x54-729 commited on
Commit
659ed91
1 Parent(s): 14e30d3

Upload internlm-chat-7b

Browse files
config.json ADDED
@@ -0,0 +1,27 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "InternLMForCausalLM"
4
+ ],
5
+ "auto_map": {
6
+ "AutoConfig": "configuration_internlm.InternLMConfig",
7
+ "AutoModelForCausalLM": "modeling_internlm.InternLMForCausalLM"
8
+ },
9
+ "bias": true,
10
+ "bos_token_id": 1,
11
+ "eos_token_id": 2,
12
+ "hidden_act": "silu",
13
+ "hidden_size": 4096,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 11008,
16
+ "max_position_embeddings": 2048,
17
+ "model_type": "internlm",
18
+ "num_attention_heads": 32,
19
+ "num_hidden_layers": 32,
20
+ "pad_token_id": 0,
21
+ "rms_norm_eps": 1e-06,
22
+ "tie_word_embeddings": false,
23
+ "torch_dtype": "bfloat16",
24
+ "transformers_version": "4.29.2",
25
+ "use_cache": true,
26
+ "vocab_size": 103168
27
+ }
configuration_internlm.py ADDED
@@ -0,0 +1,120 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # coding=utf-8
2
+ # Copyright 2022 EleutherAI and the HuggingFace Inc. team. All rights reserved.
3
+ #
4
+ # This code is based on EleutherAI's GPT-NeoX library and the GPT-NeoX
5
+ # and OPT implementations in this library. It has been modified from its
6
+ # original forms to accommodate minor architectural differences compared
7
+ # to GPT-NeoX and OPT used by the Meta AI team that trained the model.
8
+ #
9
+ # Licensed under the Apache License, Version 2.0 (the "License");
10
+ # you may not use this file except in compliance with the License.
11
+ # You may obtain a copy of the License at
12
+ #
13
+ # http://www.apache.org/licenses/LICENSE-2.0
14
+ #
15
+ # Unless required by applicable law or agreed to in writing, software
16
+ # distributed under the License is distributed on an "AS IS" BASIS,
17
+ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
18
+ # See the License for the specific language governing permissions and
19
+ # limitations under the License.
20
+ """ InternLM model configuration"""
21
+
22
+ from transformers.utils import logging
23
+ from transformers.configuration_utils import PretrainedConfig
24
+
25
+
26
+ logger = logging.get_logger(__name__)
27
+
28
+ INTERNLM_PRETRAINED_CONFIG_ARCHIVE_MAP = {}
29
+
30
+
31
+ class InternLMConfig(PretrainedConfig):
32
+ r"""
33
+ This is the configuration class to store the configuration of a [`InternLMModel`]. It is used to instantiate an InternLM
34
+ model according to the specified arguments, defining the model architecture. Instantiating a configuration with the
35
+ defaults will yield a similar configuration to that of the InternLM-7B.
36
+
37
+ Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
38
+ documentation from [`PretrainedConfig`] for more information.
39
+
40
+
41
+ Args:
42
+ vocab_size (`int`, *optional*, defaults to 32000):
43
+ Vocabulary size of the InternLM model. Defines the number of different tokens that can be represented by the
44
+ `inputs_ids` passed when calling [`InternLMModel`]
45
+ hidden_size (`int`, *optional*, defaults to 4096):
46
+ Dimension of the hidden representations.
47
+ intermediate_size (`int`, *optional*, defaults to 11008):
48
+ Dimension of the MLP representations.
49
+ num_hidden_layers (`int`, *optional*, defaults to 32):
50
+ Number of hidden layers in the Transformer encoder.
51
+ num_attention_heads (`int`, *optional*, defaults to 32):
52
+ Number of attention heads for each attention layer in the Transformer encoder.
53
+ hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
54
+ The non-linear activation function (function or string) in the decoder.
55
+ max_position_embeddings (`int`, *optional*, defaults to 2048):
56
+ The maximum sequence length that this model might ever be used with. Typically set this to something large
57
+ just in case (e.g., 512 or 1024 or 2048).
58
+ initializer_range (`float`, *optional*, defaults to 0.02):
59
+ The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
60
+ rms_norm_eps (`float`, *optional*, defaults to 1e-12):
61
+ The epsilon used by the rms normalization layers.
62
+ use_cache (`bool`, *optional*, defaults to `True`):
63
+ Whether or not the model should return the last key/values attentions (not used by all models). Only
64
+ relevant if `config.is_decoder=True`.
65
+ tie_word_embeddings(`bool`, *optional*, defaults to `False`):
66
+ Whether to tie weight embeddings
67
+ Example:
68
+
69
+ ```python
70
+ >>> from transformers import InternLMModel, InternLMConfig
71
+
72
+ >>> # Initializing a InternLM internlm-7b style configuration
73
+ >>> configuration = InternLMConfig()
74
+
75
+ >>> # Initializing a model from the internlm-7b style configuration
76
+ >>> model = InternLMModel(configuration)
77
+
78
+ >>> # Accessing the model configuration
79
+ >>> configuration = model.config
80
+ ```"""
81
+ model_type = "internlm"
82
+ _auto_class = "AutoConfig"
83
+
84
+ def __init__(
85
+ self,
86
+ vocab_size=103168,
87
+ hidden_size=4096,
88
+ intermediate_size=11008,
89
+ num_hidden_layers=32,
90
+ num_attention_heads=32,
91
+ hidden_act="silu",
92
+ max_position_embeddings=2048,
93
+ initializer_range=0.02,
94
+ rms_norm_eps=1e-6,
95
+ use_cache=True,
96
+ pad_token_id=0,
97
+ bos_token_id=1,
98
+ eos_token_id=2,
99
+ tie_word_embeddings=False,
100
+ bias=True,
101
+ **kwargs,
102
+ ):
103
+ self.vocab_size = vocab_size
104
+ self.max_position_embeddings = max_position_embeddings
105
+ self.hidden_size = hidden_size
106
+ self.intermediate_size = intermediate_size
107
+ self.num_hidden_layers = num_hidden_layers
108
+ self.num_attention_heads = num_attention_heads
109
+ self.hidden_act = hidden_act
110
+ self.initializer_range = initializer_range
111
+ self.rms_norm_eps = rms_norm_eps
112
+ self.use_cache = use_cache
113
+ self.bias = bias
114
+ super().__init__(
115
+ pad_token_id=pad_token_id,
116
+ bos_token_id=bos_token_id,
117
+ eos_token_id=eos_token_id,
118
+ tie_word_embeddings=tie_word_embeddings,
119
+ **kwargs,
120
+ )
generation_config.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 1,
4
+ "eos_token_id": 2,
5
+ "pad_token_id": 0,
6
+ "transformers_version": "4.29.2"
7
+ }
modeling_internlm.py ADDED
@@ -0,0 +1,962 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # coding=utf-8
2
+ # Copyright 2022 EleutherAI and the HuggingFace Inc. team. All rights reserved.
3
+ #
4
+ # This code is based on EleutherAI's GPT-NeoX library and the GPT-NeoX
5
+ # and OPT implementations in this library. It has been modified from its
6
+ # original forms to accommodate minor architectural differences compared
7
+ # to GPT-NeoX and OPT used by the Meta AI team that trained the model.
8
+ #
9
+ # Licensed under the Apache License, Version 2.0 (the "License");
10
+ # you may not use this file except in compliance with the License.
11
+ # You may obtain a copy of the License at
12
+ #
13
+ # http://www.apache.org/licenses/LICENSE-2.0
14
+ #
15
+ # Unless required by applicable law or agreed to in writing, software
16
+ # distributed under the License is distributed on an "AS IS" BASIS,
17
+ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
18
+ # See the License for the specific language governing permissions and
19
+ # limitations under the License.
20
+ """ PyTorch InternLM model."""
21
+ import math
22
+ from typing import List, Optional, Tuple, Union
23
+
24
+ import torch
25
+ import torch.utils.checkpoint
26
+ from torch import nn
27
+ from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
28
+
29
+ from transformers.activations import ACT2FN
30
+ from transformers.modeling_outputs import BaseModelOutputWithPast, CausalLMOutputWithPast, SequenceClassifierOutputWithPast
31
+ from transformers.modeling_utils import PreTrainedModel
32
+ from transformers.generation.streamers import BaseStreamer
33
+ from transformers.utils import add_start_docstrings, add_start_docstrings_to_model_forward, logging, replace_return_docstrings
34
+ from .configuration_internlm import InternLMConfig
35
+
36
+
37
+ logger = logging.get_logger(__name__)
38
+
39
+ _CONFIG_FOR_DOC = "InternLMConfig"
40
+
41
+ # Copied from transformers.models.bart.modeling_bart._make_causal_mask
42
+ def _make_causal_mask(
43
+ input_ids_shape: torch.Size, dtype: torch.dtype, device: torch.device, past_key_values_length: int = 0
44
+ ):
45
+ """
46
+ Make causal mask used for bi-directional self-attention.
47
+ """
48
+ bsz, tgt_len = input_ids_shape
49
+ mask = torch.full((tgt_len, tgt_len), torch.tensor(torch.finfo(dtype).min, device=device), device=device)
50
+ mask_cond = torch.arange(mask.size(-1), device=device)
51
+ mask.masked_fill_(mask_cond < (mask_cond + 1).view(mask.size(-1), 1), 0)
52
+ mask = mask.to(dtype)
53
+
54
+ if past_key_values_length > 0:
55
+ mask = torch.cat([torch.zeros(tgt_len, past_key_values_length, dtype=dtype, device=device), mask], dim=-1)
56
+ return mask[None, None, :, :].expand(bsz, 1, tgt_len, tgt_len + past_key_values_length)
57
+
58
+
59
+ # Copied from transformers.models.bart.modeling_bart._expand_mask
60
+ def _expand_mask(mask: torch.Tensor, dtype: torch.dtype, tgt_len: Optional[int] = None):
61
+ """
62
+ Expands attention_mask from `[bsz, seq_len]` to `[bsz, 1, tgt_seq_len, src_seq_len]`.
63
+ """
64
+ bsz, src_len = mask.size()
65
+ tgt_len = tgt_len if tgt_len is not None else src_len
66
+
67
+ expanded_mask = mask[:, None, None, :].expand(bsz, 1, tgt_len, src_len).to(dtype)
68
+
69
+ inverted_mask = 1.0 - expanded_mask
70
+
71
+ return inverted_mask.masked_fill(inverted_mask.to(torch.bool), torch.finfo(dtype).min)
72
+
73
+
74
+ class InternLMRMSNorm(nn.Module):
75
+ def __init__(self, hidden_size, eps=1e-6):
76
+ """
77
+ InternLMRMSNorm is equivalent to T5LayerNorm
78
+ """
79
+ super().__init__()
80
+ self.weight = nn.Parameter(torch.ones(hidden_size))
81
+ self.variance_epsilon = eps
82
+
83
+ def forward(self, hidden_states):
84
+ variance = hidden_states.to(torch.float32).pow(2).mean(-1, keepdim=True)
85
+ hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)
86
+
87
+ # convert into half-precision if necessary
88
+ if self.weight.dtype in [torch.float16, torch.bfloat16]:
89
+ hidden_states = hidden_states.to(self.weight.dtype)
90
+
91
+ return self.weight * hidden_states
92
+
93
+
94
+ class InternLMRotaryEmbedding(torch.nn.Module):
95
+ def __init__(self, dim, max_position_embeddings=2048, base=10000, device=None):
96
+ super().__init__()
97
+ inv_freq = 1.0 / (base ** (torch.arange(0, dim, 2).float().to(device) / dim))
98
+ self.register_buffer("inv_freq", inv_freq)
99
+
100
+ # Build here to make `torch.jit.trace` work.
101
+ self.max_seq_len_cached = max_position_embeddings
102
+ t = torch.arange(self.max_seq_len_cached, device=self.inv_freq.device, dtype=self.inv_freq.dtype)
103
+ freqs = torch.einsum("i,j->ij", t, self.inv_freq)
104
+ # Different from paper, but it uses a different permutation in order to obtain the same calculation
105
+ emb = torch.cat((freqs, freqs), dim=-1)
106
+ self.register_buffer("cos_cached", emb.cos()[None, None, :, :], persistent=False)
107
+ self.register_buffer("sin_cached", emb.sin()[None, None, :, :], persistent=False)
108
+
109
+ def forward(self, x, seq_len=None):
110
+ # x: [bs, num_attention_heads, seq_len, head_size]
111
+ # This `if` block is unlikely to be run after we build sin/cos in `__init__`. Keep the logic here just in case.
112
+ if seq_len > self.max_seq_len_cached:
113
+ self.max_seq_len_cached = seq_len
114
+ t = torch.arange(self.max_seq_len_cached, device=x.device, dtype=self.inv_freq.dtype)
115
+ freqs = torch.einsum("i,j->ij", t, self.inv_freq)
116
+ # Different from paper, but it uses a different permutation in order to obtain the same calculation
117
+ emb = torch.cat((freqs, freqs), dim=-1).to(x.device)
118
+ self.register_buffer("cos_cached", emb.cos()[None, None, :, :], persistent=False)
119
+ self.register_buffer("sin_cached", emb.sin()[None, None, :, :], persistent=False)
120
+ return (
121
+ self.cos_cached[:, :, :seq_len, ...].to(dtype=x.dtype),
122
+ self.sin_cached[:, :, :seq_len, ...].to(dtype=x.dtype),
123
+ )
124
+
125
+
126
+ def rotate_half(x):
127
+ """Rotates half the hidden dims of the input."""
128
+ x1 = x[..., : x.shape[-1] // 2]
129
+ x2 = x[..., x.shape[-1] // 2 :]
130
+ return torch.cat((-x2, x1), dim=-1)
131
+
132
+
133
+ def apply_rotary_pos_emb(q, k, cos, sin, position_ids):
134
+ # The first two dimensions of cos and sin are always 1, so we can `squeeze` them.
135
+ cos = cos.squeeze(1).squeeze(0) # [seq_len, dim]
136
+ sin = sin.squeeze(1).squeeze(0) # [seq_len, dim]
137
+ cos = cos[position_ids].unsqueeze(1) # [bs, 1, seq_len, dim]
138
+ sin = sin[position_ids].unsqueeze(1) # [bs, 1, seq_len, dim]
139
+ q_embed = (q * cos) + (rotate_half(q) * sin)
140
+ k_embed = (k * cos) + (rotate_half(k) * sin)
141
+ return q_embed, k_embed
142
+
143
+
144
+ class InternLMMLP(nn.Module):
145
+ def __init__(
146
+ self,
147
+ hidden_size: int,
148
+ intermediate_size: int,
149
+ hidden_act: str,
150
+ ):
151
+ super().__init__()
152
+ self.gate_proj = nn.Linear(hidden_size, intermediate_size, bias=False)
153
+ self.down_proj = nn.Linear(intermediate_size, hidden_size, bias=False)
154
+ self.up_proj = nn.Linear(hidden_size, intermediate_size, bias=False)
155
+ self.act_fn = ACT2FN[hidden_act]
156
+
157
+ def forward(self, x):
158
+ return self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))
159
+
160
+
161
+ class InternLMAttention(nn.Module):
162
+ """Multi-headed attention from 'Attention Is All You Need' paper"""
163
+
164
+ def __init__(self, config: InternLMConfig):
165
+ super().__init__()
166
+ self.config = config
167
+ self.hidden_size = config.hidden_size
168
+ self.num_heads = config.num_attention_heads
169
+ self.head_dim = self.hidden_size // self.num_heads
170
+ self.max_position_embeddings = config.max_position_embeddings
171
+
172
+ if (self.head_dim * self.num_heads) != self.hidden_size:
173
+ raise ValueError(
174
+ f"hidden_size must be divisible by num_heads (got `hidden_size`: {self.hidden_size}"
175
+ f" and `num_heads`: {self.num_heads})."
176
+ )
177
+ self.q_proj = nn.Linear(self.hidden_size, self.num_heads * self.head_dim, bias=config.bias)
178
+ self.k_proj = nn.Linear(self.hidden_size, self.num_heads * self.head_dim, bias=config.bias)
179
+ self.v_proj = nn.Linear(self.hidden_size, self.num_heads * self.head_dim, bias=config.bias)
180
+ self.o_proj = nn.Linear(self.num_heads * self.head_dim, self.hidden_size, bias=config.bias)
181
+ self.rotary_emb = InternLMRotaryEmbedding(self.head_dim, max_position_embeddings=self.max_position_embeddings)
182
+
183
+ def _shape(self, tensor: torch.Tensor, seq_len: int, bsz: int):
184
+ return tensor.view(bsz, seq_len, self.num_heads, self.head_dim).transpose(1, 2).contiguous()
185
+
186
+ def forward(
187
+ self,
188
+ hidden_states: torch.Tensor,
189
+ attention_mask: Optional[torch.Tensor] = None,
190
+ position_ids: Optional[torch.LongTensor] = None,
191
+ past_key_value: Optional[Tuple[torch.Tensor]] = None,
192
+ output_attentions: bool = False,
193
+ use_cache: bool = False,
194
+ ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
195
+ bsz, q_len, _ = hidden_states.size()
196
+
197
+ query_states = self.q_proj(hidden_states).view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
198
+ key_states = self.k_proj(hidden_states).view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
199
+ value_states = self.v_proj(hidden_states).view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
200
+
201
+ kv_seq_len = key_states.shape[-2]
202
+ if past_key_value is not None:
203
+ kv_seq_len += past_key_value[0].shape[-2]
204
+ cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
205
+ query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids)
206
+ # [bsz, nh, t, hd]
207
+
208
+ if past_key_value is not None:
209
+ # reuse k, v, self_attention
210
+ key_states = torch.cat([past_key_value[0], key_states], dim=2)
211
+ value_states = torch.cat([past_key_value[1], value_states], dim=2)
212
+
213
+ past_key_value = (key_states, value_states) if use_cache else None
214
+
215
+ attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) / math.sqrt(self.head_dim)
216
+
217
+ if attn_weights.size() != (bsz, self.num_heads, q_len, kv_seq_len):
218
+ raise ValueError(
219
+ f"Attention weights should be of size {(bsz, self.num_heads, q_len, kv_seq_len)}, but is"
220
+ f" {attn_weights.size()}"
221
+ )
222
+
223
+ if attention_mask is not None:
224
+ if attention_mask.size() != (bsz, 1, q_len, kv_seq_len):
225
+ raise ValueError(
226
+ f"Attention mask should be of size {(bsz, 1, q_len, kv_seq_len)}, but is {attention_mask.size()}"
227
+ )
228
+ attn_weights = attn_weights + attention_mask
229
+ attn_weights = torch.max(attn_weights, torch.tensor(torch.finfo(attn_weights.dtype).min))
230
+
231
+ # upcast attention to fp32
232
+ attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query_states.dtype)
233
+ attn_output = torch.matmul(attn_weights, value_states)
234
+
235
+ if attn_output.size() != (bsz, self.num_heads, q_len, self.head_dim):
236
+ raise ValueError(
237
+ f"`attn_output` should be of size {(bsz, self.num_heads, q_len, self.head_dim)}, but is"
238
+ f" {attn_output.size()}"
239
+ )
240
+
241
+ attn_output = attn_output.transpose(1, 2)
242
+ attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
243
+
244
+ attn_output = self.o_proj(attn_output)
245
+
246
+ if not output_attentions:
247
+ attn_weights = None
248
+
249
+ return attn_output, attn_weights, past_key_value
250
+
251
+
252
+ class InternLMDecoderLayer(nn.Module):
253
+ def __init__(self, config: InternLMConfig):
254
+ super().__init__()
255
+ self.hidden_size = config.hidden_size
256
+ self.self_attn = InternLMAttention(config=config)
257
+ self.mlp = InternLMMLP(
258
+ hidden_size=self.hidden_size,
259
+ intermediate_size=config.intermediate_size,
260
+ hidden_act=config.hidden_act,
261
+ )
262
+ self.input_layernorm = InternLMRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
263
+ self.post_attention_layernorm = InternLMRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
264
+
265
+ def forward(
266
+ self,
267
+ hidden_states: torch.Tensor,
268
+ attention_mask: Optional[torch.Tensor] = None,
269
+ position_ids: Optional[torch.LongTensor] = None,
270
+ past_key_value: Optional[Tuple[torch.Tensor]] = None,
271
+ output_attentions: Optional[bool] = False,
272
+ use_cache: Optional[bool] = False,
273
+ ) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]:
274
+ """
275
+ Args:
276
+ hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
277
+ attention_mask (`torch.FloatTensor`, *optional*): attention mask of size
278
+ `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
279
+ output_attentions (`bool`, *optional*):
280
+ Whether or not to return the attentions tensors of all attention layers. See `attentions` under
281
+ returned tensors for more detail.
282
+ use_cache (`bool`, *optional*):
283
+ If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding
284
+ (see `past_key_values`).
285
+ past_key_value (`Tuple(torch.FloatTensor)`, *optional*): cached past key and value projection states
286
+ """
287
+
288
+ residual = hidden_states
289
+
290
+ hidden_states = self.input_layernorm(hidden_states)
291
+
292
+ # Self Attention
293
+ hidden_states, self_attn_weights, present_key_value = self.self_attn(
294
+ hidden_states=hidden_states,
295
+ attention_mask=attention_mask,
296
+ position_ids=position_ids,
297
+ past_key_value=past_key_value,
298
+ output_attentions=output_attentions,
299
+ use_cache=use_cache,
300
+ )
301
+ hidden_states = residual + hidden_states
302
+
303
+ # Fully Connected
304
+ residual = hidden_states
305
+ hidden_states = self.post_attention_layernorm(hidden_states)
306
+ hidden_states = self.mlp(hidden_states)
307
+ hidden_states = residual + hidden_states
308
+
309
+ outputs = (hidden_states,)
310
+
311
+ if output_attentions:
312
+ outputs += (self_attn_weights,)
313
+
314
+ if use_cache:
315
+ outputs += (present_key_value,)
316
+
317
+ return outputs
318
+
319
+
320
+ INTERNLM_START_DOCSTRING = r"""
321
+ This model inherits from [`PreTrainedModel`]. Check the superclass documentation for the generic methods the
322
+ library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
323
+ etc.)
324
+
325
+ This model is also a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) subclass.
326
+ Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage
327
+ and behavior.
328
+
329
+ Parameters:
330
+ config ([`InternLMConfig`]):
331
+ Model configuration class with all the parameters of the model. Initializing with a config file does not
332
+ load the weights associated with the model, only the configuration. Check out the
333
+ [`~PreTrainedModel.from_pretrained`] method to load the model weights.
334
+ """
335
+
336
+
337
+ @add_start_docstrings(
338
+ "The bare InternLM Model outputting raw hidden-states without any specific head on top.",
339
+ INTERNLM_START_DOCSTRING,
340
+ )
341
+ class InternLMPreTrainedModel(PreTrainedModel):
342
+ config_class = InternLMConfig
343
+ base_model_prefix = "model"
344
+ supports_gradient_checkpointing = True
345
+ _no_split_modules = ["InternLMDecoderLayer"]
346
+ _keys_to_ignore_on_load_unexpected = [r"decoder\.version"]
347
+
348
+ def _init_weights(self, module):
349
+ std = self.config.initializer_range
350
+ if isinstance(module, nn.Linear):
351
+ module.weight.data.normal_(mean=0.0, std=std)
352
+ if module.bias is not None:
353
+ module.bias.data.zero_()
354
+ elif isinstance(module, nn.Embedding):
355
+ module.weight.data.normal_(mean=0.0, std=std)
356
+ if module.padding_idx is not None:
357
+ module.weight.data[module.padding_idx].zero_()
358
+
359
+ def _set_gradient_checkpointing(self, module, value=False):
360
+ if isinstance(module, InternLMModel):
361
+ module.gradient_checkpointing = value
362
+
363
+
364
+ INTERNLM_INPUTS_DOCSTRING = r"""
365
+ Args:
366
+ input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
367
+ Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide
368
+ it.
369
+
370
+ Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
371
+ [`PreTrainedTokenizer.__call__`] for details.
372
+
373
+ [What are input IDs?](../glossary#input-ids)
374
+ attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
375
+ Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
376
+
377
+ - 1 for tokens that are **not masked**,
378
+ - 0 for tokens that are **masked**.
379
+
380
+ [What are attention masks?](../glossary#attention-mask)
381
+
382
+ Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
383
+ [`PreTrainedTokenizer.__call__`] for details.
384
+
385
+ If `past_key_values` is used, optionally only the last `decoder_input_ids` have to be input (see
386
+ `past_key_values`).
387
+
388
+ If you want to change padding behavior, you should read [`modeling_opt._prepare_decoder_attention_mask`]
389
+ and modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more
390
+ information on the default strategy.
391
+
392
+ - 1 indicates the head is **not masked**,
393
+ - 0 indicates the head is **masked**.
394
+ position_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
395
+ Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0,
396
+ config.n_positions - 1]`.
397
+
398
+ [What are position IDs?](../glossary#position-ids)
399
+ past_key_values (`tuple(tuple(torch.FloatTensor))`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`):
400
+ Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of shape
401
+ `(batch_size, num_heads, sequence_length, embed_size_per_head)`) and 2 additional tensors of shape
402
+ `(batch_size, num_heads, encoder_sequence_length, embed_size_per_head)`.
403
+
404
+ Contains pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention
405
+ blocks) that can be used (see `past_key_values` input) to speed up sequential decoding.
406
+
407
+ If `past_key_values` are used, the user can optionally input only the last `decoder_input_ids` (those that
408
+ don't have their past key value states given to this model) of shape `(batch_size, 1)` instead of all
409
+ `decoder_input_ids` of shape `(batch_size, sequence_length)`.
410
+ inputs_embeds (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
411
+ Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation. This
412
+ is useful if you want more control over how to convert `input_ids` indices into associated vectors than the
413
+ model's internal embedding lookup matrix.
414
+ use_cache (`bool`, *optional*):
415
+ If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see
416
+ `past_key_values`).
417
+ output_attentions (`bool`, *optional*):
418
+ Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned
419
+ tensors for more detail.
420
+ output_hidden_states (`bool`, *optional*):
421
+ Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
422
+ more detail.
423
+ return_dict (`bool`, *optional*):
424
+ Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
425
+ """
426
+
427
+
428
+ @add_start_docstrings(
429
+ "The bare InternLM Model outputting raw hidden-states without any specific head on top.",
430
+ INTERNLM_START_DOCSTRING,
431
+ )
432
+ class InternLMModel(InternLMPreTrainedModel):
433
+ """
434
+ Transformer decoder consisting of *config.num_hidden_layers* layers. Each layer is a [`InternLMDecoderLayer`]
435
+
436
+ Args:
437
+ config: InternLMConfig
438
+ """
439
+ _auto_class = "AutoModel"
440
+
441
+ def __init__(self, config: InternLMConfig):
442
+ super().__init__(config)
443
+ self.padding_idx = config.pad_token_id
444
+ self.vocab_size = config.vocab_size
445
+
446
+ self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size, self.padding_idx)
447
+ self.layers = nn.ModuleList([InternLMDecoderLayer(config) for _ in range(config.num_hidden_layers)])
448
+ self.norm = InternLMRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
449
+
450
+ self.gradient_checkpointing = False
451
+ # Initialize weights and apply final processing
452
+ self.post_init()
453
+
454
+ def get_input_embeddings(self):
455
+ return self.embed_tokens
456
+
457
+ def set_input_embeddings(self, value):
458
+ self.embed_tokens = value
459
+
460
+ # Copied from transformers.models.bart.modeling_bart.BartDecoder._prepare_decoder_attention_mask
461
+ def _prepare_decoder_attention_mask(self, attention_mask, input_shape, inputs_embeds, past_key_values_length):
462
+ # create causal mask
463
+ # [bsz, seq_len] -> [bsz, 1, tgt_seq_len, src_seq_len]
464
+ combined_attention_mask = None
465
+ if input_shape[-1] > 1:
466
+ combined_attention_mask = _make_causal_mask(
467
+ input_shape,
468
+ inputs_embeds.dtype,
469
+ device=inputs_embeds.device,
470
+ past_key_values_length=past_key_values_length,
471
+ )
472
+
473
+ if attention_mask is not None:
474
+ # [bsz, seq_len] -> [bsz, 1, tgt_seq_len, src_seq_len]
475
+ expanded_attn_mask = _expand_mask(attention_mask, inputs_embeds.dtype, tgt_len=input_shape[-1]).to(
476
+ inputs_embeds.device
477
+ )
478
+ combined_attention_mask = (
479
+ expanded_attn_mask if combined_attention_mask is None else expanded_attn_mask + combined_attention_mask
480
+ )
481
+
482
+ return combined_attention_mask
483
+
484
+ @add_start_docstrings_to_model_forward(INTERNLM_INPUTS_DOCSTRING)
485
+ def forward(
486
+ self,
487
+ input_ids: torch.LongTensor = None,
488
+ attention_mask: Optional[torch.Tensor] = None,
489
+ position_ids: Optional[torch.LongTensor] = None,
490
+ past_key_values: Optional[List[torch.FloatTensor]] = None,
491
+ inputs_embeds: Optional[torch.FloatTensor] = None,
492
+ use_cache: Optional[bool] = None,
493
+ output_attentions: Optional[bool] = None,
494
+ output_hidden_states: Optional[bool] = None,
495
+ return_dict: Optional[bool] = None,
496
+ ) -> Union[Tuple, BaseModelOutputWithPast]:
497
+ output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
498
+ output_hidden_states = (
499
+ output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
500
+ )
501
+ use_cache = use_cache if use_cache is not None else self.config.use_cache
502
+
503
+ return_dict = return_dict if return_dict is not None else self.config.use_return_dict
504
+
505
+ # retrieve input_ids and inputs_embeds
506
+ if input_ids is not None and inputs_embeds is not None:
507
+ raise ValueError("You cannot specify both decoder_input_ids and decoder_inputs_embeds at the same time")
508
+ elif input_ids is not None:
509
+ batch_size, seq_length = input_ids.shape
510
+ elif inputs_embeds is not None:
511
+ batch_size, seq_length, _ = inputs_embeds.shape
512
+ else:
513
+ raise ValueError("You have to specify either decoder_input_ids or decoder_inputs_embeds")
514
+
515
+ seq_length_with_past = seq_length
516
+ past_key_values_length = 0
517
+
518
+ if past_key_values is not None:
519
+ past_key_values_length = past_key_values[0][0].shape[2]
520
+ seq_length_with_past = seq_length_with_past + past_key_values_length
521
+
522
+ if position_ids is None:
523
+ device = input_ids.device if input_ids is not None else inputs_embeds.device
524
+ position_ids = torch.arange(
525
+ past_key_values_length, seq_length + past_key_values_length, dtype=torch.long, device=device
526
+ )
527
+ position_ids = position_ids.unsqueeze(0).view(-1, seq_length)
528
+ else:
529
+ position_ids = position_ids.view(-1, seq_length).long()
530
+
531
+ if inputs_embeds is None:
532
+ inputs_embeds = self.embed_tokens(input_ids)
533
+ # embed positions
534
+ if attention_mask is None:
535
+ attention_mask = torch.ones(
536
+ (batch_size, seq_length_with_past), dtype=torch.bool, device=inputs_embeds.device
537
+ )
538
+ attention_mask = self._prepare_decoder_attention_mask(
539
+ attention_mask, (batch_size, seq_length), inputs_embeds, past_key_values_length
540
+ )
541
+
542
+ hidden_states = inputs_embeds
543
+
544
+ if self.gradient_checkpointing and self.training:
545
+ if use_cache:
546
+ logger.warning_once(
547
+ "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..."
548
+ )
549
+ use_cache = False
550
+
551
+ # decoder layers
552
+ all_hidden_states = () if output_hidden_states else None
553
+ all_self_attns = () if output_attentions else None
554
+ next_decoder_cache = () if use_cache else None
555
+
556
+ for idx, decoder_layer in enumerate(self.layers):
557
+ if output_hidden_states:
558
+ all_hidden_states += (hidden_states,)
559
+
560
+ past_key_value = past_key_values[idx] if past_key_values is not None else None
561
+
562
+ if self.gradient_checkpointing and self.training:
563
+
564
+ def create_custom_forward(module):
565
+ def custom_forward(*inputs):
566
+ # None for past_key_value
567
+ return module(*inputs, output_attentions, None)
568
+
569
+ return custom_forward
570
+
571
+ layer_outputs = torch.utils.checkpoint.checkpoint(
572
+ create_custom_forward(decoder_layer),
573
+ hidden_states,
574
+ attention_mask,
575
+ position_ids,
576
+ None,
577
+ )
578
+ else:
579
+ layer_outputs = decoder_layer(
580
+ hidden_states,
581
+ attention_mask=attention_mask,
582
+ position_ids=position_ids,
583
+ past_key_value=past_key_value,
584
+ output_attentions=output_attentions,
585
+ use_cache=use_cache,
586
+ )
587
+
588
+ hidden_states = layer_outputs[0]
589
+
590
+ if use_cache:
591
+ next_decoder_cache += (layer_outputs[2 if output_attentions else 1],)
592
+
593
+ if output_attentions:
594
+ all_self_attns += (layer_outputs[1],)
595
+
596
+ hidden_states = self.norm(hidden_states)
597
+
598
+ # add hidden states from the last decoder layer
599
+ if output_hidden_states:
600
+ all_hidden_states += (hidden_states,)
601
+
602
+ next_cache = next_decoder_cache if use_cache else None
603
+ if not return_dict:
604
+ return tuple(v for v in [hidden_states, next_cache, all_hidden_states, all_self_attns] if v is not None)
605
+ return BaseModelOutputWithPast(
606
+ last_hidden_state=hidden_states,
607
+ past_key_values=next_cache,
608
+ hidden_states=all_hidden_states,
609
+ attentions=all_self_attns,
610
+ )
611
+
612
+
613
+ class InternLMForCausalLM(InternLMPreTrainedModel):
614
+ _auto_class = "AutoModelForCausalLM"
615
+
616
+ def __init__(self, config):
617
+ super().__init__(config)
618
+ self.model = InternLMModel(config)
619
+
620
+ self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
621
+
622
+ # Initialize weights and apply final processing
623
+ self.post_init()
624
+
625
+ def get_input_embeddings(self):
626
+ return self.model.embed_tokens
627
+
628
+ def set_input_embeddings(self, value):
629
+ self.model.embed_tokens = value
630
+
631
+ def get_output_embeddings(self):
632
+ return self.lm_head
633
+
634
+ def set_output_embeddings(self, new_embeddings):
635
+ self.lm_head = new_embeddings
636
+
637
+ def set_decoder(self, decoder):
638
+ self.model = decoder
639
+
640
+ def get_decoder(self):
641
+ return self.model
642
+
643
+ @add_start_docstrings_to_model_forward(INTERNLM_INPUTS_DOCSTRING)
644
+ @replace_return_docstrings(output_type=CausalLMOutputWithPast, config_class=_CONFIG_FOR_DOC)
645
+ def forward(
646
+ self,
647
+ input_ids: torch.LongTensor = None,
648
+ attention_mask: Optional[torch.Tensor] = None,
649
+ position_ids: Optional[torch.LongTensor] = None,
650
+ past_key_values: Optional[List[torch.FloatTensor]] = None,
651
+ inputs_embeds: Optional[torch.FloatTensor] = None,
652
+ labels: Optional[torch.LongTensor] = None,
653
+ use_cache: Optional[bool] = None,
654
+ output_attentions: Optional[bool] = None,
655
+ output_hidden_states: Optional[bool] = None,
656
+ return_dict: Optional[bool] = None,
657
+ ) -> Union[Tuple, CausalLMOutputWithPast]:
658
+ r"""
659
+ Args:
660
+ labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
661
+ Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
662
+ config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
663
+ (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`.
664
+
665
+ Returns:
666
+
667
+ Example:
668
+
669
+ ```python
670
+ >>> from transformers import AutoTokenizer, InternLMForCausalLM
671
+
672
+ >>> model = InternLMForCausalLM.from_pretrained(PATH_TO_CONVERTED_WEIGHTS)
673
+ >>> tokenizer = AutoTokenizer.from_pretrained(PATH_TO_CONVERTED_TOKENIZER)
674
+
675
+ >>> prompt = "Hey, are you consciours? Can you talk to me?"
676
+ >>> inputs = tokenizer(prompt, return_tensors="pt")
677
+
678
+ >>> # Generate
679
+ >>> generate_ids = model.generate(inputs.input_ids, max_length=30)
680
+ >>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
681
+ "Hey, are you consciours? Can you talk to me?\nI'm not consciours, but I can talk to you."
682
+ ```"""
683
+
684
+ output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
685
+ output_hidden_states = (
686
+ output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
687
+ )
688
+ return_dict = return_dict if return_dict is not None else self.config.use_return_dict
689
+
690
+ # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
691
+ outputs = self.model(
692
+ input_ids=input_ids,
693
+ attention_mask=attention_mask,
694
+ position_ids=position_ids,
695
+ past_key_values=past_key_values,
696
+ inputs_embeds=inputs_embeds,
697
+ use_cache=use_cache,
698
+ output_attentions=output_attentions,
699
+ output_hidden_states=output_hidden_states,
700
+ return_dict=return_dict,
701
+ )
702
+
703
+ hidden_states = outputs[0]
704
+ logits = self.lm_head(hidden_states)
705
+
706
+ loss = None
707
+ if labels is not None:
708
+ # Shift so that tokens < n predict n
709
+ shift_logits = logits[..., :-1, :].contiguous()
710
+ shift_labels = labels[..., 1:].contiguous()
711
+ # Flatten the tokens
712
+ loss_fct = CrossEntropyLoss()
713
+ shift_logits = shift_logits.view(-1, self.config.vocab_size)
714
+ shift_labels = shift_labels.view(-1)
715
+ # Enable model parallelism
716
+ shift_labels = shift_labels.to(shift_logits.device)
717
+ loss = loss_fct(shift_logits, shift_labels)
718
+
719
+ if not return_dict:
720
+ output = (logits,) + outputs[1:]
721
+ return (loss,) + output if loss is not None else output
722
+
723
+ return CausalLMOutputWithPast(
724
+ loss=loss,
725
+ logits=logits,
726
+ past_key_values=outputs.past_key_values,
727
+ hidden_states=outputs.hidden_states,
728
+ attentions=outputs.attentions,
729
+ )
730
+
731
+ def prepare_inputs_for_generation(
732
+ self, input_ids, past_key_values=None, attention_mask=None, inputs_embeds=None, **kwargs
733
+ ):
734
+ if past_key_values:
735
+ input_ids = input_ids[:, -1:]
736
+
737
+ position_ids = kwargs.get("position_ids", None)
738
+ if attention_mask is not None and position_ids is None:
739
+ # create position_ids on the fly for batch generation
740
+ position_ids = attention_mask.long().cumsum(-1) - 1
741
+ position_ids.masked_fill_(attention_mask == 0, 1)
742
+ if past_key_values:
743
+ position_ids = position_ids[:, -1].unsqueeze(-1)
744
+
745
+ # if `inputs_embeds` are passed, we only want to use them in the 1st generation step
746
+ if inputs_embeds is not None and past_key_values is None:
747
+ model_inputs = {"inputs_embeds": inputs_embeds}
748
+ else:
749
+ model_inputs = {"input_ids": input_ids}
750
+
751
+ model_inputs.update(
752
+ {
753
+ "position_ids": position_ids,
754
+ "past_key_values": past_key_values,
755
+ "use_cache": kwargs.get("use_cache"),
756
+ "attention_mask": attention_mask,
757
+ }
758
+ )
759
+ return model_inputs
760
+
761
+ @staticmethod
762
+ def _reorder_cache(past_key_values, beam_idx):
763
+ reordered_past = ()
764
+ for layer_past in past_key_values:
765
+ reordered_past += (tuple(past_state.index_select(0, beam_idx) for past_state in layer_past),)
766
+ return reordered_past
767
+
768
+ def build_inputs(self, tokenizer, query: str, history: List[Tuple[str, str]] = []):
769
+ prompt = ""
770
+ for record in history:
771
+ prompt += f"""<s><|User|>:{record[0]}<eoh>\n<|Bot|>:{record[1]}<eoa>\n"""
772
+ if len(prompt) == 0:
773
+ prompt += "<s>"
774
+ prompt += f"""<|User|>:{query}<eoh>\n<|Bot|>:"""
775
+ return tokenizer([prompt], return_tensors="pt")
776
+
777
+ @torch.no_grad()
778
+ def chat(self,
779
+ tokenizer,
780
+ query: str,
781
+ history: List[Tuple[str, str]] = [],
782
+ streamer: Optional[BaseStreamer] = None,
783
+ max_new_tokens: int = 1024,
784
+ do_sample: bool = True,
785
+ temperature: float = 0.8,
786
+ top_p: float = 0.8,
787
+ **kwargs):
788
+ inputs = self.build_inputs(tokenizer, query, history)
789
+ inputs = {k: v.to(self.device) for k, v in inputs.items() if torch.is_tensor(v)}
790
+ outputs = self.generate(**inputs,
791
+ streamer=streamer,
792
+ max_new_tokens=max_new_tokens,
793
+ do_sample=do_sample,
794
+ temperature=temperature,
795
+ top_p=top_p,
796
+ **kwargs)
797
+ outputs = outputs[0].cpu().tolist()[len(inputs["input_ids"][0]):]
798
+ response = tokenizer.decode(outputs, skip_special_tokens=True)
799
+ response = response.split("<eoa>")[0]
800
+ history = history + [(query, response)]
801
+ return response, history
802
+
803
+ @torch.no_grad()
804
+ def stream_chat(self,
805
+ tokenizer,
806
+ query: str,
807
+ history: List[Tuple[str, str]] = [],
808
+ max_new_tokens: int = 1024,
809
+ do_sample: bool = True,
810
+ temperature: float = 0.8,
811
+ top_p: float = 0.8,
812
+ **kwargs):
813
+ class ChatStreamer(BaseStreamer):
814
+ def __init__(self, tokenizer) -> None:
815
+ super().__init__()
816
+ self.tokenizer = tokenizer
817
+
818
+ def put(self, value):
819
+ if len(value.shape) > 1 and value.shape[0] > 1:
820
+ raise ValueError("ChatStreamer only supports batch size 1")
821
+ elif len(value.shape) > 1:
822
+ value = value[0]
823
+ token = self.tokenizer.decode([value[-1]], skip_special_tokens=True)
824
+ if token.strip() != "<eoa>":
825
+ print(token, end="")
826
+
827
+ def end(self):
828
+ print("")
829
+
830
+ return self.chat(
831
+ tokenizer=tokenizer,
832
+ query=query,
833
+ streamer=ChatStreamer(tokenizer=tokenizer),
834
+ history=history,
835
+ max_new_tokens=max_new_tokens,
836
+ do_sample=do_sample,
837
+ temperature=temperature,
838
+ top_p=top_p,
839
+ **kwargs
840
+ )
841
+
842
+
843
+ @add_start_docstrings(
844
+ """
845
+ The InternLM Model transformer with a sequence classification head on top (linear layer).
846
+
847
+ [`InternLMForSequenceClassification`] uses the last token in order to do the classification, as other causal models
848
+ (e.g. GPT-2) do.
849
+
850
+ Since it does classification on the last token, it requires to know the position of the last token. If a
851
+ `pad_token_id` is defined in the configuration, it finds the last token that is not a padding token in each row. If
852
+ no `pad_token_id` is defined, it simply takes the last value in each row of the batch. Since it cannot guess the
853
+ padding tokens when `inputs_embeds` are passed instead of `input_ids`, it does the same (take the last value in
854
+ each row of the batch).
855
+ """,
856
+ INTERNLM_START_DOCSTRING,
857
+ )
858
+ class InternLMForSequenceClassification(InternLMPreTrainedModel):
859
+ _keys_to_ignore_on_load_missing = [r"lm_head.weight"]
860
+
861
+ def __init__(self, config):
862
+ super().__init__(config)
863
+ self.num_labels = config.num_labels
864
+ self.model = InternLMModel(config)
865
+ self.score = nn.Linear(config.hidden_size, self.num_labels, bias=False)
866
+
867
+ # Initialize weights and apply final processing
868
+ self.post_init()
869
+
870
+ def get_input_embeddings(self):
871
+ return self.model.embed_tokens
872
+
873
+ def set_input_embeddings(self, value):
874
+ self.model.embed_tokens = value
875
+
876
+ @add_start_docstrings_to_model_forward(INTERNLM_INPUTS_DOCSTRING)
877
+ def forward(
878
+ self,
879
+ input_ids: torch.LongTensor = None,
880
+ attention_mask: Optional[torch.Tensor] = None,
881
+ position_ids: Optional[torch.LongTensor] = None,
882
+ past_key_values: Optional[List[torch.FloatTensor]] = None,
883
+ inputs_embeds: Optional[torch.FloatTensor] = None,
884
+ labels: Optional[torch.LongTensor] = None,
885
+ use_cache: Optional[bool] = None,
886
+ output_attentions: Optional[bool] = None,
887
+ output_hidden_states: Optional[bool] = None,
888
+ return_dict: Optional[bool] = None,
889
+ ) -> Union[Tuple, SequenceClassifierOutputWithPast]:
890
+ r"""
891
+ labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
892
+ Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
893
+ config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
894
+ `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
895
+ """
896
+ return_dict = return_dict if return_dict is not None else self.config.use_return_dict
897
+
898
+ transformer_outputs = self.model(
899
+ input_ids,
900
+ attention_mask=attention_mask,
901
+ position_ids=position_ids,
902
+ past_key_values=past_key_values,
903
+ inputs_embeds=inputs_embeds,
904
+ use_cache=use_cache,
905
+ output_attentions=output_attentions,
906
+ output_hidden_states=output_hidden_states,
907
+ return_dict=return_dict,
908
+ )
909
+ hidden_states = transformer_outputs[0]
910
+ logits = self.score(hidden_states)
911
+
912
+ if input_ids is not None:
913
+ batch_size = input_ids.shape[0]
914
+ else:
915
+ batch_size = inputs_embeds.shape[0]
916
+
917
+ if self.config.pad_token_id is None and batch_size != 1:
918
+ raise ValueError("Cannot handle batch sizes > 1 if no padding token is defined.")
919
+ if self.config.pad_token_id is None:
920
+ sequence_lengths = -1
921
+ else:
922
+ if input_ids is not None:
923
+ sequence_lengths = (torch.ne(input_ids, self.config.pad_token_id).sum(-1) - 1).to(logits.device)
924
+ else:
925
+ sequence_lengths = -1
926
+
927
+ pooled_logits = logits[torch.arange(batch_size, device=logits.device), sequence_lengths]
928
+
929
+ loss = None
930
+ if labels is not None:
931
+ labels = labels.to(logits.device)
932
+ if self.config.problem_type is None:
933
+ if self.num_labels == 1:
934
+ self.config.problem_type = "regression"
935
+ elif self.num_labels > 1 and (labels.dtype == torch.long or labels.dtype == torch.int):
936
+ self.config.problem_type = "single_label_classification"
937
+ else:
938
+ self.config.problem_type = "multi_label_classification"
939
+
940
+ if self.config.problem_type == "regression":
941
+ loss_fct = MSELoss()
942
+ if self.num_labels == 1:
943
+ loss = loss_fct(pooled_logits.squeeze(), labels.squeeze())
944
+ else:
945
+ loss = loss_fct(pooled_logits, labels)
946
+ elif self.config.problem_type == "single_label_classification":
947
+ loss_fct = CrossEntropyLoss()
948
+ loss = loss_fct(pooled_logits.view(-1, self.num_labels), labels.view(-1))
949
+ elif self.config.problem_type == "multi_label_classification":
950
+ loss_fct = BCEWithLogitsLoss()
951
+ loss = loss_fct(pooled_logits, labels)
952
+ if not return_dict:
953
+ output = (pooled_logits,) + transformer_outputs[1:]
954
+ return ((loss,) + output) if loss is not None else output
955
+
956
+ return SequenceClassifierOutputWithPast(
957
+ loss=loss,
958
+ logits=pooled_logits,
959
+ past_key_values=transformer_outputs.past_key_values,
960
+ hidden_states=transformer_outputs.hidden_states,
961
+ attentions=transformer_outputs.attentions,
962
+ )
pytorch_model-00001-of-00002.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fc4f657d36b380a22bbf5f854a2b05b0392749eed902d2f251aa7ad921867ef7
3
+ size 9975288936
pytorch_model-00002-of-00002.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:351cbe1b91fbc1162a4b69607b3a9456fce60e08d6f3a60e49e76f7ce853f3c5
3
+ size 4668775771
pytorch_model.bin.index.json ADDED
@@ -0,0 +1,458 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metadata": {
3
+ "total_size": 14643904512
4
+ },
5
+ "weight_map": {
6
+ "lm_head.weight": "pytorch_model-00002-of-00002.bin",
7
+ "model.embed_tokens.weight": "pytorch_model-00001-of-00002.bin",
8
+ "model.layers.0.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
9
+ "model.layers.0.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
10
+ "model.layers.0.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
11
+ "model.layers.0.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
12
+ "model.layers.0.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
13
+ "model.layers.0.self_attn.k_proj.bias": "pytorch_model-00001-of-00002.bin",
14
+ "model.layers.0.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
15
+ "model.layers.0.self_attn.o_proj.bias": "pytorch_model-00001-of-00002.bin",
16
+ "model.layers.0.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
17
+ "model.layers.0.self_attn.q_proj.bias": "pytorch_model-00001-of-00002.bin",
18
+ "model.layers.0.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
19
+ "model.layers.0.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
20
+ "model.layers.0.self_attn.v_proj.bias": "pytorch_model-00001-of-00002.bin",
21
+ "model.layers.0.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
22
+ "model.layers.1.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
23
+ "model.layers.1.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
24
+ "model.layers.1.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
25
+ "model.layers.1.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
26
+ "model.layers.1.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
27
+ "model.layers.1.self_attn.k_proj.bias": "pytorch_model-00001-of-00002.bin",
28
+ "model.layers.1.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
29
+ "model.layers.1.self_attn.o_proj.bias": "pytorch_model-00001-of-00002.bin",
30
+ "model.layers.1.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
31
+ "model.layers.1.self_attn.q_proj.bias": "pytorch_model-00001-of-00002.bin",
32
+ "model.layers.1.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
33
+ "model.layers.1.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
34
+ "model.layers.1.self_attn.v_proj.bias": "pytorch_model-00001-of-00002.bin",
35
+ "model.layers.1.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
36
+ "model.layers.10.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
37
+ "model.layers.10.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
38
+ "model.layers.10.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
39
+ "model.layers.10.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
40
+ "model.layers.10.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
41
+ "model.layers.10.self_attn.k_proj.bias": "pytorch_model-00001-of-00002.bin",
42
+ "model.layers.10.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
43
+ "model.layers.10.self_attn.o_proj.bias": "pytorch_model-00001-of-00002.bin",
44
+ "model.layers.10.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
45
+ "model.layers.10.self_attn.q_proj.bias": "pytorch_model-00001-of-00002.bin",
46
+ "model.layers.10.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
47
+ "model.layers.10.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
48
+ "model.layers.10.self_attn.v_proj.bias": "pytorch_model-00001-of-00002.bin",
49
+ "model.layers.10.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
50
+ "model.layers.11.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
51
+ "model.layers.11.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
52
+ "model.layers.11.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
53
+ "model.layers.11.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
54
+ "model.layers.11.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
55
+ "model.layers.11.self_attn.k_proj.bias": "pytorch_model-00001-of-00002.bin",
56
+ "model.layers.11.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
57
+ "model.layers.11.self_attn.o_proj.bias": "pytorch_model-00001-of-00002.bin",
58
+ "model.layers.11.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
59
+ "model.layers.11.self_attn.q_proj.bias": "pytorch_model-00001-of-00002.bin",
60
+ "model.layers.11.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
61
+ "model.layers.11.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
62
+ "model.layers.11.self_attn.v_proj.bias": "pytorch_model-00001-of-00002.bin",
63
+ "model.layers.11.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
64
+ "model.layers.12.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
65
+ "model.layers.12.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
66
+ "model.layers.12.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
67
+ "model.layers.12.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
68
+ "model.layers.12.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
69
+ "model.layers.12.self_attn.k_proj.bias": "pytorch_model-00001-of-00002.bin",
70
+ "model.layers.12.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
71
+ "model.layers.12.self_attn.o_proj.bias": "pytorch_model-00001-of-00002.bin",
72
+ "model.layers.12.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
73
+ "model.layers.12.self_attn.q_proj.bias": "pytorch_model-00001-of-00002.bin",
74
+ "model.layers.12.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
75
+ "model.layers.12.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
76
+ "model.layers.12.self_attn.v_proj.bias": "pytorch_model-00001-of-00002.bin",
77
+ "model.layers.12.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
78
+ "model.layers.13.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
79
+ "model.layers.13.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
80
+ "model.layers.13.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
81
+ "model.layers.13.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
82
+ "model.layers.13.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
83
+ "model.layers.13.self_attn.k_proj.bias": "pytorch_model-00001-of-00002.bin",
84
+ "model.layers.13.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
85
+ "model.layers.13.self_attn.o_proj.bias": "pytorch_model-00001-of-00002.bin",
86
+ "model.layers.13.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
87
+ "model.layers.13.self_attn.q_proj.bias": "pytorch_model-00001-of-00002.bin",
88
+ "model.layers.13.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
89
+ "model.layers.13.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
90
+ "model.layers.13.self_attn.v_proj.bias": "pytorch_model-00001-of-00002.bin",
91
+ "model.layers.13.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
92
+ "model.layers.14.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
93
+ "model.layers.14.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
94
+ "model.layers.14.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
95
+ "model.layers.14.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
96
+ "model.layers.14.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
97
+ "model.layers.14.self_attn.k_proj.bias": "pytorch_model-00001-of-00002.bin",
98
+ "model.layers.14.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
99
+ "model.layers.14.self_attn.o_proj.bias": "pytorch_model-00001-of-00002.bin",
100
+ "model.layers.14.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
101
+ "model.layers.14.self_attn.q_proj.bias": "pytorch_model-00001-of-00002.bin",
102
+ "model.layers.14.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
103
+ "model.layers.14.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
104
+ "model.layers.14.self_attn.v_proj.bias": "pytorch_model-00001-of-00002.bin",
105
+ "model.layers.14.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
106
+ "model.layers.15.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
107
+ "model.layers.15.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
108
+ "model.layers.15.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
109
+ "model.layers.15.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
110
+ "model.layers.15.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
111
+ "model.layers.15.self_attn.k_proj.bias": "pytorch_model-00001-of-00002.bin",
112
+ "model.layers.15.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
113
+ "model.layers.15.self_attn.o_proj.bias": "pytorch_model-00001-of-00002.bin",
114
+ "model.layers.15.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
115
+ "model.layers.15.self_attn.q_proj.bias": "pytorch_model-00001-of-00002.bin",
116
+ "model.layers.15.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
117
+ "model.layers.15.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
118
+ "model.layers.15.self_attn.v_proj.bias": "pytorch_model-00001-of-00002.bin",
119
+ "model.layers.15.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
120
+ "model.layers.16.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
121
+ "model.layers.16.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
122
+ "model.layers.16.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
123
+ "model.layers.16.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
124
+ "model.layers.16.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
125
+ "model.layers.16.self_attn.k_proj.bias": "pytorch_model-00001-of-00002.bin",
126
+ "model.layers.16.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
127
+ "model.layers.16.self_attn.o_proj.bias": "pytorch_model-00001-of-00002.bin",
128
+ "model.layers.16.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
129
+ "model.layers.16.self_attn.q_proj.bias": "pytorch_model-00001-of-00002.bin",
130
+ "model.layers.16.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
131
+ "model.layers.16.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
132
+ "model.layers.16.self_attn.v_proj.bias": "pytorch_model-00001-of-00002.bin",
133
+ "model.layers.16.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
134
+ "model.layers.17.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
135
+ "model.layers.17.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
136
+ "model.layers.17.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
137
+ "model.layers.17.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
138
+ "model.layers.17.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
139
+ "model.layers.17.self_attn.k_proj.bias": "pytorch_model-00001-of-00002.bin",
140
+ "model.layers.17.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
141
+ "model.layers.17.self_attn.o_proj.bias": "pytorch_model-00001-of-00002.bin",
142
+ "model.layers.17.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
143
+ "model.layers.17.self_attn.q_proj.bias": "pytorch_model-00001-of-00002.bin",
144
+ "model.layers.17.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
145
+ "model.layers.17.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
146
+ "model.layers.17.self_attn.v_proj.bias": "pytorch_model-00001-of-00002.bin",
147
+ "model.layers.17.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
148
+ "model.layers.18.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
149
+ "model.layers.18.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
150
+ "model.layers.18.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
151
+ "model.layers.18.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
152
+ "model.layers.18.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
153
+ "model.layers.18.self_attn.k_proj.bias": "pytorch_model-00001-of-00002.bin",
154
+ "model.layers.18.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
155
+ "model.layers.18.self_attn.o_proj.bias": "pytorch_model-00001-of-00002.bin",
156
+ "model.layers.18.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
157
+ "model.layers.18.self_attn.q_proj.bias": "pytorch_model-00001-of-00002.bin",
158
+ "model.layers.18.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
159
+ "model.layers.18.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
160
+ "model.layers.18.self_attn.v_proj.bias": "pytorch_model-00001-of-00002.bin",
161
+ "model.layers.18.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
162
+ "model.layers.19.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
163
+ "model.layers.19.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
164
+ "model.layers.19.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
165
+ "model.layers.19.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
166
+ "model.layers.19.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
167
+ "model.layers.19.self_attn.k_proj.bias": "pytorch_model-00001-of-00002.bin",
168
+ "model.layers.19.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
169
+ "model.layers.19.self_attn.o_proj.bias": "pytorch_model-00001-of-00002.bin",
170
+ "model.layers.19.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
171
+ "model.layers.19.self_attn.q_proj.bias": "pytorch_model-00001-of-00002.bin",
172
+ "model.layers.19.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
173
+ "model.layers.19.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
174
+ "model.layers.19.self_attn.v_proj.bias": "pytorch_model-00001-of-00002.bin",
175
+ "model.layers.19.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
176
+ "model.layers.2.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
177
+ "model.layers.2.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
178
+ "model.layers.2.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
179
+ "model.layers.2.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
180
+ "model.layers.2.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
181
+ "model.layers.2.self_attn.k_proj.bias": "pytorch_model-00001-of-00002.bin",
182
+ "model.layers.2.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
183
+ "model.layers.2.self_attn.o_proj.bias": "pytorch_model-00001-of-00002.bin",
184
+ "model.layers.2.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
185
+ "model.layers.2.self_attn.q_proj.bias": "pytorch_model-00001-of-00002.bin",
186
+ "model.layers.2.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
187
+ "model.layers.2.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
188
+ "model.layers.2.self_attn.v_proj.bias": "pytorch_model-00001-of-00002.bin",
189
+ "model.layers.2.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
190
+ "model.layers.20.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
191
+ "model.layers.20.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
192
+ "model.layers.20.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
193
+ "model.layers.20.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
194
+ "model.layers.20.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
195
+ "model.layers.20.self_attn.k_proj.bias": "pytorch_model-00001-of-00002.bin",
196
+ "model.layers.20.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
197
+ "model.layers.20.self_attn.o_proj.bias": "pytorch_model-00001-of-00002.bin",
198
+ "model.layers.20.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
199
+ "model.layers.20.self_attn.q_proj.bias": "pytorch_model-00001-of-00002.bin",
200
+ "model.layers.20.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
201
+ "model.layers.20.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
202
+ "model.layers.20.self_attn.v_proj.bias": "pytorch_model-00001-of-00002.bin",
203
+ "model.layers.20.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
204
+ "model.layers.21.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
205
+ "model.layers.21.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
206
+ "model.layers.21.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
207
+ "model.layers.21.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
208
+ "model.layers.21.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
209
+ "model.layers.21.self_attn.k_proj.bias": "pytorch_model-00001-of-00002.bin",
210
+ "model.layers.21.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
211
+ "model.layers.21.self_attn.o_proj.bias": "pytorch_model-00001-of-00002.bin",
212
+ "model.layers.21.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
213
+ "model.layers.21.self_attn.q_proj.bias": "pytorch_model-00001-of-00002.bin",
214
+ "model.layers.21.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
215
+ "model.layers.21.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
216
+ "model.layers.21.self_attn.v_proj.bias": "pytorch_model-00001-of-00002.bin",
217
+ "model.layers.21.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
218
+ "model.layers.22.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
219
+ "model.layers.22.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
220
+ "model.layers.22.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
221
+ "model.layers.22.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
222
+ "model.layers.22.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
223
+ "model.layers.22.self_attn.k_proj.bias": "pytorch_model-00001-of-00002.bin",
224
+ "model.layers.22.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
225
+ "model.layers.22.self_attn.o_proj.bias": "pytorch_model-00001-of-00002.bin",
226
+ "model.layers.22.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
227
+ "model.layers.22.self_attn.q_proj.bias": "pytorch_model-00001-of-00002.bin",
228
+ "model.layers.22.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
229
+ "model.layers.22.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
230
+ "model.layers.22.self_attn.v_proj.bias": "pytorch_model-00001-of-00002.bin",
231
+ "model.layers.22.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
232
+ "model.layers.23.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
233
+ "model.layers.23.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
234
+ "model.layers.23.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
235
+ "model.layers.23.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
236
+ "model.layers.23.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
237
+ "model.layers.23.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
238
+ "model.layers.23.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
239
+ "model.layers.23.self_attn.o_proj.bias": "pytorch_model-00002-of-00002.bin",
240
+ "model.layers.23.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
241
+ "model.layers.23.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
242
+ "model.layers.23.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
243
+ "model.layers.23.self_attn.rotary_emb.inv_freq": "pytorch_model-00002-of-00002.bin",
244
+ "model.layers.23.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
245
+ "model.layers.23.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
246
+ "model.layers.24.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
247
+ "model.layers.24.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
248
+ "model.layers.24.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
249
+ "model.layers.24.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
250
+ "model.layers.24.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
251
+ "model.layers.24.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
252
+ "model.layers.24.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
253
+ "model.layers.24.self_attn.o_proj.bias": "pytorch_model-00002-of-00002.bin",
254
+ "model.layers.24.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
255
+ "model.layers.24.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
256
+ "model.layers.24.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
257
+ "model.layers.24.self_attn.rotary_emb.inv_freq": "pytorch_model-00002-of-00002.bin",
258
+ "model.layers.24.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
259
+ "model.layers.24.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
260
+ "model.layers.25.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
261
+ "model.layers.25.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
262
+ "model.layers.25.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
263
+ "model.layers.25.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
264
+ "model.layers.25.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
265
+ "model.layers.25.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
266
+ "model.layers.25.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
267
+ "model.layers.25.self_attn.o_proj.bias": "pytorch_model-00002-of-00002.bin",
268
+ "model.layers.25.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
269
+ "model.layers.25.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
270
+ "model.layers.25.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
271
+ "model.layers.25.self_attn.rotary_emb.inv_freq": "pytorch_model-00002-of-00002.bin",
272
+ "model.layers.25.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
273
+ "model.layers.25.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
274
+ "model.layers.26.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
275
+ "model.layers.26.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
276
+ "model.layers.26.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
277
+ "model.layers.26.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
278
+ "model.layers.26.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
279
+ "model.layers.26.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
280
+ "model.layers.26.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
281
+ "model.layers.26.self_attn.o_proj.bias": "pytorch_model-00002-of-00002.bin",
282
+ "model.layers.26.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
283
+ "model.layers.26.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
284
+ "model.layers.26.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
285
+ "model.layers.26.self_attn.rotary_emb.inv_freq": "pytorch_model-00002-of-00002.bin",
286
+ "model.layers.26.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
287
+ "model.layers.26.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
288
+ "model.layers.27.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
289
+ "model.layers.27.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
290
+ "model.layers.27.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
291
+ "model.layers.27.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
292
+ "model.layers.27.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
293
+ "model.layers.27.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
294
+ "model.layers.27.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
295
+ "model.layers.27.self_attn.o_proj.bias": "pytorch_model-00002-of-00002.bin",
296
+ "model.layers.27.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
297
+ "model.layers.27.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
298
+ "model.layers.27.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
299
+ "model.layers.27.self_attn.rotary_emb.inv_freq": "pytorch_model-00002-of-00002.bin",
300
+ "model.layers.27.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
301
+ "model.layers.27.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
302
+ "model.layers.28.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
303
+ "model.layers.28.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
304
+ "model.layers.28.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
305
+ "model.layers.28.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
306
+ "model.layers.28.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
307
+ "model.layers.28.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
308
+ "model.layers.28.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
309
+ "model.layers.28.self_attn.o_proj.bias": "pytorch_model-00002-of-00002.bin",
310
+ "model.layers.28.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
311
+ "model.layers.28.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
312
+ "model.layers.28.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
313
+ "model.layers.28.self_attn.rotary_emb.inv_freq": "pytorch_model-00002-of-00002.bin",
314
+ "model.layers.28.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
315
+ "model.layers.28.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
316
+ "model.layers.29.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
317
+ "model.layers.29.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
318
+ "model.layers.29.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
319
+ "model.layers.29.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
320
+ "model.layers.29.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
321
+ "model.layers.29.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
322
+ "model.layers.29.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
323
+ "model.layers.29.self_attn.o_proj.bias": "pytorch_model-00002-of-00002.bin",
324
+ "model.layers.29.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
325
+ "model.layers.29.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
326
+ "model.layers.29.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
327
+ "model.layers.29.self_attn.rotary_emb.inv_freq": "pytorch_model-00002-of-00002.bin",
328
+ "model.layers.29.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
329
+ "model.layers.29.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
330
+ "model.layers.3.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
331
+ "model.layers.3.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
332
+ "model.layers.3.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
333
+ "model.layers.3.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
334
+ "model.layers.3.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
335
+ "model.layers.3.self_attn.k_proj.bias": "pytorch_model-00001-of-00002.bin",
336
+ "model.layers.3.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
337
+ "model.layers.3.self_attn.o_proj.bias": "pytorch_model-00001-of-00002.bin",
338
+ "model.layers.3.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
339
+ "model.layers.3.self_attn.q_proj.bias": "pytorch_model-00001-of-00002.bin",
340
+ "model.layers.3.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
341
+ "model.layers.3.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
342
+ "model.layers.3.self_attn.v_proj.bias": "pytorch_model-00001-of-00002.bin",
343
+ "model.layers.3.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
344
+ "model.layers.30.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
345
+ "model.layers.30.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
346
+ "model.layers.30.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
347
+ "model.layers.30.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
348
+ "model.layers.30.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
349
+ "model.layers.30.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
350
+ "model.layers.30.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
351
+ "model.layers.30.self_attn.o_proj.bias": "pytorch_model-00002-of-00002.bin",
352
+ "model.layers.30.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
353
+ "model.layers.30.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
354
+ "model.layers.30.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
355
+ "model.layers.30.self_attn.rotary_emb.inv_freq": "pytorch_model-00002-of-00002.bin",
356
+ "model.layers.30.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
357
+ "model.layers.30.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
358
+ "model.layers.31.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
359
+ "model.layers.31.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
360
+ "model.layers.31.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
361
+ "model.layers.31.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
362
+ "model.layers.31.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
363
+ "model.layers.31.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
364
+ "model.layers.31.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
365
+ "model.layers.31.self_attn.o_proj.bias": "pytorch_model-00002-of-00002.bin",
366
+ "model.layers.31.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
367
+ "model.layers.31.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
368
+ "model.layers.31.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
369
+ "model.layers.31.self_attn.rotary_emb.inv_freq": "pytorch_model-00002-of-00002.bin",
370
+ "model.layers.31.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
371
+ "model.layers.31.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
372
+ "model.layers.4.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
373
+ "model.layers.4.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
374
+ "model.layers.4.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
375
+ "model.layers.4.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
376
+ "model.layers.4.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
377
+ "model.layers.4.self_attn.k_proj.bias": "pytorch_model-00001-of-00002.bin",
378
+ "model.layers.4.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
379
+ "model.layers.4.self_attn.o_proj.bias": "pytorch_model-00001-of-00002.bin",
380
+ "model.layers.4.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
381
+ "model.layers.4.self_attn.q_proj.bias": "pytorch_model-00001-of-00002.bin",
382
+ "model.layers.4.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
383
+ "model.layers.4.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
384
+ "model.layers.4.self_attn.v_proj.bias": "pytorch_model-00001-of-00002.bin",
385
+ "model.layers.4.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
386
+ "model.layers.5.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
387
+ "model.layers.5.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
388
+ "model.layers.5.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
389
+ "model.layers.5.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
390
+ "model.layers.5.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
391
+ "model.layers.5.self_attn.k_proj.bias": "pytorch_model-00001-of-00002.bin",
392
+ "model.layers.5.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
393
+ "model.layers.5.self_attn.o_proj.bias": "pytorch_model-00001-of-00002.bin",
394
+ "model.layers.5.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
395
+ "model.layers.5.self_attn.q_proj.bias": "pytorch_model-00001-of-00002.bin",
396
+ "model.layers.5.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
397
+ "model.layers.5.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
398
+ "model.layers.5.self_attn.v_proj.bias": "pytorch_model-00001-of-00002.bin",
399
+ "model.layers.5.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
400
+ "model.layers.6.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
401
+ "model.layers.6.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
402
+ "model.layers.6.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
403
+ "model.layers.6.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
404
+ "model.layers.6.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
405
+ "model.layers.6.self_attn.k_proj.bias": "pytorch_model-00001-of-00002.bin",
406
+ "model.layers.6.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
407
+ "model.layers.6.self_attn.o_proj.bias": "pytorch_model-00001-of-00002.bin",
408
+ "model.layers.6.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
409
+ "model.layers.6.self_attn.q_proj.bias": "pytorch_model-00001-of-00002.bin",
410
+ "model.layers.6.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
411
+ "model.layers.6.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
412
+ "model.layers.6.self_attn.v_proj.bias": "pytorch_model-00001-of-00002.bin",
413
+ "model.layers.6.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
414
+ "model.layers.7.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
415
+ "model.layers.7.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
416
+ "model.layers.7.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
417
+ "model.layers.7.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
418
+ "model.layers.7.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
419
+ "model.layers.7.self_attn.k_proj.bias": "pytorch_model-00001-of-00002.bin",
420
+ "model.layers.7.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
421
+ "model.layers.7.self_attn.o_proj.bias": "pytorch_model-00001-of-00002.bin",
422
+ "model.layers.7.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
423
+ "model.layers.7.self_attn.q_proj.bias": "pytorch_model-00001-of-00002.bin",
424
+ "model.layers.7.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
425
+ "model.layers.7.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
426
+ "model.layers.7.self_attn.v_proj.bias": "pytorch_model-00001-of-00002.bin",
427
+ "model.layers.7.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
428
+ "model.layers.8.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
429
+ "model.layers.8.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
430
+ "model.layers.8.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
431
+ "model.layers.8.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
432
+ "model.layers.8.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
433
+ "model.layers.8.self_attn.k_proj.bias": "pytorch_model-00001-of-00002.bin",
434
+ "model.layers.8.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
435
+ "model.layers.8.self_attn.o_proj.bias": "pytorch_model-00001-of-00002.bin",
436
+ "model.layers.8.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
437
+ "model.layers.8.self_attn.q_proj.bias": "pytorch_model-00001-of-00002.bin",
438
+ "model.layers.8.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
439
+ "model.layers.8.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
440
+ "model.layers.8.self_attn.v_proj.bias": "pytorch_model-00001-of-00002.bin",
441
+ "model.layers.8.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
442
+ "model.layers.9.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
443
+ "model.layers.9.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
444
+ "model.layers.9.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
445
+ "model.layers.9.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
446
+ "model.layers.9.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
447
+ "model.layers.9.self_attn.k_proj.bias": "pytorch_model-00001-of-00002.bin",
448
+ "model.layers.9.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
449
+ "model.layers.9.self_attn.o_proj.bias": "pytorch_model-00001-of-00002.bin",
450
+ "model.layers.9.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
451
+ "model.layers.9.self_attn.q_proj.bias": "pytorch_model-00001-of-00002.bin",
452
+ "model.layers.9.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
453
+ "model.layers.9.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
454
+ "model.layers.9.self_attn.v_proj.bias": "pytorch_model-00001-of-00002.bin",
455
+ "model.layers.9.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
456
+ "model.norm.weight": "pytorch_model-00002-of-00002.bin"
457
+ }
458
+ }
special_tokens_map.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": "<s>",
3
+ "eos_token": "</s>",
4
+ "pad_token": "</s>",
5
+ "unk_token": "<unk>"
6
+ }
tokenization_internlm.py ADDED
@@ -0,0 +1,242 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # coding=utf-8
2
+ # Copyright 2022 EleutherAI and the HuggingFace Inc. team. All rights reserved.
3
+ #
4
+ # This code is based on EleutherAI's GPT-NeoX library and the GPT-NeoX
5
+ # and OPT implementations in this library. It has been modified from its
6
+ # original forms to accommodate minor architectural differences compared
7
+ # to GPT-NeoX and OPT used by the Meta AI team that trained the model.
8
+ #
9
+ # Licensed under the Apache License, Version 2.0 (the "License");
10
+ # you may not use this file except in compliance with the License.
11
+ # You may obtain a copy of the License at
12
+ #
13
+ # http://www.apache.org/licenses/LICENSE-2.0
14
+ #
15
+ # Unless required by applicable law or agreed to in writing, software
16
+ # distributed under the License is distributed on an "AS IS" BASIS,
17
+ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
18
+ # See the License for the specific language governing permissions and
19
+ # limitations under the License.
20
+
21
+ """Tokenization classes for IntermLM."""
22
+ import os
23
+ from shutil import copyfile
24
+ from typing import Any, Dict, List, Optional, Tuple
25
+
26
+ import sentencepiece as spm
27
+
28
+ from transformers.tokenization_utils import PreTrainedTokenizer
29
+ from transformers.utils import logging
30
+
31
+
32
+ logger = logging.get_logger(__name__)
33
+
34
+ VOCAB_FILES_NAMES = {"vocab_file": "./tokenizer.model"}
35
+
36
+ PRETRAINED_VOCAB_FILES_MAP = {}
37
+
38
+
39
+ class InternLMTokenizer(PreTrainedTokenizer):
40
+ """
41
+ Construct a InternLM tokenizer. Based on byte-level Byte-Pair-Encoding.
42
+
43
+ Args:
44
+ vocab_file (`str`):
45
+ Path to the vocabulary file.
46
+ """
47
+
48
+ vocab_files_names = VOCAB_FILES_NAMES
49
+ pretrained_vocab_files_map = PRETRAINED_VOCAB_FILES_MAP
50
+ model_input_names = ["input_ids", "attention_mask"]
51
+ _auto_class = "AutoTokenizer"
52
+
53
+ def __init__(
54
+ self,
55
+ vocab_file,
56
+ unk_token="<unk>",
57
+ bos_token="<s>",
58
+ eos_token="</s>",
59
+ pad_token="</s>",
60
+ sp_model_kwargs: Optional[Dict[str, Any]] = None,
61
+ add_bos_token=True,
62
+ add_eos_token=False,
63
+ decode_with_prefix_space=False,
64
+ clean_up_tokenization_spaces=False,
65
+ **kwargs,
66
+ ):
67
+ self.sp_model_kwargs = {} if sp_model_kwargs is None else sp_model_kwargs
68
+ super().__init__(
69
+ bos_token=bos_token,
70
+ eos_token=eos_token,
71
+ unk_token=unk_token,
72
+ pad_token=pad_token,
73
+ clean_up_tokenization_spaces=clean_up_tokenization_spaces,
74
+ **kwargs,
75
+ )
76
+ self.vocab_file = vocab_file
77
+ self.add_bos_token = add_bos_token
78
+ self.add_eos_token = add_eos_token
79
+ self.decode_with_prefix_space = decode_with_prefix_space
80
+ self.sp_model = spm.SentencePieceProcessor(**self.sp_model_kwargs)
81
+ self.sp_model.Load(vocab_file)
82
+ self._no_prefix_space_tokens = None
83
+
84
+ """ Initialisation"""
85
+
86
+ @property
87
+ def no_prefix_space_tokens(self):
88
+ if self._no_prefix_space_tokens is None:
89
+ vocab = self.convert_ids_to_tokens(list(range(self.vocab_size)))
90
+ self._no_prefix_space_tokens = {i for i, tok in enumerate(vocab) if not tok.startswith("▁")}
91
+ return self._no_prefix_space_tokens
92
+
93
+ @property
94
+ def vocab_size(self):
95
+ """Returns vocab size"""
96
+ return self.sp_model.get_piece_size()
97
+
98
+ @property
99
+ def bos_token_id(self) -> Optional[int]:
100
+ return self.sp_model.bos_id()
101
+
102
+ @property
103
+ def eos_token_id(self) -> Optional[int]:
104
+ return self.sp_model.eos_id()
105
+
106
+ def get_vocab(self):
107
+ """Returns vocab as a dict"""
108
+ vocab = {self.convert_ids_to_tokens(i): i for i in range(self.vocab_size)}
109
+ vocab.update(self.added_tokens_encoder)
110
+ return vocab
111
+
112
+ def _tokenize(self, text):
113
+ """Returns a tokenized string."""
114
+ return self.sp_model.encode(text, out_type=str)
115
+
116
+ def _convert_token_to_id(self, token):
117
+ """Converts a token (str) in an id using the vocab."""
118
+ return self.sp_model.piece_to_id(token)
119
+
120
+ def _convert_id_to_token(self, index):
121
+ """Converts an index (integer) in a token (str) using the vocab."""
122
+ token = self.sp_model.IdToPiece(index)
123
+ return token
124
+
125
+ def _maybe_add_prefix_space(self, tokens, decoded):
126
+ if tokens and tokens[0] not in self.no_prefix_space_tokens:
127
+ return " " + decoded
128
+ else:
129
+ return decoded
130
+
131
+ def convert_tokens_to_string(self, tokens):
132
+ """Converts a sequence of tokens (string) in a single string."""
133
+ current_sub_tokens = []
134
+ out_string = ""
135
+ prev_is_special = False
136
+ for token in tokens:
137
+ # make sure that special tokens are not decoded using sentencepiece model
138
+ if token in self.all_special_tokens:
139
+ if not prev_is_special:
140
+ out_string += " "
141
+ out_string += self.sp_model.decode(current_sub_tokens) + token
142
+ prev_is_special = True
143
+ current_sub_tokens = []
144
+ else:
145
+ current_sub_tokens.append(token)
146
+ prev_is_special = False
147
+ out_string += self.sp_model.decode(current_sub_tokens)
148
+ out_string = self.clean_up_tokenization(out_string)
149
+ out_string = self._maybe_add_prefix_space(tokens=tokens, decoded=out_string)
150
+ return out_string[1:]
151
+
152
+ def save_vocabulary(self, save_directory, filename_prefix: Optional[str] = None) -> Tuple[str]:
153
+ """
154
+ Save the vocabulary and special tokens file to a directory.
155
+
156
+ Args:
157
+ save_directory (`str`):
158
+ The directory in which to save the vocabulary.
159
+
160
+ Returns:
161
+ `Tuple(str)`: Paths to the files saved.
162
+ """
163
+ if not os.path.isdir(save_directory):
164
+ logger.error(f"Vocabulary path ({save_directory}) should be a directory")
165
+ return
166
+ out_vocab_file = os.path.join(
167
+ save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
168
+ )
169
+
170
+ if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file) and os.path.isfile(self.vocab_file):
171
+ copyfile(self.vocab_file, out_vocab_file)
172
+ elif not os.path.isfile(self.vocab_file):
173
+ with open(out_vocab_file, "wb") as fi:
174
+ content_spiece_model = self.sp_model.serialized_model_proto()
175
+ fi.write(content_spiece_model)
176
+
177
+ return (out_vocab_file,)
178
+
179
+ def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=None):
180
+ if self.add_bos_token:
181
+ bos_token_ids = [self.bos_token_id]
182
+ else:
183
+ bos_token_ids = []
184
+
185
+ output = bos_token_ids + token_ids_0
186
+
187
+ if token_ids_1 is not None:
188
+ output = output + token_ids_1
189
+
190
+ if self.add_eos_token:
191
+ output = output + [self.eos_token_id]
192
+
193
+ return output
194
+
195
+ def get_special_tokens_mask(
196
+ self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None, already_has_special_tokens: bool = False
197
+ ) -> List[int]:
198
+ """
199
+ Retrieve sequence ids from a token list that has no special tokens added. This method is called when adding
200
+ special tokens using the tokenizer `prepare_for_model` method.
201
+
202
+ Args:
203
+ token_ids_0 (`List[int]`):
204
+ List of IDs.
205
+ token_ids_1 (`List[int]`, *optional*):
206
+ Optional second list of IDs for sequence pairs.
207
+ already_has_special_tokens (`bool`, *optional*, defaults to `False`):
208
+ Whether or not the token list is already formatted with special tokens for the model.
209
+
210
+ Returns:
211
+ `List[int]`: A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token.
212
+ """
213
+ if already_has_special_tokens:
214
+ return super().get_special_tokens_mask(
215
+ token_ids_0=token_ids_0, token_ids_1=token_ids_1, already_has_special_tokens=True
216
+ )
217
+
218
+ if token_ids_1 is None:
219
+ return [1] + ([0] * len(token_ids_0)) + [1]
220
+ return [1] + ([0] * len(token_ids_0)) + [1, 1] + ([0] * len(token_ids_1)) + [1]
221
+
222
+ def create_token_type_ids_from_sequences(
223
+ self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None
224
+ ) -> List[int]:
225
+ """
226
+ Create a mask from the two sequences passed to be used in a sequence-pair classification task. T5 does not make
227
+ use of token type ids, therefore a list of zeros is returned.
228
+
229
+ Args:
230
+ token_ids_0 (`List[int]`):
231
+ List of IDs.
232
+ token_ids_1 (`List[int]`, *optional*):
233
+ Optional second list of IDs for sequence pairs.
234
+
235
+ Returns:
236
+ `List[int]`: List of zeros.
237
+ """
238
+ eos = [self.eos_token_id]
239
+
240
+ if token_ids_1 is None:
241
+ return len(token_ids_0 + eos) * [0]
242
+ return len(token_ids_0 + eos + token_ids_1 + eos) * [0]
tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aab622d98c98677a1a51f969e25765154487bf3e85c7819db105db2fcacba83f
3
+ size 1658691
tokenizer_config.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "auto_map": {
3
+ "AutoTokenizer": [
4
+ "tokenization_internlm.InternLMTokenizer",
5
+ null
6
+ ]
7
+ },
8
+ "bos_token": "<s>",
9
+ "clean_up_tokenization_spaces": false,
10
+ "eos_token": "</s>",
11
+ "model_max_length": 1000000000000000019884624838656,
12
+ "pad_token": "</s>",
13
+ "tokenizer_class": "InternLMTokenizer",
14
+ "unk_token": "<unk>"
15
+ }