k) * sin)
|
| 169 |
-
return q_embed, k_embed
|
| 170 |
-
|
| 171 |
-
|
| 172 |
-
# Copied from transformers.models.llama.modeling_llama.repeat_kv
|
| 173 |
-
def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
|
| 174 |
-
"""
|
| 175 |
-
This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch,
|
| 176 |
-
num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim)
|
| 177 |
-
"""
|
| 178 |
-
batch, num_key_value_heads, slen, head_dim = hidden_states.shape
|
| 179 |
-
if n_rep == 1:
|
| 180 |
-
return hidden_states
|
| 181 |
-
hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)
|
| 182 |
-
return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)
|
| 183 |
-
|
| 184 |
-
|
| 185 |
-
class GrokAttention(nn.Module):
|
| 186 |
-
"""
|
| 187 |
-
Multi-headed attention from 'Attention Is All You Need' paper.
|
| 188 |
-
"""
|
| 189 |
-
|
| 190 |
-
def __init__(self, config: GrokConfig, layer_idx: Optional[int] = None):
|
| 191 |
-
super().__init__()
|
| 192 |
-
self.config = config
|
| 193 |
-
self.layer_idx = layer_idx
|
| 194 |
-
if layer_idx is None:
|
| 195 |
-
logger.warning_once(
|
| 196 |
-
f"Instantiating {self.__class__.__name__} without passing a `layer_idx` is not recommended and will "
|
| 197 |
-
"lead to errors during the forward call if caching is used. Please make sure to provide a `layer_idx` "
|
| 198 |
-
"when creating this class."
|
| 199 |
-
)
|
| 200 |
-
|
| 201 |
-
self.hidden_size = config.hidden_size
|
| 202 |
-
self.num_heads = config.num_attention_heads
|
| 203 |
-
self.head_dim = self.hidden_size // self.num_heads
|
| 204 |
-
self.num_key_value_heads = config.num_key_value_heads
|
| 205 |
-
self.num_key_value_groups = self.num_heads // self.num_key_value_heads
|
| 206 |
-
self.max_position_embeddings = config.max_position_embeddings
|
| 207 |
-
self.rope_theta = config.rope_theta
|
| 208 |
-
self.attn_output_multiplier = config.attn_output_multiplier
|
| 209 |
-
self.is_causal = True
|
| 210 |
-
self.attention_dropout = config.attention_dropout
|
| 211 |
-
|
| 212 |
-
if (self.head_dim * self.num_heads) != self.hidden_size:
|
| 213 |
-
raise ValueError(
|
| 214 |
-
f"hidden_size must be divisible by num_heads (got `hidden_size`: {self.hidden_size}"
|
| 215 |
-
f" and `num_heads`: {self.num_heads})."
|
| 216 |
-
)
|
| 217 |
-
self.query = nn.Linear(self.hidden_size, self.num_heads * self.head_dim, bias=False)
|
| 218 |
-
self.key = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False)
|
| 219 |
-
self.value = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False)
|
| 220 |
-
self.linear = nn.Linear(self.num_heads * self.head_dim, self.hidden_size, bias=False)
|
| 221 |
-
|
| 222 |
-
self.rotary_emb = GrokRotaryEmbedding(
|
| 223 |
-
self.head_dim,
|
| 224 |
-
max_position_embeddings=self.max_position_embeddings,
|
| 225 |
-
base=self.rope_theta,
|
| 226 |
-
)
|
| 227 |
-
|
| 228 |
-
def _shape(self, tensor: torch.Tensor, seq_len: int, bsz: int):
|
| 229 |
-
return tensor.view(bsz, seq_len, self.num_heads, self.head_dim).transpose(1, 2).contiguous()
|
| 230 |
-
|
| 231 |
-
def forward(
|
| 232 |
-
self,
|
| 233 |
-
hidden_states: torch.Tensor,
|
| 234 |
-
attention_mask: Optional[torch.Tensor] = None,
|
| 235 |
-
position_ids: Optional[torch.LongTensor] = None,
|
| 236 |
-
past_key_value: Optional[Cache] = None,
|
| 237 |
-
output_attentions: bool = False,
|
| 238 |
-
use_cache: bool = False,
|
| 239 |
-
**kwargs,
|
| 240 |
-
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
|
| 241 |
-
if "padding_mask" in kwargs:
|
| 242 |
-
warnings.warn(
|
| 243 |
-
"Passing `padding_mask` is deprecated and will be removed in v4.37. Please make sure use `attention_mask` instead.`"
|
| 244 |
-
)
|
| 245 |
-
bsz, q_len, _ = hidden_states.size()
|
| 246 |
-
|
| 247 |
-
query_states = self.query(hidden_states)
|
| 248 |
-
key_states = self.key(hidden_states)
|
| 249 |
-
value_states = self.value(hidden_states)
|
| 250 |
-
|
| 251 |
-
query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
|
| 252 |
-
key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
|
| 253 |
-
value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
|
| 254 |
-
|
| 255 |
-
kv_seq_len = key_states.shape[-2]
|
| 256 |
-
if past_key_value is not None:
|
| 257 |
-
if self.layer_idx is None:
|
| 258 |
-
raise ValueError(
|
| 259 |
-
f"The cache structure has changed since version v4.36. If you are using {self.__class__.__name__} "
|
| 260 |
-
"for auto-regressive decoding with k/v caching, please make sure to initialize the attention class "
|
| 261 |
-
"with a layer index."
|
| 262 |
-
)
|
| 263 |
-
kv_seq_len += past_key_value.get_usable_length(kv_seq_len, self.layer_idx)
|
| 264 |
-
cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
|
| 265 |
-
query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids)
|
| 266 |
-
|
| 267 |
-
if past_key_value is not None:
|
| 268 |
-
cache_kwargs = {"sin": sin, "cos": cos} # Specific to RoPE models
|
| 269 |
-
key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
|
| 270 |
-
|
| 271 |
-
# repeat k/v heads if n_kv_heads < n_heads
|
| 272 |
-
key_states = repeat_kv(key_states, self.num_key_value_groups)
|
| 273 |
-
value_states = repeat_kv(value_states, self.num_key_value_groups)
|
| 274 |
-
|
| 275 |
-
attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) * self.attn_output_multiplier
|
| 276 |
-
attn_weights = 30 * torch.tanh(attn_weights / 30)
|
| 277 |
-
|
| 278 |
-
if attn_weights.size() != (bsz, self.num_heads, q_len, kv_seq_len):
|
| 279 |
-
raise ValueError(
|
| 280 |
-
f"Attention weights should be of size {(bsz, self.num_heads, q_len, kv_seq_len)}, but is"
|
| 281 |
-
f" {attn_weights.size()}"
|
| 282 |
-
)
|
| 283 |
-
|
| 284 |
-
if attention_mask is not None:
|
| 285 |
-
if attention_mask.size() != (bsz, 1, q_len, kv_seq_len):
|
| 286 |
-
raise ValueError(
|
| 287 |
-
f"Attention mask should be of size {(bsz, 1, q_len, kv_seq_len)}, but is {attention_mask.size()}"
|
| 288 |
-
)
|
| 289 |
-
|
| 290 |
-
attn_weights = attn_weights + attention_mask
|
| 291 |
-
|
| 292 |
-
# upcast attention to fp32
|
| 293 |
-
attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query_states.dtype)
|
| 294 |
-
attn_weights = nn.functional.dropout(attn_weights, p=self.attention_dropout, training=self.training)
|
| 295 |
-
attn_output = torch.matmul(attn_weights, value_states)
|
| 296 |
-
|
| 297 |
-
if attn_output.size() != (bsz, self.num_heads, q_len, self.head_dim):
|
| 298 |
-
raise ValueError(
|
| 299 |
-
f"`attn_output` should be of size {(bsz, self.num_heads, q_len, self.head_dim)}, but is"
|
| 300 |
-
f" {attn_output.size()}"
|
| 301 |
-
)
|
| 302 |
-
|
| 303 |
-
attn_output = attn_output.transpose(1, 2).contiguous()
|
| 304 |
-
attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
|
| 305 |
-
|
| 306 |
-
attn_output = self.linear(attn_output)
|
| 307 |
-
|
| 308 |
-
if not output_attentions:
|
| 309 |
-
attn_weights = None
|
| 310 |
-
|
| 311 |
-
return attn_output, attn_weights, past_key_value
|
| 312 |
-
|
| 313 |
-
|
| 314 |
-
class GrokBlockSparseTop2MLP(nn.Module):
|
| 315 |
-
def __init__(self, config: GrokConfig):
|
| 316 |
-
super().__init__()
|
| 317 |
-
self.ffn_dim = config.intermediate_size
|
| 318 |
-
self.hidden_dim = config.hidden_size
|
| 319 |
-
|
| 320 |
-
self.linear_v = nn.Linear(self.hidden_dim, self.ffn_dim, bias=False)
|
| 321 |
-
self.linear_1 = nn.Linear(self.ffn_dim, self.hidden_dim, bias=False)
|
| 322 |
-
self.linear = nn.Linear(self.hidden_dim, self.ffn_dim, bias=False)
|
| 323 |
-
|
| 324 |
-
self.act_fn = ACT2FN[config.hidden_act]
|
| 325 |
-
|
| 326 |
-
def forward(self, hidden_states):
|
| 327 |
-
current_hidden_states = self.act_fn(self.linear(hidden_states)) * self.linear_v(hidden_states)
|
| 328 |
-
current_hidden_states = self.linear_1(current_hidden_states)
|
| 329 |
-
return current_hidden_states
|
| 330 |
-
|
| 331 |
-
|
| 332 |
-
class GrokDecoderLayer(nn.Module):
|
| 333 |
-
def __init__(self, config: GrokConfig, layer_idx: int):
|
| 334 |
-
super().__init__()
|
| 335 |
-
self.hidden_size = config.hidden_size
|
| 336 |
-
self.ffn_dim = config.intermediate_size
|
| 337 |
-
self.num_experts = config.num_local_experts
|
| 338 |
-
self.top_k = config.num_experts_per_tok
|
| 339 |
-
|
| 340 |
-
self.multi_head_attention = GrokAttention(config, layer_idx)
|
| 341 |
-
self.router = nn.Linear(self.hidden_size, self.num_experts, dtype=torch.float32, bias=False)
|
| 342 |
-
self.moe = nn.ModuleList([GrokBlockSparseTop2MLP(config) for _ in range(self.num_experts)])
|
| 343 |
-
|
| 344 |
-
self.rms_norm = GrokRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
|
| 345 |
-
self.rms_norm_1 = GrokRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
|
| 346 |
-
self.rms_norm_2 = GrokRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
|
| 347 |
-
self.rms_norm_3 = GrokRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
|
| 348 |
-
|
| 349 |
-
def forward(
|
| 350 |
-
self,
|
| 351 |
-
hidden_states: torch.Tensor,
|
| 352 |
-
attention_mask: Optional[torch.Tensor] = None,
|
| 353 |
-
position_ids: Optional[torch.LongTensor] = None,
|
| 354 |
-
past_key_value: Optional[Tuple[torch.Tensor]] = None,
|
| 355 |
-
output_attentions: Optional[bool] = False,
|
| 356 |
-
output_router_logits: Optional[bool] = False,
|
| 357 |
-
use_cache: Optional[bool] = False,
|
| 358 |
-
**kwargs,
|
| 359 |
-
) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]:
|
| 360 |
-
if "padding_mask" in kwargs:
|
| 361 |
-
warnings.warn(
|
| 362 |
-
"Passing `padding_mask` is deprecated and will be removed in v4.37. Please make sure use `attention_mask` instead.`"
|
| 363 |
-
)
|
| 364 |
-
"""
|
| 365 |
-
Args:
|
| 366 |
-
hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
|
| 367 |
-
attention_mask (`torch.FloatTensor`, *optional*): attention mask of size
|
| 368 |
-
`(batch, sequence_length)` where padding elements are indicated by 0.
|
| 369 |
-
past_key_value (`Tuple(torch.FloatTensor)`, *optional*): cached past key and value projection states
|
| 370 |
-
output_attentions (`bool`, *optional*):
|
| 371 |
-
Whether or not to return the attentions tensors of all attention layers. See `attentions` under
|
| 372 |
-
returned tensors for more detail.
|
| 373 |
-
output_router_logits (`bool`, *optional*):
|
| 374 |
-
Whether or not to return the logits of all the routers. They are useful for computing the router loss, and
|
| 375 |
-
should not be returned during inference.
|
| 376 |
-
use_cache (`bool`, *optional*):
|
| 377 |
-
If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding
|
| 378 |
-
(see `past_key_values`).
|
| 379 |
-
"""
|
| 380 |
-
|
| 381 |
-
residual = hidden_states
|
| 382 |
-
|
| 383 |
-
hidden_states = self.rms_norm(hidden_states)
|
| 384 |
-
|
| 385 |
-
# Self Attention
|
| 386 |
-
hidden_states, self_attn_weights, present_key_value = self.multi_head_attention(
|
| 387 |
-
hidden_states=hidden_states,
|
| 388 |
-
attention_mask=attention_mask,
|
| 389 |
-
position_ids=position_ids,
|
| 390 |
-
past_key_value=past_key_value,
|
| 391 |
-
output_attentions=output_attentions,
|
| 392 |
-
use_cache=use_cache,
|
| 393 |
-
)
|
| 394 |
-
hidden_states = residual + self.rms_norm_1(hidden_states)
|
| 395 |
-
|
| 396 |
-
# Fully Connected
|
| 397 |
-
residual = hidden_states
|
| 398 |
-
hidden_states = self.rms_norm_2(hidden_states)
|
| 399 |
-
|
| 400 |
-
batch_size, sequence_length, hidden_dim = hidden_states.shape
|
| 401 |
-
hidden_states = hidden_states.view(-1, hidden_dim)
|
| 402 |
-
# router_logits: (batch * sequence_length, n_experts)
|
| 403 |
-
router_logits = self.router(hidden_states.to(torch.float))
|
| 404 |
-
|
| 405 |
-
routing_weights = F.softmax(router_logits, dim=1, dtype=torch.float)
|
| 406 |
-
routing_weights, selected_experts = torch.topk(routing_weights, self.top_k, dim=-1)
|
| 407 |
-
# we cast back to the input dtype
|
| 408 |
-
routing_weights = routing_weights.to(hidden_states.dtype)
|
| 409 |
-
|
| 410 |
-
final_hidden_states = torch.zeros(
|
| 411 |
-
(batch_size * sequence_length, hidden_dim), dtype=hidden_states.dtype, device=hidden_states.device
|
| 412 |
-
)
|
| 413 |
-
|
| 414 |
-
# One hot encode the selected experts to create an expert mask
|
| 415 |
-
# this will be used to easily index which expert is going to be sollicitated
|
| 416 |
-
expert_mask = torch.nn.functional.one_hot(selected_experts, num_classes=self.num_experts).permute(2, 1, 0)
|
| 417 |
-
|
| 418 |
-
# Loop over all available experts in the model and perform the computation on each expert
|
| 419 |
-
for expert_idx in range(self.num_experts):
|
| 420 |
-
expert_layer = self.moe[expert_idx]
|
| 421 |
-
idx, top_x = torch.where(expert_mask[expert_idx])
|
| 422 |
-
|
| 423 |
-
if top_x.shape[0] == 0:
|
| 424 |
-
continue
|
| 425 |
-
|
| 426 |
-
# in torch it is faster to index using lists than torch tensors
|
| 427 |
-
top_x_list = top_x.tolist()
|
| 428 |
-
idx_list = idx.tolist()
|
| 429 |
-
|
| 430 |
-
# Index the correct hidden states and compute the expert hidden state for
|
| 431 |
-
# the current expert. We need to make sure to multiply the output hidden
|
| 432 |
-
# states by `routing_weights` on the corresponding tokens (top-1 and top-2)
|
| 433 |
-
current_state = hidden_states[None, top_x_list].reshape(-1, hidden_dim)
|
| 434 |
-
current_hidden_states = expert_layer(current_state) * routing_weights[top_x_list, idx_list, None]
|
| 435 |
-
|
| 436 |
-
# However `index_add_` only support torch tensors for indexing so we'll use
|
| 437 |
-
# the `top_x` tensor here.
|
| 438 |
-
final_hidden_states.index_add_(0, top_x, current_hidden_states.to(hidden_states.dtype))
|
| 439 |
-
hidden_states = final_hidden_states.reshape(batch_size, sequence_length, hidden_dim)
|
| 440 |
-
|
| 441 |
-
hidden_states = residual + self.rms_norm_3(hidden_states)
|
| 442 |
-
|
| 443 |
-
outputs = (hidden_states,)
|
| 444 |
-
|
| 445 |
-
if output_attentions:
|
| 446 |
-
outputs += (self_attn_weights,)
|
| 447 |
-
|
| 448 |
-
if use_cache:
|
| 449 |
-
outputs += (present_key_value,)
|
| 450 |
-
|
| 451 |
-
if output_router_logits:
|
| 452 |
-
outputs += (router_logits,)
|
| 453 |
-
|
| 454 |
-
return outputs
|
| 455 |
-
|
| 456 |
-
|
| 457 |
-
# Copied from transformers.models.mistral.modeling_mistral.MistralPreTrainedModel with Mistral->Grok
|
| 458 |
-
class GrokPreTrainedModel(PreTrainedModel):
|
| 459 |
-
config_class = GrokConfig
|
| 460 |
-
base_model_prefix = "transformer"
|
| 461 |
-
supports_gradient_checkpointing = True
|
| 462 |
-
_no_split_modules = ["GrokDecoderLayer"]
|
| 463 |
-
_skip_keys_device_placement = "past_key_values"
|
| 464 |
-
_keys_to_ignore_on_load_missing = [r"lm_head.*."]
|
| 465 |
-
_supports_flash_attn_2 = False
|
| 466 |
-
_supports_sdpa = False
|
| 467 |
-
|
| 468 |
-
def _init_weights(self, module):
|
| 469 |
-
pass
|
| 470 |
-
|
| 471 |
-
|
| 472 |
-
# Copied from transformers.models.mistral.modeling_mistral.MistralModel with Mistral->Grok
|
| 473 |
-
class GrokModel(GrokPreTrainedModel):
|
| 474 |
-
"""
|
| 475 |
-
Transformer decoder consisting of *config.num_hidden_layers* layers. Each layer is a [`GrokDecoderLayer`]
|
| 476 |
-
|
| 477 |
-
Args:
|
| 478 |
-
config: GrokConfig
|
| 479 |
-
"""
|
| 480 |
-
|
| 481 |
-
def __init__(self, config: GrokConfig):
|
| 482 |
-
super().__init__(config)
|
| 483 |
-
self.padding_idx = config.pad_token_id
|
| 484 |
-
self.vocab_size = config.vocab_size
|
| 485 |
-
self.embedding_multiplier_scale = config.embedding_multiplier_scale
|
| 486 |
-
|
| 487 |
-
self.in_out_embed = nn.Embedding(config.vocab_size, config.hidden_size, self.padding_idx)
|
| 488 |
-
self.decoder_layer = nn.ModuleList(
|
| 489 |
-
[GrokDecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
|
| 490 |
-
)
|
| 491 |
-
self.rms_norm = GrokRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
|
| 492 |
-
|
| 493 |
-
self.gradient_checkpointing = False
|
| 494 |
-
# Initialize weights and apply final processing
|
| 495 |
-
self.post_init()
|
| 496 |
-
|
| 497 |
-
def get_input_embeddings(self):
|
| 498 |
-
return self.in_out_embed
|
| 499 |
-
|
| 500 |
-
def set_input_embeddings(self, value):
|
| 501 |
-
self.in_out_embed = value
|
| 502 |
-
|
| 503 |
-
# Ignore copy
|
| 504 |
-
def forward(
|
| 505 |
-
self,
|
| 506 |
-
input_ids: torch.LongTensor = None,
|
| 507 |
-
attention_mask: Optional[torch.Tensor] = None,
|
| 508 |
-
position_ids: Optional[torch.LongTensor] = None,
|
| 509 |
-
past_key_values: Optional[List[torch.FloatTensor]] = None,
|
| 510 |
-
inputs_embeds: Optional[torch.FloatTensor] = None,
|
| 511 |
-
use_cache: Optional[bool] = None,
|
| 512 |
-
output_attentions: Optional[bool] = None,
|
| 513 |
-
output_hidden_states: Optional[bool] = None,
|
| 514 |
-
output_router_logits: Optional[bool] = None,
|
| 515 |
-
return_dict: Optional[bool] = None,
|
| 516 |
-
) -> Union[Tuple, MoeModelOutputWithPast]:
|
| 517 |
-
output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
|
| 518 |
-
output_router_logits = (
|
| 519 |
-
output_router_logits if output_router_logits is not None else self.config.output_router_logits
|
| 520 |
-
)
|
| 521 |
-
output_hidden_states = (
|
| 522 |
-
output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
|
| 523 |
-
)
|
| 524 |
-
use_cache = use_cache if use_cache is not None else self.config.use_cache
|
| 525 |
-
|
| 526 |
-
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
|
| 527 |
-
|
| 528 |
-
# retrieve input_ids and inputs_embeds
|
| 529 |
-
if input_ids is not None and inputs_embeds is not None:
|
| 530 |
-
raise ValueError("You cannot specify both decoder_input_ids and decoder_inputs_embeds at the same time")
|
| 531 |
-
elif input_ids is not None:
|
| 532 |
-
batch_size, seq_length = input_ids.shape
|
| 533 |
-
elif inputs_embeds is not None:
|
| 534 |
-
batch_size, seq_length, _ = inputs_embeds.shape
|
| 535 |
-
else:
|
| 536 |
-
raise ValueError("You have to specify either decoder_input_ids or decoder_inputs_embeds")
|
| 537 |
-
|
| 538 |
-
past_key_values_length = 0
|
| 539 |
-
|
| 540 |
-
if self.gradient_checkpointing and self.training:
|
| 541 |
-
if use_cache:
|
| 542 |
-
logger.warning_once(
|
| 543 |
-
"`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..."
|
| 544 |
-
)
|
| 545 |
-
use_cache = False
|
| 546 |
-
|
| 547 |
-
if use_cache:
|
| 548 |
-
use_legacy_cache = not isinstance(past_key_values, Cache)
|
| 549 |
-
if use_legacy_cache:
|
| 550 |
-
past_key_values = DynamicCache.from_legacy_cache(past_key_values)
|
| 551 |
-
past_key_values_length = past_key_values.get_usable_length(seq_length)
|
| 552 |
-
|
| 553 |
-
if position_ids is None:
|
| 554 |
-
device = input_ids.device if input_ids is not None else inputs_embeds.device
|
| 555 |
-
position_ids = torch.arange(
|
| 556 |
-
past_key_values_length, seq_length + past_key_values_length, dtype=torch.long, device=device
|
| 557 |
-
)
|
| 558 |
-
position_ids = position_ids.unsqueeze(0).view(-1, seq_length)
|
| 559 |
-
else:
|
| 560 |
-
position_ids = position_ids.view(-1, seq_length).long()
|
| 561 |
-
|
| 562 |
-
if inputs_embeds is None:
|
| 563 |
-
inputs_embeds = self.in_out_embed(input_ids)
|
| 564 |
-
|
| 565 |
-
attention_mask = _prepare_4d_causal_attention_mask(
|
| 566 |
-
attention_mask,
|
| 567 |
-
(batch_size, seq_length),
|
| 568 |
-
inputs_embeds,
|
| 569 |
-
past_key_values_length,
|
| 570 |
-
)
|
| 571 |
-
|
| 572 |
-
hidden_states = inputs_embeds
|
| 573 |
-
hidden_states *= self.embedding_multiplier_scale
|
| 574 |
-
|
| 575 |
-
# decoder layers
|
| 576 |
-
all_hidden_states = () if output_hidden_states else None
|
| 577 |
-
all_self_attns = () if output_attentions else None
|
| 578 |
-
all_router_logits = () if output_router_logits else None
|
| 579 |
-
next_decoder_cache = None
|
| 580 |
-
|
| 581 |
-
for decoder_layer in self.decoder_layer:
|
| 582 |
-
if output_hidden_states:
|
| 583 |
-
all_hidden_states += (hidden_states,)
|
| 584 |
-
|
| 585 |
-
if self.gradient_checkpointing and self.training:
|
| 586 |
-
layer_outputs = self._gradient_checkpointing_func(
|
| 587 |
-
decoder_layer.__call__,
|
| 588 |
-
hidden_states,
|
| 589 |
-
attention_mask,
|
| 590 |
-
position_ids,
|
| 591 |
-
past_key_values,
|
| 592 |
-
output_attentions,
|
| 593 |
-
output_router_logits,
|
| 594 |
-
use_cache,
|
| 595 |
-
)
|
| 596 |
-
else:
|
| 597 |
-
layer_outputs = decoder_layer(
|
| 598 |
-
hidden_states,
|
| 599 |
-
attention_mask=attention_mask,
|
| 600 |
-
position_ids=position_ids,
|
| 601 |
-
past_key_value=past_key_values,
|
| 602 |
-
output_attentions=output_attentions,
|
| 603 |
-
output_router_logits=output_router_logits,
|
| 604 |
-
use_cache=use_cache,
|
| 605 |
-
)
|
| 606 |
-
|
| 607 |
-
hidden_states = layer_outputs[0]
|
| 608 |
-
|
| 609 |
-
if use_cache:
|
| 610 |
-
next_decoder_cache = layer_outputs[2 if output_attentions else 1]
|
| 611 |
-
|
| 612 |
-
if output_attentions:
|
| 613 |
-
all_self_attns += (layer_outputs[1],)
|
| 614 |
-
|
| 615 |
-
if output_router_logits:
|
| 616 |
-
all_router_logits += (layer_outputs[-1],)
|
| 617 |
-
|
| 618 |
-
hidden_states = self.rms_norm(hidden_states)
|
| 619 |
-
|
| 620 |
-
# add hidden states from the last decoder layer
|
| 621 |
-
if output_hidden_states:
|
| 622 |
-
all_hidden_states += (hidden_states,)
|
| 623 |
-
|
| 624 |
-
next_cache = None
|
| 625 |
-
if use_cache:
|
| 626 |
-
next_cache = next_decoder_cache.to_legacy_cache() if use_legacy_cache else next_decoder_cache
|
| 627 |
-
|
| 628 |
-
if not return_dict:
|
| 629 |
-
return tuple(
|
| 630 |
-
v
|
| 631 |
-
for v in [hidden_states, next_cache, all_hidden_states, all_self_attns, all_router_logits]
|
| 632 |
-
if v is not None
|
| 633 |
-
)
|
| 634 |
-
return MoeModelOutputWithPast(
|
| 635 |
-
last_hidden_state=hidden_states,
|
| 636 |
-
past_key_values=next_cache,
|
| 637 |
-
hidden_states=all_hidden_states,
|
| 638 |
-
attentions=all_self_attns,
|
| 639 |
-
router_logits=all_router_logits,
|
| 640 |
-
)
|
| 641 |
-
|
| 642 |
-
|
| 643 |
-
class GrokForCausalLM(GrokPreTrainedModel):
|
| 644 |
-
_tied_weights_keys = ["lm_head.weight"]
|
| 645 |
-
|
| 646 |
-
def __init__(self, config):
|
| 647 |
-
super().__init__(config)
|
| 648 |
-
self.transformer = GrokModel(config)
|
| 649 |
-
self.vocab_size = config.vocab_size
|
| 650 |
-
self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
|
| 651 |
-
self.router_aux_loss_coef = config.router_aux_loss_coef
|
| 652 |
-
self.num_experts = config.num_local_experts
|
| 653 |
-
self.num_experts_per_tok = config.num_experts_per_tok
|
| 654 |
-
self.output_multiplier_scale = config.output_multiplier_scale
|
| 655 |
-
# Initialize weights and apply final processing
|
| 656 |
-
self.post_init()
|
| 657 |
-
|
| 658 |
-
def get_input_embeddings(self):
|
| 659 |
-
return self.transformer.in_out_embed
|
| 660 |
-
|
| 661 |
-
def set_input_embeddings(self, value):
|
| 662 |
-
self.transformer.in_out_embed = value
|
| 663 |
-
|
| 664 |
-
def get_output_embeddings(self):
|
| 665 |
-
return self.lm_head
|
| 666 |
-
|
| 667 |
-
def set_output_embeddings(self, new_embeddings):
|
| 668 |
-
self.lm_head = new_embeddings
|
| 669 |
-
|
| 670 |
-
def set_decoder(self, decoder):
|
| 671 |
-
self.transformer = decoder
|
| 672 |
-
|
| 673 |
-
def get_decoder(self):
|
| 674 |
-
return self.transformer
|
| 675 |
-
|
| 676 |
-
def _tie_weights(self):
|
| 677 |
-
self._tie_or_clone_weights(self.lm_head, self.get_input_embeddings())
|
| 678 |
-
|
| 679 |
-
# Ignore copy
|
| 680 |
-
def forward(
|
| 681 |
-
self,
|
| 682 |
-
input_ids: torch.LongTensor = None,
|
| 683 |
-
attention_mask: Optional[torch.Tensor] = None,
|
| 684 |
-
position_ids: Optional[torch.LongTensor] = None,
|
| 685 |
-
past_key_values: Optional[List[torch.FloatTensor]] = None,
|
| 686 |
-
inputs_embeds: Optional[torch.FloatTensor] = None,
|
| 687 |
-
labels: Optional[torch.LongTensor] = None,
|
| 688 |
-
use_cache: Optional[bool] = None,
|
| 689 |
-
output_attentions: Optional[bool] = None,
|
| 690 |
-
output_hidden_states: Optional[bool] = None,
|
| 691 |
-
output_router_logits: Optional[bool] = None,
|
| 692 |
-
return_dict: Optional[bool] = None,
|
| 693 |
-
) -> Union[Tuple, MoeCausalLMOutputWithPast]:
|
| 694 |
-
r"""
|
| 695 |
-
Args:
|
| 696 |
-
labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
|
| 697 |
-
Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
|
| 698 |
-
config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
|
| 699 |
-
(masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`.
|
| 700 |
-
|
| 701 |
-
Returns:
|
| 702 |
-
|
| 703 |
-
"""
|
| 704 |
-
|
| 705 |
-
output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
|
| 706 |
-
output_router_logits = (
|
| 707 |
-
output_router_logits if output_router_logits is not None else self.config.output_router_logits
|
| 708 |
-
)
|
| 709 |
-
|
| 710 |
-
output_hidden_states = (
|
| 711 |
-
output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
|
| 712 |
-
)
|
| 713 |
-
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
|
| 714 |
-
|
| 715 |
-
# decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
|
| 716 |
-
outputs = self.transformer(
|
| 717 |
-
input_ids=input_ids,
|
| 718 |
-
attention_mask=attention_mask,
|
| 719 |
-
position_ids=position_ids,
|
| 720 |
-
past_key_values=past_key_values,
|
| 721 |
-
inputs_embeds=inputs_embeds,
|
| 722 |
-
use_cache=use_cache,
|
| 723 |
-
output_attentions=output_attentions,
|
| 724 |
-
output_hidden_states=output_hidden_states,
|
| 725 |
-
output_router_logits=output_router_logits,
|
| 726 |
-
return_dict=return_dict,
|
| 727 |
-
)
|
| 728 |
-
|
| 729 |
-
hidden_states = outputs[0]
|
| 730 |
-
logits = self.lm_head(hidden_states)
|
| 731 |
-
logits = logits * self.output_multiplier_scale
|
| 732 |
-
logits = logits.float()
|
| 733 |
-
|
| 734 |
-
loss = None
|
| 735 |
-
if labels is not None:
|
| 736 |
-
# Shift so that tokens < n predict n
|
| 737 |
-
shift_logits = logits[..., :-1, :].contiguous()
|
| 738 |
-
shift_labels = labels[..., 1:].contiguous()
|
| 739 |
-
# Flatten the tokens
|
| 740 |
-
loss_fct = CrossEntropyLoss()
|
| 741 |
-
shift_logits = shift_logits.view(-1, self.config.vocab_size)
|
| 742 |
-
shift_labels = shift_labels.view(-1)
|
| 743 |
-
# Enable model parallelism
|
| 744 |
-
shift_labels = shift_labels.to(shift_logits.device)
|
| 745 |
-
loss = loss_fct(shift_logits, shift_labels)
|
| 746 |
-
|
| 747 |
-
aux_loss = None
|
| 748 |
-
if output_router_logits:
|
| 749 |
-
aux_loss = load_balancing_loss_func(
|
| 750 |
-
outputs.router_logits if return_dict else outputs[-1],
|
| 751 |
-
self.num_experts,
|
| 752 |
-
self.num_experts_per_tok,
|
| 753 |
-
attention_mask,
|
| 754 |
-
)
|
| 755 |
-
if labels is not None:
|
| 756 |
-
loss += self.router_aux_loss_coef * aux_loss.to(loss.device) # make sure to reside in the same device
|
| 757 |
-
|
| 758 |
-
if not return_dict:
|
| 759 |
-
output = (logits,) + outputs[1:]
|
| 760 |
-
if output_router_logits:
|
| 761 |
-
output = (aux_loss,) + output
|
| 762 |
-
return (loss,) + output if loss is not None else output
|
| 763 |
-
|
| 764 |
-
return MoeCausalLMOutputWithPast(
|
| 765 |
-
loss=loss,
|
| 766 |
-
aux_loss=aux_loss,
|
| 767 |
-
logits=logits,
|
| 768 |
-
past_key_values=outputs.past_key_values,
|
| 769 |
-
hidden_states=outputs.hidden_states,
|
| 770 |
-
attentions=outputs.attentions,
|
| 771 |
-
router_logits=outputs.router_logits,
|
| 772 |
-
)
|
| 773 |
-
|
| 774 |
-
def prepare_inputs_for_generation(
|
| 775 |
-
self, input_ids, past_key_values=None, attention_mask=None, inputs_embeds=None, **kwargs
|
| 776 |
-
):
|
| 777 |
-
# Omit tokens covered by past_key_values
|
| 778 |
-
if past_key_values is not None:
|
| 779 |
-
if isinstance(past_key_values, Cache):
|
| 780 |
-
cache_length = past_key_values.get_seq_length()
|
| 781 |
-
past_length = past_key_values.seen_tokens
|
| 782 |
-
max_cache_length = past_key_values.get_max_length()
|
| 783 |
-
else:
|
| 784 |
-
cache_length = past_length = past_key_values[0][0].shape[2]
|
| 785 |
-
max_cache_length = None
|
| 786 |
-
|
| 787 |
-
# Keep only the unprocessed tokens:
|
| 788 |
-
# 1 - If the length of the attention_mask exceeds the length of input_ids, then we are in a setting where
|
| 789 |
-
# some of the inputs are exclusively passed as part of the cache (e.g. when passing input_embeds as
|
| 790 |
-
# input)
|
| 791 |
-
if attention_mask is not None and attention_mask.shape[1] > input_ids.shape[1]:
|
| 792 |
-
input_ids = input_ids[:, -(attention_mask.shape[1] - past_length) :]
|
| 793 |
-
# 2 - If the past_length is smaller than input_ids', then input_ids holds all input tokens. We can discard
|
| 794 |
-
# input_ids based on the past_length.
|
| 795 |
-
elif past_length < input_ids.shape[1]:
|
| 796 |
-
input_ids = input_ids[:, past_length:]
|
| 797 |
-
# 3 - Otherwise (past_length >= input_ids.shape[1]), let's assume input_ids only has unprocessed tokens.
|
| 798 |
-
|
| 799 |
-
# If we are about to go beyond the maximum cache length, we need to crop the input attention mask.
|
| 800 |
-
if (
|
| 801 |
-
max_cache_length is not None
|
| 802 |
-
and attention_mask is not None
|
| 803 |
-
and cache_length + input_ids.shape[1] > max_cache_length
|
| 804 |
-
):
|
| 805 |
-
attention_mask = attention_mask[:, -max_cache_length:]
|
| 806 |
-
|
| 807 |
-
position_ids = kwargs.get("position_ids", None)
|
| 808 |
-
if attention_mask is not None and position_ids is None:
|
| 809 |
-
# create position_ids on the fly for batch generation
|
| 810 |
-
position_ids = attention_mask.long().cumsum(-1) - 1
|
| 811 |
-
position_ids.masked_fill_(attention_mask == 0, 1)
|
| 812 |
-
if past_key_values:
|
| 813 |
-
position_ids = position_ids[:, -input_ids.shape[1] :]
|
| 814 |
-
|
| 815 |
-
# if `inputs_embeds` are passed, we only want to use them in the 1st generation step
|
| 816 |
-
if inputs_embeds is not None and past_key_values is None:
|
| 817 |
-
model_inputs = {"inputs_embeds": inputs_embeds}
|
| 818 |
-
else:
|
| 819 |
-
model_inputs = {"input_ids": input_ids}
|
| 820 |
-
|
| 821 |
-
model_inputs.update(
|
| 822 |
-
{
|
| 823 |
-
"position_ids": position_ids,
|
| 824 |
-
"past_key_values": past_key_values,
|
| 825 |
-
"use_cache": kwargs.get("use_cache"),
|
| 826 |
-
"attention_mask": attention_mask,
|
| 827 |
-
}
|
| 828 |
-
)
|
| 829 |
-
return model_inputs
|
| 830 |
-
|
| 831 |
-
@staticmethod
|
| 832 |
-
def _reorder_cache(past_key_values, beam_idx):
|
| 833 |
-
reordered_past = ()
|
| 834 |
-
for layer_past in past_key_values:
|
| 835 |
-
reordered_past += (
|
| 836 |
-
tuple(past_state.index_select(0, beam_idx.to(past_state.device)) for past_state in layer_past),
|
| 837 |
-
)
|
| 838 |
-
return reordered_past
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
pytorch_model-00001-of-00019.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:49bf50ef3d70afd8e7d199223f8472bc7296f2407ec2f43809ca62276ebaa9a9
|
| 3 |
-
size 24688084712
|
|
|
|
|
|
|
|
|
|
|
|
pytorch_model-00002-of-00019.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:cd2e170b63a6532af4f1e7cde4fb51c34bd810aaa0ebbf33345f36f1d08fe9ac
|
| 3 |
-
size 35962479774
|
|
|
|
|
|
|
|
|
|
|
|
pytorch_model-00003-of-00019.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:abb2c6f340f4d0f122c92479f0c2e9324d8f415af1bed4a645e6a374dd7693c3
|
| 3 |
-
size 32917463282
|
|
|
|
|
|
|
|
|
|
|
|
pytorch_model-00004-of-00019.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:b9fdff65e71dc305f23181d18ea28b35c5b740f3d624a6d7e92495ccb7cb7a07
|
| 3 |
-
size 35962479902
|
|
|
|
|
|
|
|
|
|
|
|
pytorch_model-00005-of-00019.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:88ec608b37942f333a7274017d4d7ef47e8c420ec3d1c251a0002e08da59a852
|
| 3 |
-
size 32917463410
|
|
|
|
|
|
|
|
|
|
|
|
pytorch_model-00006-of-00019.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:52ad50ece14037cc7d7165d720b02c19619e45d354ff1f46b6b568975106e1a2
|
| 3 |
-
size 35962479902
|
|
|
|
|
|
|
|
|
|
|
|
pytorch_model-00007-of-00019.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:15b5b209f1c94aed54d67b5bccd67be3696c042ce8af4d04317a834aa70c9feb
|
| 3 |
-
size 32917463410
|
|
|
|
|
|
|
|
|
|
|
|
pytorch_model-00008-of-00019.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:1a78713c58d353f9835ac0913f80e45c57aa34a93813bba937e39ac37ed09c46
|
| 3 |
-
size 35962479902
|
|
|
|
|
|
|
|
|
|
|
|
pytorch_model-00009-of-00019.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:ccaa848a8f0f1245019b69e8bf883f51d3a36ba9f251813cd48c28fca597b1c1
|
| 3 |
-
size 32917463410
|
|
|
|
|
|
|
|
|
|
|
|
pytorch_model-00010-of-00019.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:0e11fc345ffdf8cb429deeca6ad79ebbf77a4684e845e6857a4003235c6e074e
|
| 3 |
-
size 35962479902
|
|
|
|
|
|
|
|
|
|
|
|
pytorch_model-00011-of-00019.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:8d180e294abda6722fe27332a1d7311f6256c490db667ce8041345fd912f06e2
|
| 3 |
-
size 32917463410
|
|
|
|
|
|
|
|
|
|
|
|
pytorch_model-00012-of-00019.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:aaaeeece29933f8ed8e3bd2ba431f80ee692f525b0a33ff9fbc9e4e319c354d4
|
| 3 |
-
size 35962479902
|
|
|
|
|
|
|
|
|
|
|
|
pytorch_model-00013-of-00019.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:1dd78618c0245d6388b13a88eba6577b958a10cd88342a060c6ca9ad559c0feb
|
| 3 |
-
size 32917463410
|
|
|
|
|
|
|
|
|
|
|
|
pytorch_model-00014-of-00019.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:0a0c17526baf9136881fd3438bf347e527bcdbb6c014bc25e2fea8d621ee85a6
|
| 3 |
-
size 35962479902
|
|
|
|
|
|
|
|
|
|
|
|
pytorch_model-00015-of-00019.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:af1b44d316af3cb0d66e4e2f3703e0956a70b77b0f1754cea611aea499ec25a9
|
| 3 |
-
size 32917463410
|
|
|
|
|
|
|
|
|
|
|
|
pytorch_model-00016-of-00019.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:3082fac8e8f9a2d0271b1a88ce1812b6ef41ff40cdfca7e4d9ce16d0241e36b1
|
| 3 |
-
size 35962479902
|
|
|
|
|
|
|
|
|
|
|
|
pytorch_model-00017-of-00019.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:931d80d3cc5157b219ccaaa3c5b6ab5c3996300162cc6f6cf27780f2f0de7262
|
| 3 |
-
size 32917463410
|
|
|
|
|
|
|
|
|
|
|
|
pytorch_model-00018-of-00019.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:74d3dfe9067c0185982e8a6fcf91612da6e118782eb92849452289af1bc75d06
|
| 3 |
-
size 35962479902
|
|
|
|
|
|
|
|
|
|
|
|
pytorch_model-00019-of-00019.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:adf12c09aae1a83384d945c440524cd8e2e3ce4dffb83bad3f96857a8a55a770
|
| 3 |
-
size 19679996468
|
|
|
|
|
|
|
|
|
|
|
|
pytorch_model.bin.index.json
DELETED
|
The diff for this file is too large to render.
See raw diff
|
|
|
special_tokens_map.json
DELETED
|
@@ -1,23 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"bos_token": {
|
| 3 |
-
"content": "<s>",
|
| 4 |
-
"lstrip": false,
|
| 5 |
-
"normalized": false,
|
| 6 |
-
"rstrip": false,
|
| 7 |
-
"single_word": false
|
| 8 |
-
},
|
| 9 |
-
"eos_token": {
|
| 10 |
-
"content": "</s>",
|
| 11 |
-
"lstrip": false,
|
| 12 |
-
"normalized": false,
|
| 13 |
-
"rstrip": false,
|
| 14 |
-
"single_word": false
|
| 15 |
-
},
|
| 16 |
-
"unk_token": {
|
| 17 |
-
"content": "<unk>",
|
| 18 |
-
"lstrip": false,
|
| 19 |
-
"normalized": false,
|
| 20 |
-
"rstrip": false,
|
| 21 |
-
"single_word": false
|
| 22 |
-
}
|
| 23 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
tokenizer.json
DELETED
|
The diff for this file is too large to render.
See raw diff
|
|
|
tokenizer.model
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:5c69d7cbad192fa2c9d14e2a77fbfdc11597c68907b043ddb0260e3d28eddd7f
|
| 3 |
-
size 2229219
|
|
|
|
|
|
|
|
|
|
|
|