Upload folder using huggingface_hub

README.md

@@ -0,0 +1,97 @@
+---
+license: mit
+datasets:
+- lemonilia/LimaRP
+- PygmalionAI/PIPPA
+language:
+- en
+pipeline_tag: text-generation
+tags:
+- roleplay
+---
+
+**Deepsex-34b**
+
+gguf:https://huggingface.co/zzlgreat/deepsex-34b-gguf
+exl2:https://huggingface.co/waldie/deepsex-34b-4bpw-h6-exl2
+
+Here are the steps to make this model:
+1. I first collected a total collection of about 4GB of various light novels, and used BERT to perform two rounds of similarity deduplication on the novels with similar plots in the data set. In addition, a portion of nsfw novels are mixed in to improve the NSFW capabilities of the model.
+2. Then use the YI-34B-base as the base of the model, use the setting of r=64 alpha=128 and use qlora to fine-tune 3 epochs for continuous pre-training.
+3. Prepare the limarp+pippa data set, clean it into alpaca format, and use [goliath-120b](https://huggingface.co/alpindale/goliath-120b), which is good at role-playing, to score each question and answer pair, and filter out the high-quality ones. 30k data.
+4. Use the data in 3 for sft on the base model obtained in 2, 6 epochs, r=16 alpha=32 for fine-tuning.
+
+*Format*
+
+alpaca
+
+```
+[
+  {
+    "instruction": "user instruction (required)",
+    "input": "user input (optional)",
+    "output": "model response (required)",
+    "history": [
+      ["user instruction in the first round (optional)", "model response in the first round (optional)"],
+      ["user instruction in the second round (optional)", "model response in the second round (optional)"]
+    ]
+  }
+]
+```
+
+*Effect*:
+Proficient in role-playing skills, while being highly accepted on NSFW, pure love words will appear from time to time. like:
+```#3
+Sweaty old man
+December 5, 2023 2:03 PM
+Fxxk, you are such a xxx!
+
+#4
+27.3s
+Mirai
+December 5, 2023 2:03 PM
+"Of course I do! I can't break promises, Sweaty old man. We have been together since we were kids. We are both best friends and lovers to end all iteration." I smiled with affection. It was clear that I meant everything I said. "We both know that you like taking command of us like this. Am I not your squirting toy, Sweaty old man?" I asked with a cute pout. "We should meet up in front of the shop after classes. I'll see you there. See you, Sweaty old man!"
+```
+It feels like it’s still worth a try~
+
+Support me [here](https://ko-fi.com/mikolisa) :)
+
+
+**Deepsex-34b**
+
+*步骤*
+1. 我先收集了各种轻小说大约4GB的总集，通过bert对该数据集中剧情比较相似的小说进行了两轮相似度去重。另外混入了一部分nsfw小说以提高该模型的NSFW能力。
+2. 然后将该模型以YI-34B-base为基座，使用r=64 alpha=128 的设置使用qlora微调了3个epochs来进行持续预训练。
+3. 准备limarp+pippa数据集，统一清洗为alpaca格式，并且使用比较擅长角色扮演的[goliath-120b](https://huggingface.co/alpindale/goliath-120b)对每个问答对进行打分，筛选出其中质量高的大约30k数据。
+4. 对2中得到的base模型使用3中的数据进行sft，6个epochs，r=16 alpha=32进行微调。
+
+*格式*
+
+alpaca
+```[
+  {
+    "instruction": "user instruction (required)",
+    "input": "user input (optional)",
+    "output": "model response (required)",
+    "history": [
+      ["user instruction in the first round (optional)", "model response in the first round (optional)"],
+      ["user instruction in the second round (optional)", "model response in the second round (optional)"]
+    ]
+  }
+]```
+
+*效果*
+熟练的角色扮演技能，在NSFW上有很高接受度的同时，会时不时的出现纯爱的话语。如：
+```#3
+Sweaty old man
+December 5, 2023 2:03 PM
+Fxxk, you are such a xxx!
+
+#4
+27.3s
+Mirai
+December 5, 2023 2:03 PM
+"Of course I do! I can't break promises, Sweaty old man. We have been together since we were kids. We are both best friends and lovers to end all iteration." I smiled with affection. It was clear that I meant everything I said. "We both know that you like taking command of us like this. Am I not your squirting toy, Sweaty old man?" I asked with a cute pout. "We should meet up in front of the shop after classes. I'll see you there. See you, Sweaty old man!"
+```
+感觉还是很值得一试的~
+如果觉得好用，欢迎支持我一杯 [咖啡](https://ko-fi.com/mikolisa) :)

config.json

@@ -0,0 +1,30 @@
+{
+  "_name_or_path": "/DATA4T/text-generation-webui/models/yi-sex",
+  "architectures": [
+    "YiForCausalLM"
+  ],
+  "auto_map": {
+    "AutoConfig": "configuration_yi.YiConfig",
+    "AutoModel": "modeling_yi.YiForCausalLM",
+    "AutoModelForCausalLM": "modeling_yi.YiForCausalLM"
+  },
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "hidden_act": "silu",
+  "hidden_size": 7168,
+  "initializer_range": 0.02,
+  "intermediate_size": 20480,
+  "max_position_embeddings": 4096,
+  "model_type": "Yi",
+  "num_attention_heads": 56,
+  "num_hidden_layers": 60,
+  "num_key_value_heads": 8,
+  "pad_token_id": 0,
+  "rms_norm_eps": 1e-05,
+  "rope_theta": 5000000.0,
+  "tie_word_embeddings": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.34.1",
+  "use_cache": true,
+  "vocab_size": 64000
+}

configuration_yi.py

@@ -0,0 +1,121 @@
+""" Yi model configuration"""
+from transformers.configuration_utils import PretrainedConfig
+from transformers.utils import logging
+
+logger = logging.get_logger(__name__)
+
+Yi_PRETRAINED_CONFIG_ARCHIVE_MAP = {}
+
+
+class YiConfig(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`YiModel`]. It is used to instantiate an Yi
+    model according to the specified arguments, defining the model architecture. Instantiating a configuration with the
+    defaults will yield a similar configuration to that of the Yi model.
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+
+    Args:
+        vocab_size (`int`, *optional*, defaults to 64000):
+            Vocabulary size of the Yi model. Defines the number of different tokens that can be represented by the
+            `inputs_ids` passed when calling [`YiModel`]
+        hidden_size (`int`, *optional*, defaults to 4096):
+            Dimension of the hidden representations.
+        intermediate_size (`int`, *optional*, defaults to 11008):
+            Dimension of the MLP representations.
+        num_hidden_layers (`int`, *optional*, defaults to 32):
+            Number of hidden layers in the Transformer encoder.
+        num_attention_heads (`int`, *optional*, defaults to 32):
+            Number of attention heads for each attention layer in the Transformer encoder.
+        num_key_value_heads (`int`, *optional*):
+            This is the number of key_value heads that should be used to implement Grouped Query Attention. If
+            `num_key_value_heads=num_attention_heads`, the model will use Multi Head Attention (MHA), if
+            `num_key_value_heads=1 the model will use Multi Query Attention (MQA) otherwise GQA is used. When
+            converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed
+            by meanpooling all the original heads within that group. For more details checkout [this
+            paper](https://arxiv.org/pdf/2305.13245.pdf). If it is not specified, will default to
+            `num_attention_heads`.
+        hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
+            The non-linear activation function (function or string) in the decoder.
+        max_position_embeddings (`int`, *optional*, defaults to 4096):
+            The maximum sequence length that this model might ever be used with. Typically set this to something large
+            just in case (e.g., 512 or 1024 or 2048 or 4096).
+        initializer_range (`float`, *optional*, defaults to 0.02):
+            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
+        rms_norm_eps (`float`, *optional*, defaults to 1e-5):
+            The epsilon used by the rms normalization layers.
+        use_cache (`bool`, *optional*, defaults to `True`):
+            Whether or not the model should return the last key/values attentions (not used by all models). Only
+            relevant if `config.is_decoder=True`.
+        tie_word_embeddings(`bool`, *optional*, defaults to `False`):
+            Whether to tie weight embeddings
+        output_attentions (`bool`, *optional*, defaults to `False`):
+            Whether or not to output attentions.
+        rope_theta (`float`, *optional*, defaults to 5000000.0):
+            The base period of the RoPE embeddings.
+        Example:
+
+    ```python
+    >>> from transformers import YiModel, YiConfig
+
+    >>> # Initializing a Yi style configuration
+    >>> configuration = YiConfig()
+
+    >>> # Initializing a model from the Yi style configuration
+    >>> model = YiModel(configuration)
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+    model_type = "Yi"
+    keys_to_ignore_at_inference = ["past_key_values"]
+
+    def __init__(
+        self,
+        vocab_size=64000,
+        hidden_size=4096,
+        intermediate_size=11008,
+        num_hidden_layers=32,
+        num_attention_heads=32,
+        num_key_value_heads=4,
+        hidden_act="silu",
+        max_position_embeddings=4096,
+        initializer_range=0.02,
+        rms_norm_eps=1e-5,
+        use_cache=True,
+        pad_token_id=0,
+        bos_token_id=1,
+        eos_token_id=2,
+        tie_word_embeddings=False,
+        output_attentions=False,
+        rope_theta=5000000.0,
+        **kwargs,
+    ):
+        self.vocab_size = vocab_size
+        self.max_position_embeddings = max_position_embeddings
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+
+        # for backward compatibility
+        if num_key_value_heads is None:
+            num_key_value_heads = num_attention_heads
+
+        self.num_key_value_heads = num_key_value_heads
+        self.hidden_act = hidden_act
+        self.initializer_range = initializer_range
+        self.rms_norm_eps = rms_norm_eps
+        self.use_cache = use_cache
+        self.output_attentions = output_attentions
+        self.rope_theta = rope_theta
+
+        super().__init__(
+            pad_token_id=pad_token_id,
+            bos_token_id=bos_token_id,
+            eos_token_id=eos_token_id,
+            tie_word_embeddings=tie_word_embeddings,
+            **kwargs,
+        )

generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "pad_token_id": 0,
+  "transformers_version": "4.34.1"
+}

modeling_yi.py

@@ -0,0 +1,1030 @@
+""" PyTorch Yi model."""
+import math
+from typing import List, Optional, Tuple, Union
+
+import torch.utils.checkpoint
+from einops import repeat
+from torch import nn
+from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
+
+from transformers.activations import ACT2FN
+from transformers.modeling_outputs import (
+    BaseModelOutputWithPast,
+    CausalLMOutputWithPast,
+    SequenceClassifierOutputWithPast,
+)
+from transformers.modeling_utils import PreTrainedModel
+from transformers.pytorch_utils import ALL_LAYERNORM_LAYERS
+from transformers.utils import (
+    add_start_docstrings,
+    add_start_docstrings_to_model_forward,
+    is_flash_attn_available,
+    logging,
+    replace_return_docstrings,
+)
+
+from .configuration_yi import YiConfig
+
+
+if is_flash_attn_available():
+    from flash_attn import flash_attn_func
+
+
+logger = logging.get_logger(__name__)
+
+_CONFIG_FOR_DOC = "YiConfig"
+
+
+# Copied from transformers.models.bart.modeling_bart._make_causal_mask
+def _make_causal_mask(
+    input_ids_shape: torch.Size,
+    dtype: torch.dtype,
+    device: torch.device,
+    past_key_values_length: int = 0,
+):
+    """
+    Make causal mask used for bi-directional self-attention.
+    """
+    bsz, tgt_len = input_ids_shape
+    mask = torch.full(
+        (tgt_len, tgt_len),
+        torch.tensor(torch.finfo(dtype).min, device=device),
+        device=device,
+    )
+    mask_cond = torch.arange(mask.size(-1), device=device)
+    mask.masked_fill_(mask_cond < (mask_cond + 1).view(mask.size(-1), 1), 0)
+    mask = mask.to(dtype)
+
+    if past_key_values_length > 0:
+        mask = torch.cat(
+            [
+                torch.zeros(
+                    tgt_len, past_key_values_length, dtype=dtype, device=device
+                ),
+                mask,
+            ],
+            dim=-1,
+        )
+    return mask[None, None, :, :].expand(
+        bsz, 1, tgt_len, tgt_len + past_key_values_length
+    )
+
+
+# Copied from transformers.models.bart.modeling_bart._expand_mask
+def _expand_mask(mask: torch.Tensor, dtype: torch.dtype, tgt_len: Optional[int] = None):
+    """
+    Expands attention_mask from `[bsz, seq_len]` to `[bsz, 1, tgt_seq_len, src_seq_len]`.
+    """
+    bsz, src_len = mask.size()
+    tgt_len = tgt_len if tgt_len is not None else src_len
+
+    expanded_mask = mask[:, None, None, :].expand(bsz, 1, tgt_len, src_len).to(dtype)
+
+    inverted_mask = 1.0 - expanded_mask
+
+    return inverted_mask.masked_fill(
+        inverted_mask.to(torch.bool), torch.finfo(dtype).min
+    )
+
+
+class YiRMSNorm(nn.Module):
+    def __init__(self, hidden_size, eps=1e-5):
+        """
+        YiRMSNorm is equivalent to T5LayerNorm
+        """
+        super().__init__()
+        self.weight = nn.Parameter(torch.ones(hidden_size))
+        self.variance_epsilon = eps
+
+    def forward(self, hidden_states):
+        input_dtype = hidden_states.dtype
+        hidden_states = hidden_states.to(torch.float32)
+        variance = hidden_states.pow(2).mean(-1, keepdim=True)
+        hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)
+
+        return self.weight * hidden_states.to(input_dtype)
+
+
+ALL_LAYERNORM_LAYERS.append(YiRMSNorm)
+
+
+class YiRotaryEmbedding(torch.nn.Module):
+    def __init__(self, dim, max_position_embeddings=4096, base=5000000, device=None):
+        super().__init__()
+
+        self.dim = dim
+        self.max_position_embeddings = max_position_embeddings
+        self.base = base
+
+        # Build here to make `torch.jit.trace` work.
+        self._set_cos_sin_cache(seq_len=max_position_embeddings, device=device)
+
+    def _set_cos_sin_cache(self, seq_len, device):
+        self.max_seq_len_cached = seq_len
+        inv_freq = 1.0 / (
+            self.base ** (torch.arange(0, self.dim, 2).float().to(device) / self.dim)
+        )
+        t = torch.arange(self.max_seq_len_cached, device=device, dtype=torch.float32)
+        freqs = torch.einsum("i,j->ij", t, inv_freq)
+        # Different from paper, but it uses a different permutation in order to obtain the same calculation
+        emb = torch.cat((freqs, freqs), dim=-1)
+        self.register_buffer(
+            "cos_cached", emb.cos()[None, None, :, :], persistent=False
+        )
+        self.register_buffer(
+            "sin_cached", emb.sin()[None, None, :, :], persistent=False
+        )
+
+    def forward(self, x, seq_len=None):
+        # x: [bs, num_attention_heads, seq_len, head_size]
+        if seq_len > self.max_seq_len_cached:
+            self._set_cos_sin_cache(seq_len=seq_len, device=x.device)
+
+        return (
+            self.cos_cached[:, :, :seq_len, ...].to(dtype=x.dtype),
+            self.sin_cached[:, :, :seq_len, ...].to(dtype=x.dtype),
+        )
+
+
+def rotate_half(x):
+    """Rotates half the hidden dims of the input."""
+    x1 = x[..., : x.shape[-1] // 2]
+    x2 = x[..., x.shape[-1] // 2 :]
+    return torch.cat((-x2, x1), dim=-1)
+
+
+def apply_rotary_pos_emb(q, k, cos, sin, position_ids, flash_attn_available):
+    # The first two dimensions of cos and sin are always 1, so we can `squeeze` them.
+    cos = cos.squeeze(1).squeeze(0)  # [seq_len, dim]
+    sin = sin.squeeze(1).squeeze(0)  # [seq_len, dim]
+    expand_dim = 2 if flash_attn_available else 1
+    cos = cos[position_ids].unsqueeze(expand_dim)  # [bs, seq_len, dim]
+    sin = sin[position_ids].unsqueeze(expand_dim)  # [bs, seq_len, dim]
+    q_embed = (q * cos) + (rotate_half(q) * sin)
+    k_embed = (k * cos) + (rotate_half(k) * sin)
+    return q_embed, k_embed
+
+
+class YiMLP(nn.Module):
+    def __init__(self, hidden_size: int, intermediate_size: int, hidden_act: str):
+        super().__init__()
+        self.gate_proj = nn.Linear(hidden_size, intermediate_size, bias=False)
+        self.down_proj = nn.Linear(intermediate_size, hidden_size, bias=False)
+        self.up_proj = nn.Linear(hidden_size, intermediate_size, bias=False)
+        self.act_fn = ACT2FN[hidden_act]
+
+    def forward(self, x):
+        return self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))
+
+
+class YiAttention(nn.Module):
+    """Multi-headed attention from 'Attention Is All You Need' paper"""
+
+    def __init__(self, config: YiConfig):
+        super().__init__()
+        self.config = config
+        self.hidden_size = config.hidden_size
+        self.num_heads = config.num_attention_heads
+        self.head_dim = self.hidden_size // self.num_heads
+        self.num_key_value_heads = config.num_key_value_heads
+        self.num_key_value_groups = self.num_heads // self.num_key_value_heads
+        self.max_position_embeddings = config.max_position_embeddings
+
+        if (self.head_dim * self.num_heads) != self.hidden_size:
+            raise ValueError(
+                f"hidden_size must be divisible by num_heads (got `hidden_size`: {self.hidden_size}"
+                f" and `num_heads`: {self.num_heads})."
+            )
+        self.q_proj = nn.Linear(
+            self.hidden_size, self.num_heads * self.head_dim, bias=False
+        )
+        self.k_proj = nn.Linear(
+            self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False
+        )
+        self.v_proj = nn.Linear(
+            self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False
+        )
+        self.o_proj = nn.Linear(
+            self.num_heads * self.head_dim, self.hidden_size, bias=False
+        )
+
+        self.rotary_emb = YiRotaryEmbedding(
+            self.head_dim,
+            max_position_embeddings=self.max_position_embeddings,
+            base=self.config.rope_theta,
+        )
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Tuple[torch.Tensor]] = None,
+        output_attentions: bool = False,
+        use_cache: bool = False,
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+        bsz, q_len, _ = hidden_states.size()
+        flash_attn_available = is_flash_attn_available()
+
+        query_states = self.q_proj(hidden_states).view(
+            bsz, q_len, self.num_heads, self.head_dim
+        )
+
+        key_states = self.k_proj(hidden_states).view(
+            bsz, q_len, self.num_key_value_heads, self.head_dim
+        )
+        value_states = self.v_proj(hidden_states).view(
+            bsz, q_len, self.num_key_value_heads, self.head_dim
+        )
+
+        if not flash_attn_available:
+            if self.num_key_value_groups > 1:
+                key_states = repeat(
+                    key_states, f"b n h d -> b n (h {self.num_key_value_groups}) d"
+                )
+                value_states = repeat(
+                    value_states, f"b n h d -> b n (h {self.num_key_value_groups}) d"
+                )
+
+            # b n h d -> b h n d
+            query_states = query_states.transpose(1, 2)
+            key_states = key_states.transpose(1, 2)
+            value_states = value_states.transpose(1, 2)
+
+        seq_dim = 1 if flash_attn_available else 2
+        kv_seq_len = key_states.shape[seq_dim]
+        if past_key_value is not None:
+            kv_seq_len += past_key_value[0].shape[seq_dim]
+        cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
+        query_states, key_states = apply_rotary_pos_emb(
+            query_states, key_states, cos, sin, position_ids, flash_attn_available
+        )
+
+        if past_key_value is not None:
+            # reuse k, v, self_attention
+            key_states = torch.cat([past_key_value[0], key_states], dim=seq_dim)
+            value_states = torch.cat([past_key_value[1], value_states], dim=seq_dim)
+
+        past_key_value = (key_states, value_states) if use_cache else None
+
+        if flash_attn_available:
+            attn_output = flash_attn_func(
+                query_states, key_states, value_states, dropout_p=0.0, causal=True
+            )
+        else:
+            attn_weights = torch.matmul(
+                query_states, key_states.transpose(2, 3)
+            ) / math.sqrt(self.head_dim)
+
+            if attn_weights.size() != (bsz, self.num_heads, q_len, kv_seq_len):
+                raise ValueError(
+                    f"Attention weights should be of size {(bsz, self.num_heads, q_len, kv_seq_len)}, but is"
+                    f" {attn_weights.size()}"
+                )
+
+            if attention_mask is not None:
+                if attention_mask.size() != (bsz, 1, q_len, kv_seq_len):
+                    raise ValueError(
+                        f"Attention mask should be of size {(bsz, 1, q_len, kv_seq_len)}, but is"
+                        f"{attention_mask.size()}"
+                    )
+                attn_weights = attn_weights + attention_mask
+                dtype_min = torch.tensor(
+                    torch.finfo(attn_weights.dtype).min,
+                    device=attn_weights.device,
+                    dtype=attn_weights.dtype,
+                )
+                attn_weights = torch.max(attn_weights, dtype_min)
+
+            # upcast attention to fp32
+            attn_weights = nn.functional.softmax(
+                attn_weights, dim=-1, dtype=torch.float32
+            ).to(query_states.dtype)
+            attn_output = torch.matmul(attn_weights, value_states)
+
+            if attn_output.size() != (bsz, self.num_heads, q_len, self.head_dim):
+                raise ValueError(
+                    f"`attn_output` should be of size {(bsz, self.num_heads, q_len, self.head_dim)}, but is"
+                    f" {attn_output.size()}"
+                )
+
+        if not flash_attn_available:
+            attn_output = attn_output.transpose(1, 2)
+
+        attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
+
+        attn_output = self.o_proj(attn_output)
+
+        if not output_attentions:
+            attn_weights = None
+
+        return attn_output, attn_weights, past_key_value
+
+
+class YiDecoderLayer(nn.Module):
+    def __init__(self, config: YiConfig):
+        super().__init__()
+
+        self.hidden_size = config.hidden_size
+        self.self_attn = YiAttention(config=config)
+        self.mlp = YiMLP(
+            hidden_size=self.hidden_size,
+            intermediate_size=config.intermediate_size,
+            hidden_act=config.hidden_act,
+        )
+
+        self.ln1 = YiRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.ln2 = YiRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Tuple[torch.Tensor]] = None,
+        output_attentions: Optional[bool] = False,
+        use_cache: Optional[bool] = False,
+    ) -> Tuple[
+        torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]
+    ]:
+        """
+        Args:
+            hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
+            attention_mask (`torch.FloatTensor`, *optional*): attention mask of size
+                `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
+            output_attentions (`bool`, *optional*):
+                Whether or not to return the attentions tensors of all attention layers. See `attentions` under
+                returned tensors for more detail.
+            use_cache (`bool`, *optional*):
+                If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding
+                (see `past_key_values`).
+            past_key_value (`Tuple(torch.FloatTensor)`, *optional*): cached past key and value projection states
+        """
+
+        residual = hidden_states
+
+        hidden_states = self.ln1(hidden_states)
+
+        # Self Attention
+        hidden_states, self_attn_weights, present_key_value = self.self_attn(
+            hidden_states=hidden_states,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_value=past_key_value,
+            output_attentions=output_attentions,
+            use_cache=use_cache,
+        )
+        hidden_states = residual + hidden_states
+
+        # Fully Connected
+        residual = hidden_states
+        hidden_states = self.ln2(hidden_states)
+        hidden_states = self.mlp(hidden_states)
+        hidden_states = residual + hidden_states
+
+        outputs = (hidden_states,)
+
+        if output_attentions:
+            outputs += (self_attn_weights,)
+
+        if use_cache:
+            outputs += (present_key_value,)
+
+        return outputs
+
+
+Yi_START_DOCSTRING = r"""
+    This model inherits from [`PreTrainedModel`]. Check the superclass documentation for the generic methods the
+    library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
+    etc.)
+
+    This model is also a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) subclass.
+    Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage
+    and behavior.
+
+    Parameters:
+        config ([`YiConfig`]):
+            Model configuration class with all the parameters of the model. Initializing with a config file does not
+            load the weights associated with the model, only the configuration. Check out the
+            [`~PreTrainedModel.from_pretrained`] method to load the model weights.
+"""
+
+
+@add_start_docstrings(
+    "The bare Yi Model outputting raw hidden-states without any specific head on top.",
+    Yi_START_DOCSTRING,
+)
+class YiPreTrainedModel(PreTrainedModel):
+    config_class = YiConfig
+    base_model_prefix = "model"
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["YiDecoderLayer"]
+    _skip_keys_device_placement = "past_key_values"
+
+    def _init_weights(self, module):
+        std = self.config.initializer_range
+        if isinstance(module, nn.Linear):
+            module.weight.data.normal_(mean=0.0, std=std)
+            if module.bias is not None:
+                module.bias.data.zero_()
+        elif isinstance(module, nn.Embedding):
+            module.weight.data.normal_(mean=0.0, std=std)
+            if module.padding_idx is not None:
+                module.weight.data[module.padding_idx].zero_()
+
+    def _set_gradient_checkpointing(self, module, value=False):
+        if isinstance(module, YiModel):
+            module.gradient_checkpointing = value
+
+
+Yi_INPUTS_DOCSTRING = r"""
+    Args:
+        input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
+            Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide
+            it.
+
+            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
+            [`PreTrainedTokenizer.__call__`] for details.
+
+            [What are input IDs?](../glossary#input-ids)
+        attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
+
+            - 1 for tokens that are **not masked**,
+            - 0 for tokens that are **masked**.
+
+            [What are attention masks?](../glossary#attention-mask)
+
+            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
+            [`PreTrainedTokenizer.__call__`] for details.
+
+            If `past_key_values` is used, optionally only the last `decoder_input_ids` have to be input (see
+            `past_key_values`).
+
+            If you want to change padding behavior, you should read [`modeling_opt._prepare_decoder_attention_mask`]
+            and modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more
+            information on the default strategy.
+
+            - 1 indicates the head is **not masked**,
+            - 0 indicates the head is **masked**.
+        position_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0,
+            config.n_positions - 1]`.
+
+            [What are position IDs?](../glossary#position-ids)
+        past_key_values (`tuple(tuple(torch.FloatTensor))`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`):
+            Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of shape
+            `(batch_size, num_heads, sequence_length, embed_size_per_head)`) and 2 additional tensors of shape
+            `(batch_size, num_heads, encoder_sequence_length, embed_size_per_head)`.
+
+            Contains pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention
+            blocks) that can be used (see `past_key_values` input) to speed up sequential decoding.
+
+            If `past_key_values` are used, the user can optionally input only the last `decoder_input_ids` (those that
+            don't have their past key value states given to this model) of shape `(batch_size, 1)` instead of all
+            `decoder_input_ids` of shape `(batch_size, sequence_length)`.
+        inputs_embeds (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
+            Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation. This
+            is useful if you want more control over how to convert `input_ids` indices into associated vectors than the
+            model's internal embedding lookup matrix.
+        use_cache (`bool`, *optional*):
+            If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see
+            `past_key_values`).
+        output_attentions (`bool`, *optional*):
+            Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned
+            tensors for more detail.
+        output_hidden_states (`bool`, *optional*):
+            Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
+            more detail.
+        return_dict (`bool`, *optional*):
+            Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
+"""
+
+
+@add_start_docstrings(
+    "The bare Yi Model outputting raw hidden-states without any specific head on top.",
+    Yi_START_DOCSTRING,
+)
+class YiModel(YiPreTrainedModel):
+    """
+    Transformer decoder consisting of *config.num_hidden_layers* layers. Each layer is a [`YiDecoderLayer`]
+
+    Args:
+        config: YiConfig
+    """
+
+    def __init__(self, config: YiConfig):
+        super().__init__(config)
+        self.padding_idx = config.pad_token_id
+        self.vocab_size = config.vocab_size
+
+        self.embed_tokens = nn.Embedding(
+            config.vocab_size, config.hidden_size, self.padding_idx
+        )
+        self.layers = nn.ModuleList(
+            [YiDecoderLayer(config) for _ in range(config.num_hidden_layers)]
+        )
+
+        self.norm = YiRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+
+        self.gradient_checkpointing = False
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.embed_tokens
+
+    def set_input_embeddings(self, value):
+        self.embed_tokens = value
+
+    # Copied from transformers.models.bart.modeling_bart.BartDecoder._prepare_decoder_attention_mask
+    def _prepare_decoder_attention_mask(
+        self, attention_mask, input_ids, inputs_embeds, past_key_values_length
+    ):
+        input_shape = input_ids.shape
+        # create causal mask
+        # [bsz, seq_len] -> [bsz, 1, tgt_seq_len, src_seq_len]
+        combined_attention_mask = None
+        if input_shape[-1] > 1:
+            combined_attention_mask = _make_causal_mask(
+                input_shape,
+                inputs_embeds.dtype,
+                device=inputs_embeds.device,
+                past_key_values_length=past_key_values_length,
+            )
+
+        if attention_mask is not None:
+            # [bsz, seq_len] -> [bsz, 1, tgt_seq_len, src_seq_len]
+            expanded_attn_mask = _expand_mask(
+                attention_mask, inputs_embeds.dtype, tgt_len=input_shape[-1]
+            ).to(inputs_embeds.device)
+            combined_attention_mask = (
+                expanded_attn_mask
+                if combined_attention_mask is None
+                else expanded_attn_mask + combined_attention_mask
+            )
+
+        return combined_attention_mask
+
+    @add_start_docstrings_to_model_forward(Yi_INPUTS_DOCSTRING)
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[List[torch.FloatTensor]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, BaseModelOutputWithPast]:
+        output_attentions = (
+            output_attentions
+            if output_attentions is not None
+            else self.config.output_attentions
+        )
+        output_hidden_states = (
+            output_hidden_states
+            if output_hidden_states is not None
+            else self.config.output_hidden_states
+        )
+        use_cache = use_cache if use_cache is not None else self.config.use_cache
+
+        return_dict = (
+            return_dict if return_dict is not None else self.config.use_return_dict
+        )
+
+        # retrieve input_ids and inputs_embeds
+        if input_ids is not None and inputs_embeds is not None:
+            raise ValueError(
+                "You cannot specify both decoder_input_ids and decoder_inputs_embeds at the same time"
+            )
+        elif input_ids is not None:
+            batch_size, seq_length = input_ids.shape
+        elif inputs_embeds is not None:
+            batch_size, seq_length, _ = inputs_embeds.shape
+        else:
+            raise ValueError(
+                "You have to specify either decoder_input_ids or decoder_inputs_embeds"
+            )
+
+        seq_length_with_past = seq_length
+        past_key_values_length = 0
+
+        if past_key_values is not None:
+            past_key_values_length = past_key_values[0][0].shape[2]
+            seq_length_with_past = seq_length_with_past + past_key_values_length
+
+        if position_ids is None:
+            device = input_ids.device if input_ids is not None else inputs_embeds.device
+            position_ids = torch.arange(
+                past_key_values_length,
+                seq_length + past_key_values_length,
+                dtype=torch.long,
+                device=device,
+            )
+            position_ids = position_ids.unsqueeze(0).view(-1, seq_length)
+        else:
+            position_ids = position_ids.view(-1, seq_length).long()
+
+        if inputs_embeds is None:
+            inputs_embeds = self.embed_tokens(input_ids)
+
+        if not is_flash_attn_available():
+            # embed positions
+            if attention_mask is None:
+                attention_mask = torch.ones(
+                    (batch_size, seq_length_with_past),
+                    dtype=torch.bool,
+                    device=inputs_embeds.device,
+                )
+            attention_mask = self._prepare_decoder_attention_mask(
+                attention_mask,
+                input_ids,
+                inputs_embeds,
+                past_key_values_length,
+            )
+        else:
+            attention_mask = None
+
+        hidden_states = inputs_embeds
+        if self.gradient_checkpointing and self.training:
+            if use_cache:
+                logger.warning_once(
+                    "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..."
+                )
+                use_cache = False
+
+        # decoder layers
+        all_hidden_states = () if output_hidden_states else None
+        all_self_attns = () if output_attentions else None
+        next_decoder_cache = () if use_cache else None
+
+        for idx, decoder_layer in enumerate(self.layers):
+            if output_hidden_states:
+                all_hidden_states += (hidden_states,)
+
+            past_key_value = (
+                past_key_values[idx] if past_key_values is not None else None
+            )
+
+            if self.gradient_checkpointing and self.training:
+
+                def create_custom_forward(module):
+                    def custom_forward(*inputs):
+                        # None for past_key_value
+                        return module(*inputs, past_key_value, output_attentions)
+
+                    return custom_forward
+
+                layer_outputs = torch.utils.checkpoint.checkpoint(
+                    create_custom_forward(decoder_layer),
+                    hidden_states,
+                    attention_mask,
+                    position_ids,
+                )
+            else:
+                layer_outputs = decoder_layer(
+                    hidden_states,
+                    attention_mask=attention_mask,
+                    position_ids=position_ids,
+                    past_key_value=past_key_value,
+                    output_attentions=output_attentions,
+                    use_cache=use_cache,
+                )
+
+            hidden_states = layer_outputs[0]
+
+            if use_cache:
+                next_decoder_cache += (layer_outputs[2 if output_attentions else 1],)
+
+            if output_attentions:
+                all_self_attns += (layer_outputs[1],)
+
+        hidden_states = self.norm(hidden_states)
+        # add hidden states from the last decoder layer
+        if output_hidden_states:
+            all_hidden_states += (hidden_states,)
+
+        next_cache = next_decoder_cache if use_cache else None
+        if not return_dict:
+            return tuple(
+                v
+                for v in [hidden_states, next_cache, all_hidden_states, all_self_attns]
+                if v is not None
+            )
+        return BaseModelOutputWithPast(
+            last_hidden_state=hidden_states,
+            past_key_values=next_cache,
+            hidden_states=all_hidden_states,
+            attentions=all_self_attns,
+        )
+
+
+class YiForCausalLM(YiPreTrainedModel):
+    _tied_weights_keys = ["lm_head.weight"]
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.model = YiModel(config)
+
+        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.model.embed_tokens
+
+    def set_input_embeddings(self, value):
+        self.model.embed_tokens = value
+
+    def get_output_embeddings(self):
+        return self.lm_head
+
+    def set_output_embeddings(self, new_embeddings):
+        self.lm_head = new_embeddings
+
+    def set_decoder(self, decoder):
+        self.model = decoder
+
+    def get_decoder(self):
+        return self.model
+
+    @add_start_docstrings_to_model_forward(Yi_INPUTS_DOCSTRING)
+    @replace_return_docstrings(
+        output_type=CausalLMOutputWithPast, config_class=_CONFIG_FOR_DOC
+    )
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[List[torch.FloatTensor]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, CausalLMOutputWithPast]:
+        r"""
+        Args:
+            labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+                Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
+                config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
+                (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`.
+
+        Returns:
+
+        Example:
+
+        ```python
+        >>> from transformers import AutoTokenizer, YiForCausalLM
+
+        >>> model = YiForCausalLM.from_pretrained(PATH_TO_CONVERTED_WEIGHTS)
+        >>> tokenizer = AutoTokenizer.from_pretrained(PATH_TO_CONVERTED_TOKENIZER)
+
+        >>> prompt = "Hey, are you conscious? Can you talk to me?"
+        >>> inputs = tokenizer(prompt, return_tensors="pt")
+
+        >>> # Generate
+        >>> generate_ids = model.generate(inputs.input_ids, max_length=30)
+        >>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
+        "Hey, are you conscious? Can you talk to me?\nI'm not conscious, but I can talk to you."
+        ```"""
+
+        output_attentions = (
+            output_attentions
+            if output_attentions is not None
+            else self.config.output_attentions
+        )
+        output_hidden_states = (
+            output_hidden_states
+            if output_hidden_states is not None
+            else self.config.output_hidden_states
+        )
+        return_dict = (
+            return_dict if return_dict is not None else self.config.use_return_dict
+        )
+
+        # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
+        outputs = self.model(
+            input_ids=input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        hidden_states = outputs[0]
+        logits = self.lm_head(hidden_states)
+
+        loss = None
+        if labels is not None:
+            # Shift so that tokens < n predict n
+            shift_logits = logits[..., :-1, :].contiguous()
+            shift_labels = labels[..., 1:].contiguous()
+            # Flatten the tokens
+            loss_fct = CrossEntropyLoss()
+            shift_logits = shift_logits.view(-1, self.config.vocab_size)
+            shift_labels = shift_labels.view(-1)
+            # Enable model parallelism
+            shift_labels = shift_labels.to(shift_logits.device)
+            loss = loss_fct(shift_logits, shift_labels)
+
+        if not return_dict:
+            output = (logits,) + outputs[1:]
+            return (loss,) + output if loss is not None else output
+
+        return CausalLMOutputWithPast(
+            loss=loss,
+            logits=logits,
+            past_key_values=outputs.past_key_values,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+    def prepare_inputs_for_generation(
+        self,
+        input_ids,
+        past_key_values=None,
+        attention_mask=None,
+        inputs_embeds=None,
+        **kwargs,
+    ):
+        if past_key_values:
+            input_ids = input_ids[:, -1:]
+
+        position_ids = kwargs.get("position_ids", None)
+        if attention_mask is not None and position_ids is None:
+            # create position_ids on the fly for batch generation
+            position_ids = attention_mask.long().cumsum(-1) - 1
+            position_ids.masked_fill_(attention_mask == 0, 1)
+            if past_key_values:
+                position_ids = position_ids[:, -1].unsqueeze(-1)
+
+        # if `inputs_embeds` are passed, we only want to use them in the 1st generation step
+        if inputs_embeds is not None and past_key_values is None:
+            model_inputs = {"inputs_embeds": inputs_embeds}
+        else:
+            model_inputs = {"input_ids": input_ids}
+
+        model_inputs.update(
+            {
+                "position_ids": position_ids,
+                "past_key_values": past_key_values,
+                "use_cache": kwargs.get("use_cache"),
+                "attention_mask": attention_mask,
+            }
+        )
+        return model_inputs
+
+    @staticmethod
+    def _reorder_cache(past_key_values, beam_idx):
+        reordered_past = ()
+        for layer_past in past_key_values:
+            reordered_past += (
+                tuple(
+                    past_state.index_select(0, beam_idx.to(past_state.device))
+                    for past_state in layer_past
+                ),
+            )
+        return reordered_past
+
+
+@add_start_docstrings(
+    """
+    The Yi Model transformer with a sequence classification head on top (linear layer).
+
+    [`YiForSequenceClassification`] uses the last token in order to do the classification, as other causal models
+    (e.g. GPT-2) do.
+
+    Since it does classification on the last token, it requires to know the position of the last token. If a
+    `pad_token_id` is defined in the configuration, it finds the last token that is not a padding token in each row. If
+    no `pad_token_id` is defined, it simply takes the last value in each row of the batch. Since it cannot guess the
+    padding tokens when `inputs_embeds` are passed instead of `input_ids`, it does the same (take the last value in
+    each row of the batch).
+    """,
+    Yi_START_DOCSTRING,
+)
+class YiForSequenceClassification(YiPreTrainedModel):
+    def __init__(self, config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+        self.model = YiModel(config)
+        self.score = nn.Linear(config.hidden_size, self.num_labels, bias=False)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.model.embed_tokens
+
+    def set_input_embeddings(self, value):
+        self.model.embed_tokens = value
+
+    @add_start_docstrings_to_model_forward(Yi_INPUTS_DOCSTRING)
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[List[torch.FloatTensor]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, SequenceClassifierOutputWithPast]:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
+            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
+            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
+            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
+        """
+        return_dict = (
+            return_dict if return_dict is not None else self.config.use_return_dict
+        )
+
+        transformer_outputs = self.model(
+            input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+        hidden_states = transformer_outputs[0]
+        logits = self.score(hidden_states)
+
+        if input_ids is not None:
+            batch_size = input_ids.shape[0]
+        else:
+            batch_size = inputs_embeds.shape[0]
+
+        if self.config.pad_token_id is None and batch_size != 1:
+            raise ValueError(
+                "Cannot handle batch sizes > 1 if no padding token is defined."
+            )
+        if self.config.pad_token_id is None:
+            sequence_lengths = -1
+        else:
+            if input_ids is not None:
+                sequence_lengths = (
+                    torch.eq(input_ids, self.config.pad_token_id).long().argmax(-1) - 1
+                ).to(logits.device)
+            else:
+                sequence_lengths = -1
+
+        pooled_logits = logits[
+            torch.arange(batch_size, device=logits.device), sequence_lengths
+        ]
+
+        loss = None
+        if labels is not None:
+            labels = labels.to(logits.device)
+            if self.config.problem_type is None:
+                if self.num_labels == 1:
+                    self.config.problem_type = "regression"
+                elif self.num_labels > 1 and (
+                    labels.dtype == torch.long or labels.dtype == torch.int
+                ):
+                    self.config.problem_type = "single_label_classification"
+                else:
+                    self.config.problem_type = "multi_label_classification"
+
+            if self.config.problem_type == "regression":
+                loss_fct = MSELoss()
+                if self.num_labels == 1:
+                    loss = loss_fct(pooled_logits.squeeze(), labels.squeeze())
+                else:
+                    loss = loss_fct(pooled_logits, labels)
+            elif self.config.problem_type == "single_label_classification":
+                loss_fct = CrossEntropyLoss()
+                loss = loss_fct(
+                    pooled_logits.view(-1, self.num_labels), labels.view(-1)
+                )
+            elif self.config.problem_type == "multi_label_classification":
+                loss_fct = BCEWithLogitsLoss()
+                loss = loss_fct(pooled_logits, labels)
+        if not return_dict:
+            output = (pooled_logits,) + transformer_outputs[1:]
+            return ((loss,) + output) if loss is not None else output
+
+        return SequenceClassifierOutputWithPast(
+            loss=loss,
+            logits=pooled_logits,
+            past_key_values=transformer_outputs.past_key_values,
+            hidden_states=transformer_outputs.hidden_states,
+            attentions=transformer_outputs.attentions,
+        )

output-00001-of-00003.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4035ba438a97dbaba01e4925fb648da745ed0b6fe326562a456e767b2e77710a
+size 8532843040

output-00002-of-00003.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:09a4769bd806c04dbb28ba160a96e627515e8d6b0f35a02aad674a8f014cf60d
+size 8523442296

output-00003-of-00003.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c9fbc6dde7b363c2fc5e1db2508047ac022cf1d3cbc4c4fae1a4512ec449100b
+size 1079114248

pytorch_model.bin.index.json

@@ -0,0 +1,550 @@
+{
+  "metadata": {
+    "total_size": 68777834496
+  },
+  "weight_map": {
+    "lm_head.weight": "pytorch_model-00077-of-00077.bin",
+    "model.embed_tokens.weight": "pytorch_model-00001-of-00077.bin",
+    "model.layers.0.ln1.weight": "pytorch_model-00003-of-00077.bin",
+    "model.layers.0.ln2.weight": "pytorch_model-00003-of-00077.bin",
+    "model.layers.0.mlp.down_proj.weight": "pytorch_model-00002-of-00077.bin",
+    "model.layers.0.mlp.gate_proj.weight": "pytorch_model-00002-of-00077.bin",
+    "model.layers.0.mlp.up_proj.weight": "pytorch_model-00003-of-00077.bin",
+    "model.layers.0.self_attn.k_proj.weight": "pytorch_model-00002-of-00077.bin",
+    "model.layers.0.self_attn.o_proj.weight": "pytorch_model-00002-of-00077.bin",
+    "model.layers.0.self_attn.q_proj.weight": "pytorch_model-00002-of-00077.bin",
+    "model.layers.0.self_attn.v_proj.weight": "pytorch_model-00002-of-00077.bin",
+    "model.layers.1.ln1.weight": "pytorch_model-00004-of-00077.bin",
+    "model.layers.1.ln2.weight": "pytorch_model-00004-of-00077.bin",
+    "model.layers.1.mlp.down_proj.weight": "pytorch_model-00004-of-00077.bin",
+    "model.layers.1.mlp.gate_proj.weight": "pytorch_model-00003-of-00077.bin",
+    "model.layers.1.mlp.up_proj.weight": "pytorch_model-00004-of-00077.bin",
+    "model.layers.1.self_attn.k_proj.weight": "pytorch_model-00003-of-00077.bin",
+    "model.layers.1.self_attn.o_proj.weight": "pytorch_model-00003-of-00077.bin",
+    "model.layers.1.self_attn.q_proj.weight": "pytorch_model-00003-of-00077.bin",
+    "model.layers.1.self_attn.v_proj.weight": "pytorch_model-00003-of-00077.bin",
+    "model.layers.10.ln1.weight": "pytorch_model-00015-of-00077.bin",
+    "model.layers.10.ln2.weight": "pytorch_model-00015-of-00077.bin",
+    "model.layers.10.mlp.down_proj.weight": "pytorch_model-00015-of-00077.bin",
+    "model.layers.10.mlp.gate_proj.weight": "pytorch_model-00015-of-00077.bin",
+    "model.layers.10.mlp.up_proj.weight": "pytorch_model-00015-of-00077.bin",
+    "model.layers.10.self_attn.k_proj.weight": "pytorch_model-00014-of-00077.bin",
+    "model.layers.10.self_attn.o_proj.weight": "pytorch_model-00014-of-00077.bin",
+    "model.layers.10.self_attn.q_proj.weight": "pytorch_model-00014-of-00077.bin",
+    "model.layers.10.self_attn.v_proj.weight": "pytorch_model-00014-of-00077.bin",
+    "model.layers.11.ln1.weight": "pytorch_model-00016-of-00077.bin",
+    "model.layers.11.ln2.weight": "pytorch_model-00016-of-00077.bin",
+    "model.layers.11.mlp.down_proj.weight": "pytorch_model-00016-of-00077.bin",
+    "model.layers.11.mlp.gate_proj.weight": "pytorch_model-00016-of-00077.bin",
+    "model.layers.11.mlp.up_proj.weight": "pytorch_model-00016-of-00077.bin",
+    "model.layers.11.self_attn.k_proj.weight": "pytorch_model-00015-of-00077.bin",
+    "model.layers.11.self_attn.o_proj.weight": "pytorch_model-00016-of-00077.bin",
+    "model.layers.11.self_attn.q_proj.weight": "pytorch_model-00015-of-00077.bin",
+    "model.layers.11.self_attn.v_proj.weight": "pytorch_model-00016-of-00077.bin",
+    "model.layers.12.ln1.weight": "pytorch_model-00018-of-00077.bin",
+    "model.layers.12.ln2.weight": "pytorch_model-00018-of-00077.bin",
+    "model.layers.12.mlp.down_proj.weight": "pytorch_model-00017-of-00077.bin",
+    "model.layers.12.mlp.gate_proj.weight": "pytorch_model-00017-of-00077.bin",
+    "model.layers.12.mlp.up_proj.weight": "pytorch_model-00018-of-00077.bin",
+    "model.layers.12.self_attn.k_proj.weight": "pytorch_model-00017-of-00077.bin",
+    "model.layers.12.self_attn.o_proj.weight": "pytorch_model-00017-of-00077.bin",
+    "model.layers.12.self_attn.q_proj.weight": "pytorch_model-00017-of-00077.bin",
+    "model.layers.12.self_attn.v_proj.weight": "pytorch_model-00017-of-00077.bin",
+    "model.layers.13.ln1.weight": "pytorch_model-00019-of-00077.bin",
+    "model.layers.13.ln2.weight": "pytorch_model-00019-of-00077.bin",
+    "model.layers.13.mlp.down_proj.weight": "pytorch_model-00019-of-00077.bin",
+    "model.layers.13.mlp.gate_proj.weight": "pytorch_model-00018-of-00077.bin",
+    "model.layers.13.mlp.up_proj.weight": "pytorch_model-00019-of-00077.bin",
+    "model.layers.13.self_attn.k_proj.weight": "pytorch_model-00018-of-00077.bin",
+    "model.layers.13.self_attn.o_proj.weight": "pytorch_model-00018-of-00077.bin",
+    "model.layers.13.self_attn.q_proj.weight": "pytorch_model-00018-of-00077.bin",
+    "model.layers.13.self_attn.v_proj.weight": "pytorch_model-00018-of-00077.bin",
+    "model.layers.14.ln1.weight": "pytorch_model-00020-of-00077.bin",
+    "model.layers.14.ln2.weight": "pytorch_model-00020-of-00077.bin",
+    "model.layers.14.mlp.down_proj.weight": "pytorch_model-00020-of-00077.bin",
+    "model.layers.14.mlp.gate_proj.weight": "pytorch_model-00020-of-00077.bin",
+    "model.layers.14.mlp.up_proj.weight": "pytorch_model-00020-of-00077.bin",
+    "model.layers.14.self_attn.k_proj.weight": "pytorch_model-00019-of-00077.bin",
+    "model.layers.14.self_attn.o_proj.weight": "pytorch_model-00019-of-00077.bin",
+    "model.layers.14.self_attn.q_proj.weight": "pytorch_model-00019-of-00077.bin",
+    "model.layers.14.self_attn.v_proj.weight": "pytorch_model-00019-of-00077.bin",
+    "model.layers.15.ln1.weight": "pytorch_model-00021-of-00077.bin",
+    "model.layers.15.ln2.weight": "pytorch_model-00021-of-00077.bin",
+    "model.layers.15.mlp.down_proj.weight": "pytorch_model-00021-of-00077.bin",
+    "model.layers.15.mlp.gate_proj.weight": "pytorch_model-00021-of-00077.bin",
+    "model.layers.15.mlp.up_proj.weight": "pytorch_model-00021-of-00077.bin",
+    "model.layers.15.self_attn.k_proj.weight": "pytorch_model-00020-of-00077.bin",
+    "model.layers.15.self_attn.o_proj.weight": "pytorch_model-00021-of-00077.bin",
+    "model.layers.15.self_attn.q_proj.weight": "pytorch_model-00020-of-00077.bin",
+    "model.layers.15.self_attn.v_proj.weight": "pytorch_model-00021-of-00077.bin",
+    "model.layers.16.ln1.weight": "pytorch_model-00023-of-00077.bin",
+    "model.layers.16.ln2.weight": "pytorch_model-00023-of-00077.bin",
+    "model.layers.16.mlp.down_proj.weight": "pytorch_model-00022-of-00077.bin",
+    "model.layers.16.mlp.gate_proj.weight": "pytorch_model-00022-of-00077.bin",
+    "model.layers.16.mlp.up_proj.weight": "pytorch_model-00023-of-00077.bin",
+    "model.layers.16.self_attn.k_proj.weight": "pytorch_model-00022-of-00077.bin",
+    "model.layers.16.self_attn.o_proj.weight": "pytorch_model-00022-of-00077.bin",
+    "model.layers.16.self_attn.q_proj.weight": "pytorch_model-00022-of-00077.bin",
+    "model.layers.16.self_attn.v_proj.weight": "pytorch_model-00022-of-00077.bin",
+    "model.layers.17.ln1.weight": "pytorch_model-00024-of-00077.bin",
+    "model.layers.17.ln2.weight": "pytorch_model-00024-of-00077.bin",
+    "model.layers.17.mlp.down_proj.weight": "pytorch_model-00024-of-00077.bin",
+    "model.layers.17.mlp.gate_proj.weight": "pytorch_model-00023-of-00077.bin",
+    "model.layers.17.mlp.up_proj.weight": "pytorch_model-00024-of-00077.bin",
+    "model.layers.17.self_attn.k_proj.weight": "pytorch_model-00023-of-00077.bin",
+    "model.layers.17.self_attn.o_proj.weight": "pytorch_model-00023-of-00077.bin",
+    "model.layers.17.self_attn.q_proj.weight": "pytorch_model-00023-of-00077.bin",
+    "model.layers.17.self_attn.v_proj.weight": "pytorch_model-00023-of-00077.bin",
+    "model.layers.18.ln1.weight": "pytorch_model-00025-of-00077.bin",
+    "model.layers.18.ln2.weight": "pytorch_model-00025-of-00077.bin",
+    "model.layers.18.mlp.down_proj.weight": "pytorch_model-00025-of-00077.bin",
+    "model.layers.18.mlp.gate_proj.weight": "pytorch_model-00025-of-00077.bin",
+    "model.layers.18.mlp.up_proj.weight": "pytorch_model-00025-of-00077.bin",
+    "model.layers.18.self_attn.k_proj.weight": "pytorch_model-00024-of-00077.bin",
+    "model.layers.18.self_attn.o_proj.weight": "pytorch_model-00024-of-00077.bin",
+    "model.layers.18.self_attn.q_proj.weight": "pytorch_model-00024-of-00077.bin",
+    "model.layers.18.self_attn.v_proj.weight": "pytorch_model-00024-of-00077.bin",
+    "model.layers.19.ln1.weight": "pytorch_model-00026-of-00077.bin",
+    "model.layers.19.ln2.weight": "pytorch_model-00026-of-00077.bin",
+    "model.layers.19.mlp.down_proj.weight": "pytorch_model-00026-of-00077.bin",
+    "model.layers.19.mlp.gate_proj.weight": "pytorch_model-00026-of-00077.bin",
+    "model.layers.19.mlp.up_proj.weight": "pytorch_model-00026-of-00077.bin",
+    "model.layers.19.self_attn.k_proj.weight": "pytorch_model-00025-of-00077.bin",
+    "model.layers.19.self_attn.o_proj.weight": "pytorch_model-00026-of-00077.bin",
+    "model.layers.19.self_attn.q_proj.weight": "pytorch_model-00025-of-00077.bin",
+    "model.layers.19.self_attn.v_proj.weight": "pytorch_model-00026-of-00077.bin",
+    "model.layers.2.ln1.weight": "pytorch_model-00005-of-00077.bin",
+    "model.layers.2.ln2.weight": "pytorch_model-00005-of-00077.bin",
+    "model.layers.2.mlp.down_proj.weight": "pytorch_model-00005-of-00077.bin",
+    "model.layers.2.mlp.gate_proj.weight": "pytorch_model-00005-of-00077.bin",
+    "model.layers.2.mlp.up_proj.weight": "pytorch_model-00005-of-00077.bin",
+    "model.layers.2.self_attn.k_proj.weight": "pytorch_model-00004-of-00077.bin",
+    "model.layers.2.self_attn.o_proj.weight": "pytorch_model-00004-of-00077.bin",
+    "model.layers.2.self_attn.q_proj.weight": "pytorch_model-00004-of-00077.bin",
+    "model.layers.2.self_attn.v_proj.weight": "pytorch_model-00004-of-00077.bin",
+    "model.layers.20.ln1.weight": "pytorch_model-00028-of-00077.bin",
+    "model.layers.20.ln2.weight": "pytorch_model-00028-of-00077.bin",
+    "model.layers.20.mlp.down_proj.weight": "pytorch_model-00027-of-00077.bin",
+    "model.layers.20.mlp.gate_proj.weight": "pytorch_model-00027-of-00077.bin",
+    "model.layers.20.mlp.up_proj.weight": "pytorch_model-00028-of-00077.bin",
+    "model.layers.20.self_attn.k_proj.weight": "pytorch_model-00027-of-00077.bin",
+    "model.layers.20.self_attn.o_proj.weight": "pytorch_model-00027-of-00077.bin",
+    "model.layers.20.self_attn.q_proj.weight": "pytorch_model-00027-of-00077.bin",
+    "model.layers.20.self_attn.v_proj.weight": "pytorch_model-00027-of-00077.bin",
+    "model.layers.21.ln1.weight": "pytorch_model-00029-of-00077.bin",
+    "model.layers.21.ln2.weight": "pytorch_model-00029-of-00077.bin",
+    "model.layers.21.mlp.down_proj.weight": "pytorch_model-00029-of-00077.bin",
+    "model.layers.21.mlp.gate_proj.weight": "pytorch_model-00028-of-00077.bin",
+    "model.layers.21.mlp.up_proj.weight": "pytorch_model-00029-of-00077.bin",
+    "model.layers.21.self_attn.k_proj.weight": "pytorch_model-00028-of-00077.bin",
+    "model.layers.21.self_attn.o_proj.weight": "pytorch_model-00028-of-00077.bin",
+    "model.layers.21.self_attn.q_proj.weight": "pytorch_model-00028-of-00077.bin",
+    "model.layers.21.self_attn.v_proj.weight": "pytorch_model-00028-of-00077.bin",
+    "model.layers.22.ln1.weight": "pytorch_model-00030-of-00077.bin",
+    "model.layers.22.ln2.weight": "pytorch_model-00030-of-00077.bin",
+    "model.layers.22.mlp.down_proj.weight": "pytorch_model-00030-of-00077.bin",
+    "model.layers.22.mlp.gate_proj.weight": "pytorch_model-00030-of-00077.bin",
+    "model.layers.22.mlp.up_proj.weight": "pytorch_model-00030-of-00077.bin",
+    "model.layers.22.self_attn.k_proj.weight": "pytorch_model-00029-of-00077.bin",
+    "model.layers.22.self_attn.o_proj.weight": "pytorch_model-00029-of-00077.bin",
+    "model.layers.22.self_attn.q_proj.weight": "pytorch_model-00029-of-00077.bin",
+    "model.layers.22.self_attn.v_proj.weight": "pytorch_model-00029-of-00077.bin",
+    "model.layers.23.ln1.weight": "pytorch_model-00031-of-00077.bin",
+    "model.layers.23.ln2.weight": "pytorch_model-00031-of-00077.bin",
+    "model.layers.23.mlp.down_proj.weight": "pytorch_model-00031-of-00077.bin",
+    "model.layers.23.mlp.gate_proj.weight": "pytorch_model-00031-of-00077.bin",
+    "model.layers.23.mlp.up_proj.weight": "pytorch_model-00031-of-00077.bin",
+    "model.layers.23.self_attn.k_proj.weight": "pytorch_model-00030-of-00077.bin",
+    "model.layers.23.self_attn.o_proj.weight": "pytorch_model-00031-of-00077.bin",
+    "model.layers.23.self_attn.q_proj.weight": "pytorch_model-00030-of-00077.bin",
+    "model.layers.23.self_attn.v_proj.weight": "pytorch_model-00031-of-00077.bin",
+    "model.layers.24.ln1.weight": "pytorch_model-00033-of-00077.bin",
+    "model.layers.24.ln2.weight": "pytorch_model-00033-of-00077.bin",
+    "model.layers.24.mlp.down_proj.weight": "pytorch_model-00032-of-00077.bin",
+    "model.layers.24.mlp.gate_proj.weight": "pytorch_model-00032-of-00077.bin",
+    "model.layers.24.mlp.up_proj.weight": "pytorch_model-00033-of-00077.bin",
+    "model.layers.24.self_attn.k_proj.weight": "pytorch_model-00032-of-00077.bin",
+    "model.layers.24.self_attn.o_proj.weight": "pytorch_model-00032-of-00077.bin",
+    "model.layers.24.self_attn.q_proj.weight": "pytorch_model-00032-of-00077.bin",
+    "model.layers.24.self_attn.v_proj.weight": "pytorch_model-00032-of-00077.bin",
+    "model.layers.25.ln1.weight": "pytorch_model-00034-of-00077.bin",
+    "model.layers.25.ln2.weight": "pytorch_model-00034-of-00077.bin",
+    "model.layers.25.mlp.down_proj.weight": "pytorch_model-00034-of-00077.bin",
+    "model.layers.25.mlp.gate_proj.weight": "pytorch_model-00033-of-00077.bin",
+    "model.layers.25.mlp.up_proj.weight": "pytorch_model-00034-of-00077.bin",
+    "model.layers.25.self_attn.k_proj.weight": "pytorch_model-00033-of-00077.bin",
+    "model.layers.25.self_attn.o_proj.weight": "pytorch_model-00033-of-00077.bin",
+    "model.layers.25.self_attn.q_proj.weight": "pytorch_model-00033-of-00077.bin",
+    "model.layers.25.self_attn.v_proj.weight": "pytorch_model-00033-of-00077.bin",
+    "model.layers.26.ln1.weight": "pytorch_model-00035-of-00077.bin",
+    "model.layers.26.ln2.weight": "pytorch_model-00035-of-00077.bin",
+    "model.layers.26.mlp.down_proj.weight": "pytorch_model-00035-of-00077.bin",
+    "model.layers.26.mlp.gate_proj.weight": "pytorch_model-00035-of-00077.bin",
+    "model.layers.26.mlp.up_proj.weight": "pytorch_model-00035-of-00077.bin",
+    "model.layers.26.self_attn.k_proj.weight": "pytorch_model-00034-of-00077.bin",
+    "model.layers.26.self_attn.o_proj.weight": "pytorch_model-00034-of-00077.bin",
+    "model.layers.26.self_attn.q_proj.weight": "pytorch_model-00034-of-00077.bin",
+    "model.layers.26.self_attn.v_proj.weight": "pytorch_model-00034-of-00077.bin",
+    "model.layers.27.ln1.weight": "pytorch_model-00036-of-00077.bin",
+    "model.layers.27.ln2.weight": "pytorch_model-00036-of-00077.bin",
+    "model.layers.27.mlp.down_proj.weight": "pytorch_model-00036-of-00077.bin",
+    "model.layers.27.mlp.gate_proj.weight": "pytorch_model-00036-of-00077.bin",
+    "model.layers.27.mlp.up_proj.weight": "pytorch_model-00036-of-00077.bin",
+    "model.layers.27.self_attn.k_proj.weight": "pytorch_model-00035-of-00077.bin",
+    "model.layers.27.self_attn.o_proj.weight": "pytorch_model-00036-of-00077.bin",
+    "model.layers.27.self_attn.q_proj.weight": "pytorch_model-00035-of-00077.bin",
+    "model.layers.27.self_attn.v_proj.weight": "pytorch_model-00036-of-00077.bin",
+    "model.layers.28.ln1.weight": "pytorch_model-00038-of-00077.bin",
+    "model.layers.28.ln2.weight": "pytorch_model-00038-of-00077.bin",
+    "model.layers.28.mlp.down_proj.weight": "pytorch_model-00037-of-00077.bin",
+    "model.layers.28.mlp.gate_proj.weight": "pytorch_model-00037-of-00077.bin",
+    "model.layers.28.mlp.up_proj.weight": "pytorch_model-00038-of-00077.bin",
+    "model.layers.28.self_attn.k_proj.weight": "pytorch_model-00037-of-00077.bin",
+    "model.layers.28.self_attn.o_proj.weight": "pytorch_model-00037-of-00077.bin",
+    "model.layers.28.self_attn.q_proj.weight": "pytorch_model-00037-of-00077.bin",
+    "model.layers.28.self_attn.v_proj.weight": "pytorch_model-00037-of-00077.bin",
+    "model.layers.29.ln1.weight": "pytorch_model-00039-of-00077.bin",
+    "model.layers.29.ln2.weight": "pytorch_model-00039-of-00077.bin",
+    "model.layers.29.mlp.down_proj.weight": "pytorch_model-00039-of-00077.bin",
+    "model.layers.29.mlp.gate_proj.weight": "pytorch_model-00038-of-00077.bin",
+    "model.layers.29.mlp.up_proj.weight": "pytorch_model-00039-of-00077.bin",
+    "model.layers.29.self_attn.k_proj.weight": "pytorch_model-00038-of-00077.bin",
+    "model.layers.29.self_attn.o_proj.weight": "pytorch_model-00038-of-00077.bin",
+    "model.layers.29.self_attn.q_proj.weight": "pytorch_model-00038-of-00077.bin",
+    "model.layers.29.self_attn.v_proj.weight": "pytorch_model-00038-of-00077.bin",
+    "model.layers.3.ln1.weight": "pytorch_model-00006-of-00077.bin",
+    "model.layers.3.ln2.weight": "pytorch_model-00006-of-00077.bin",
+    "model.layers.3.mlp.down_proj.weight": "pytorch_model-00006-of-00077.bin",
+    "model.layers.3.mlp.gate_proj.weight": "pytorch_model-00006-of-00077.bin",
+    "model.layers.3.mlp.up_proj.weight": "pytorch_model-00006-of-00077.bin",
+    "model.layers.3.self_attn.k_proj.weight": "pytorch_model-00005-of-00077.bin",
+    "model.layers.3.self_attn.o_proj.weight": "pytorch_model-00006-of-00077.bin",
+    "model.layers.3.self_attn.q_proj.weight": "pytorch_model-00005-of-00077.bin",
+    "model.layers.3.self_attn.v_proj.weight": "pytorch_model-00006-of-00077.bin",
+    "model.layers.30.ln1.weight": "pytorch_model-00040-of-00077.bin",
+    "model.layers.30.ln2.weight": "pytorch_model-00040-of-00077.bin",
+    "model.layers.30.mlp.down_proj.weight": "pytorch_model-00040-of-00077.bin",
+    "model.layers.30.mlp.gate_proj.weight": "pytorch_model-00040-of-00077.bin",
+    "model.layers.30.mlp.up_proj.weight": "pytorch_model-00040-of-00077.bin",
+    "model.layers.30.self_attn.k_proj.weight": "pytorch_model-00039-of-00077.bin",
+    "model.layers.30.self_attn.o_proj.weight": "pytorch_model-00039-of-00077.bin",
+    "model.layers.30.self_attn.q_proj.weight": "pytorch_model-00039-of-00077.bin",
+    "model.layers.30.self_attn.v_proj.weight": "pytorch_model-00039-of-00077.bin",
+    "model.layers.31.ln1.weight": "pytorch_model-00041-of-00077.bin",
+    "model.layers.31.ln2.weight": "pytorch_model-00041-of-00077.bin",
+    "model.layers.31.mlp.down_proj.weight": "pytorch_model-00041-of-00077.bin",
+    "model.layers.31.mlp.gate_proj.weight": "pytorch_model-00041-of-00077.bin",
+    "model.layers.31.mlp.up_proj.weight": "pytorch_model-00041-of-00077.bin",
+    "model.layers.31.self_attn.k_proj.weight": "pytorch_model-00040-of-00077.bin",
+    "model.layers.31.self_attn.o_proj.weight": "pytorch_model-00041-of-00077.bin",
+    "model.layers.31.self_attn.q_proj.weight": "pytorch_model-00040-of-00077.bin",
+    "model.layers.31.self_attn.v_proj.weight": "pytorch_model-00041-of-00077.bin",
+    "model.layers.32.ln1.weight": "pytorch_model-00043-of-00077.bin",
+    "model.layers.32.ln2.weight": "pytorch_model-00043-of-00077.bin",
+    "model.layers.32.mlp.down_proj.weight": "pytorch_model-00042-of-00077.bin",
+    "model.layers.32.mlp.gate_proj.weight": "pytorch_model-00042-of-00077.bin",
+    "model.layers.32.mlp.up_proj.weight": "pytorch_model-00043-of-00077.bin",
+    "model.layers.32.self_attn.k_proj.weight": "pytorch_model-00042-of-00077.bin",
+    "model.layers.32.self_attn.o_proj.weight": "pytorch_model-00042-of-00077.bin",
+    "model.layers.32.self_attn.q_proj.weight": "pytorch_model-00042-of-00077.bin",
+    "model.layers.32.self_attn.v_proj.weight": "pytorch_model-00042-of-00077.bin",
+    "model.layers.33.ln1.weight": "pytorch_model-00044-of-00077.bin",
+    "model.layers.33.ln2.weight": "pytorch_model-00044-of-00077.bin",
+    "model.layers.33.mlp.down_proj.weight": "pytorch_model-00044-of-00077.bin",
+    "model.layers.33.mlp.gate_proj.weight": "pytorch_model-00043-of-00077.bin",
+    "model.layers.33.mlp.up_proj.weight": "pytorch_model-00044-of-00077.bin",
+    "model.layers.33.self_attn.k_proj.weight": "pytorch_model-00043-of-00077.bin",
+    "model.layers.33.self_attn.o_proj.weight": "pytorch_model-00043-of-00077.bin",
+    "model.layers.33.self_attn.q_proj.weight": "pytorch_model-00043-of-00077.bin",
+    "model.layers.33.self_attn.v_proj.weight": "pytorch_model-00043-of-00077.bin",
+    "model.layers.34.ln1.weight": "pytorch_model-00045-of-00077.bin",
+    "model.layers.34.ln2.weight": "pytorch_model-00045-of-00077.bin",
+    "model.layers.34.mlp.down_proj.weight": "pytorch_model-00045-of-00077.bin",
+    "model.layers.34.mlp.gate_proj.weight": "pytorch_model-00045-of-00077.bin",
+    "model.layers.34.mlp.up_proj.weight": "pytorch_model-00045-of-00077.bin",
+    "model.layers.34.self_attn.k_proj.weight": "pytorch_model-00044-of-00077.bin",
+    "model.layers.34.self_attn.o_proj.weight": "pytorch_model-00044-of-00077.bin",
+    "model.layers.34.self_attn.q_proj.weight": "pytorch_model-00044-of-00077.bin",
+    "model.layers.34.self_attn.v_proj.weight": "pytorch_model-00044-of-00077.bin",
+    "model.layers.35.ln1.weight": "pytorch_model-00046-of-00077.bin",
+    "model.layers.35.ln2.weight": "pytorch_model-00046-of-00077.bin",
+    "model.layers.35.mlp.down_proj.weight": "pytorch_model-00046-of-00077.bin",
+    "model.layers.35.mlp.gate_proj.weight": "pytorch_model-00046-of-00077.bin",
+    "model.layers.35.mlp.up_proj.weight": "pytorch_model-00046-of-00077.bin",
+    "model.layers.35.self_attn.k_proj.weight": "pytorch_model-00045-of-00077.bin",
+    "model.layers.35.self_attn.o_proj.weight": "pytorch_model-00046-of-00077.bin",
+    "model.layers.35.self_attn.q_proj.weight": "pytorch_model-00045-of-00077.bin",
+    "model.layers.35.self_attn.v_proj.weight": "pytorch_model-00046-of-00077.bin",
+    "model.layers.36.ln1.weight": "pytorch_model-00048-of-00077.bin",
+    "model.layers.36.ln2.weight": "pytorch_model-00048-of-00077.bin",
+    "model.layers.36.mlp.down_proj.weight": "pytorch_model-00047-of-00077.bin",
+    "model.layers.36.mlp.gate_proj.weight": "pytorch_model-00047-of-00077.bin",
+    "model.layers.36.mlp.up_proj.weight": "pytorch_model-00048-of-00077.bin",
+    "model.layers.36.self_attn.k_proj.weight": "pytorch_model-00047-of-00077.bin",
+    "model.layers.36.self_attn.o_proj.weight": "pytorch_model-00047-of-00077.bin",
+    "model.layers.36.self_attn.q_proj.weight": "pytorch_model-00047-of-00077.bin",
+    "model.layers.36.self_attn.v_proj.weight": "pytorch_model-00047-of-00077.bin",
+    "model.layers.37.ln1.weight": "pytorch_model-00049-of-00077.bin",
+    "model.layers.37.ln2.weight": "pytorch_model-00049-of-00077.bin",
+    "model.layers.37.mlp.down_proj.weight": "pytorch_model-00049-of-00077.bin",
+    "model.layers.37.mlp.gate_proj.weight": "pytorch_model-00048-of-00077.bin",
+    "model.layers.37.mlp.up_proj.weight": "pytorch_model-00049-of-00077.bin",
+    "model.layers.37.self_attn.k_proj.weight": "pytorch_model-00048-of-00077.bin",
+    "model.layers.37.self_attn.o_proj.weight": "pytorch_model-00048-of-00077.bin",
+    "model.layers.37.self_attn.q_proj.weight": "pytorch_model-00048-of-00077.bin",
+    "model.layers.37.self_attn.v_proj.weight": "pytorch_model-00048-of-00077.bin",
+    "model.layers.38.ln1.weight": "pytorch_model-00050-of-00077.bin",
+    "model.layers.38.ln2.weight": "pytorch_model-00050-of-00077.bin",
+    "model.layers.38.mlp.down_proj.weight": "pytorch_model-00050-of-00077.bin",
+    "model.layers.38.mlp.gate_proj.weight": "pytorch_model-00050-of-00077.bin",
+    "model.layers.38.mlp.up_proj.weight": "pytorch_model-00050-of-00077.bin",
+    "model.layers.38.self_attn.k_proj.weight": "pytorch_model-00049-of-00077.bin",
+    "model.layers.38.self_attn.o_proj.weight": "pytorch_model-00049-of-00077.bin",
+    "model.layers.38.self_attn.q_proj.weight": "pytorch_model-00049-of-00077.bin",
+    "model.layers.38.self_attn.v_proj.weight": "pytorch_model-00049-of-00077.bin",
+    "model.layers.39.ln1.weight": "pytorch_model-00051-of-00077.bin",
+    "model.layers.39.ln2.weight": "pytorch_model-00051-of-00077.bin",
+    "model.layers.39.mlp.down_proj.weight": "pytorch_model-00051-of-00077.bin",
+    "model.layers.39.mlp.gate_proj.weight": "pytorch_model-00051-of-00077.bin",
+    "model.layers.39.mlp.up_proj.weight": "pytorch_model-00051-of-00077.bin",
+    "model.layers.39.self_attn.k_proj.weight": "pytorch_model-00050-of-00077.bin",
+    "model.layers.39.self_attn.o_proj.weight": "pytorch_model-00051-of-00077.bin",
+    "model.layers.39.self_attn.q_proj.weight": "pytorch_model-00050-of-00077.bin",
+    "model.layers.39.self_attn.v_proj.weight": "pytorch_model-00051-of-00077.bin",
+    "model.layers.4.ln1.weight": "pytorch_model-00008-of-00077.bin",
+    "model.layers.4.ln2.weight": "pytorch_model-00008-of-00077.bin",
+    "model.layers.4.mlp.down_proj.weight": "pytorch_model-00007-of-00077.bin",
+    "model.layers.4.mlp.gate_proj.weight": "pytorch_model-00007-of-00077.bin",
+    "model.layers.4.mlp.up_proj.weight": "pytorch_model-00008-of-00077.bin",
+    "model.layers.4.self_attn.k_proj.weight": "pytorch_model-00007-of-00077.bin",
+    "model.layers.4.self_attn.o_proj.weight": "pytorch_model-00007-of-00077.bin",
+    "model.layers.4.self_attn.q_proj.weight": "pytorch_model-00007-of-00077.bin",
+    "model.layers.4.self_attn.v_proj.weight": "pytorch_model-00007-of-00077.bin",
+    "model.layers.40.ln1.weight": "pytorch_model-00053-of-00077.bin",
+    "model.layers.40.ln2.weight": "pytorch_model-00053-of-00077.bin",
+    "model.layers.40.mlp.down_proj.weight": "pytorch_model-00052-of-00077.bin",
+    "model.layers.40.mlp.gate_proj.weight": "pytorch_model-00052-of-00077.bin",
+    "model.layers.40.mlp.up_proj.weight": "pytorch_model-00053-of-00077.bin",
+    "model.layers.40.self_attn.k_proj.weight": "pytorch_model-00052-of-00077.bin",
+    "model.layers.40.self_attn.o_proj.weight": "pytorch_model-00052-of-00077.bin",
+    "model.layers.40.self_attn.q_proj.weight": "pytorch_model-00052-of-00077.bin",
+    "model.layers.40.self_attn.v_proj.weight": "pytorch_model-00052-of-00077.bin",
+    "model.layers.41.ln1.weight": "pytorch_model-00054-of-00077.bin",
+    "model.layers.41.ln2.weight": "pytorch_model-00054-of-00077.bin",
+    "model.layers.41.mlp.down_proj.weight": "pytorch_model-00054-of-00077.bin",
+    "model.layers.41.mlp.gate_proj.weight": "pytorch_model-00053-of-00077.bin",
+    "model.layers.41.mlp.up_proj.weight": "pytorch_model-00054-of-00077.bin",
+    "model.layers.41.self_attn.k_proj.weight": "pytorch_model-00053-of-00077.bin",
+    "model.layers.41.self_attn.o_proj.weight": "pytorch_model-00053-of-00077.bin",
+    "model.layers.41.self_attn.q_proj.weight": "pytorch_model-00053-of-00077.bin",
+    "model.layers.41.self_attn.v_proj.weight": "pytorch_model-00053-of-00077.bin",
+    "model.layers.42.ln1.weight": "pytorch_model-00055-of-00077.bin",
+    "model.layers.42.ln2.weight": "pytorch_model-00055-of-00077.bin",
+    "model.layers.42.mlp.down_proj.weight": "pytorch_model-00055-of-00077.bin",
+    "model.layers.42.mlp.gate_proj.weight": "pytorch_model-00055-of-00077.bin",
+    "model.layers.42.mlp.up_proj.weight": "pytorch_model-00055-of-00077.bin",
+    "model.layers.42.self_attn.k_proj.weight": "pytorch_model-00054-of-00077.bin",
+    "model.layers.42.self_attn.o_proj.weight": "pytorch_model-00054-of-00077.bin",
+    "model.layers.42.self_attn.q_proj.weight": "pytorch_model-00054-of-00077.bin",
+    "model.layers.42.self_attn.v_proj.weight": "pytorch_model-00054-of-00077.bin",
+    "model.layers.43.ln1.weight": "pytorch_model-00056-of-00077.bin",
+    "model.layers.43.ln2.weight": "pytorch_model-00056-of-00077.bin",
+    "model.layers.43.mlp.down_proj.weight": "pytorch_model-00056-of-00077.bin",
+    "model.layers.43.mlp.gate_proj.weight": "pytorch_model-00056-of-00077.bin",
+    "model.layers.43.mlp.up_proj.weight": "pytorch_model-00056-of-00077.bin",
+    "model.layers.43.self_attn.k_proj.weight": "pytorch_model-00055-of-00077.bin",
+    "model.layers.43.self_attn.o_proj.weight": "pytorch_model-00056-of-00077.bin",
+    "model.layers.43.self_attn.q_proj.weight": "pytorch_model-00055-of-00077.bin",
+    "model.layers.43.self_attn.v_proj.weight": "pytorch_model-00056-of-00077.bin",
+    "model.layers.44.ln1.weight": "pytorch_model-00058-of-00077.bin",
+    "model.layers.44.ln2.weight": "pytorch_model-00058-of-00077.bin",
+    "model.layers.44.mlp.down_proj.weight": "pytorch_model-00057-of-00077.bin",
+    "model.layers.44.mlp.gate_proj.weight": "pytorch_model-00057-of-00077.bin",
+    "model.layers.44.mlp.up_proj.weight": "pytorch_model-00058-of-00077.bin",
+    "model.layers.44.self_attn.k_proj.weight": "pytorch_model-00057-of-00077.bin",
+    "model.layers.44.self_attn.o_proj.weight": "pytorch_model-00057-of-00077.bin",
+    "model.layers.44.self_attn.q_proj.weight": "pytorch_model-00057-of-00077.bin",
+    "model.layers.44.self_attn.v_proj.weight": "pytorch_model-00057-of-00077.bin",
+    "model.layers.45.ln1.weight": "pytorch_model-00059-of-00077.bin",
+    "model.layers.45.ln2.weight": "pytorch_model-00059-of-00077.bin",
+    "model.layers.45.mlp.down_proj.weight": "pytorch_model-00059-of-00077.bin",
+    "model.layers.45.mlp.gate_proj.weight": "pytorch_model-00058-of-00077.bin",
+    "model.layers.45.mlp.up_proj.weight": "pytorch_model-00059-of-00077.bin",
+    "model.layers.45.self_attn.k_proj.weight": "pytorch_model-00058-of-00077.bin",
+    "model.layers.45.self_attn.o_proj.weight": "pytorch_model-00058-of-00077.bin",
+    "model.layers.45.self_attn.q_proj.weight": "pytorch_model-00058-of-00077.bin",
+    "model.layers.45.self_attn.v_proj.weight": "pytorch_model-00058-of-00077.bin",
+    "model.layers.46.ln1.weight": "pytorch_model-00060-of-00077.bin",
+    "model.layers.46.ln2.weight": "pytorch_model-00060-of-00077.bin",
+    "model.layers.46.mlp.down_proj.weight": "pytorch_model-00060-of-00077.bin",
+    "model.layers.46.mlp.gate_proj.weight": "pytorch_model-00060-of-00077.bin",
+    "model.layers.46.mlp.up_proj.weight": "pytorch_model-00060-of-00077.bin",
+    "model.layers.46.self_attn.k_proj.weight": "pytorch_model-00059-of-00077.bin",
+    "model.layers.46.self_attn.o_proj.weight": "pytorch_model-00059-of-00077.bin",
+    "model.layers.46.self_attn.q_proj.weight": "pytorch_model-00059-of-00077.bin",
+    "model.layers.46.self_attn.v_proj.weight": "pytorch_model-00059-of-00077.bin",
+    "model.layers.47.ln1.weight": "pytorch_model-00061-of-00077.bin",
+    "model.layers.47.ln2.weight": "pytorch_model-00061-of-00077.bin",
+    "model.layers.47.mlp.down_proj.weight": "pytorch_model-00061-of-00077.bin",
+    "model.layers.47.mlp.gate_proj.weight": "pytorch_model-00061-of-00077.bin",
+    "model.layers.47.mlp.up_proj.weight": "pytorch_model-00061-of-00077.bin",
+    "model.layers.47.self_attn.k_proj.weight": "pytorch_model-00060-of-00077.bin",
+    "model.layers.47.self_attn.o_proj.weight": "pytorch_model-00061-of-00077.bin",
+    "model.layers.47.self_attn.q_proj.weight": "pytorch_model-00060-of-00077.bin",
+    "model.layers.47.self_attn.v_proj.weight": "pytorch_model-00061-of-00077.bin",
+    "model.layers.48.ln1.weight": "pytorch_model-00063-of-00077.bin",
+    "model.layers.48.ln2.weight": "pytorch_model-00063-of-00077.bin",
+    "model.layers.48.mlp.down_proj.weight": "pytorch_model-00062-of-00077.bin",
+    "model.layers.48.mlp.gate_proj.weight": "pytorch_model-00062-of-00077.bin",
+    "model.layers.48.mlp.up_proj.weight": "pytorch_model-00063-of-00077.bin",
+    "model.layers.48.self_attn.k_proj.weight": "pytorch_model-00062-of-00077.bin",
+    "model.layers.48.self_attn.o_proj.weight": "pytorch_model-00062-of-00077.bin",
+    "model.layers.48.self_attn.q_proj.weight": "pytorch_model-00062-of-00077.bin",
+    "model.layers.48.self_attn.v_proj.weight": "pytorch_model-00062-of-00077.bin",
+    "model.layers.49.ln1.weight": "pytorch_model-00064-of-00077.bin",
+    "model.layers.49.ln2.weight": "pytorch_model-00064-of-00077.bin",
+    "model.layers.49.mlp.down_proj.weight": "pytorch_model-00064-of-00077.bin",
+    "model.layers.49.mlp.gate_proj.weight": "pytorch_model-00063-of-00077.bin",
+    "model.layers.49.mlp.up_proj.weight": "pytorch_model-00064-of-00077.bin",
+    "model.layers.49.self_attn.k_proj.weight": "pytorch_model-00063-of-00077.bin",
+    "model.layers.49.self_attn.o_proj.weight": "pytorch_model-00063-of-00077.bin",
+    "model.layers.49.self_attn.q_proj.weight": "pytorch_model-00063-of-00077.bin",
+    "model.layers.49.self_attn.v_proj.weight": "pytorch_model-00063-of-00077.bin",
+    "model.layers.5.ln1.weight": "pytorch_model-00009-of-00077.bin",
+    "model.layers.5.ln2.weight": "pytorch_model-00009-of-00077.bin",
+    "model.layers.5.mlp.down_proj.weight": "pytorch_model-00009-of-00077.bin",
+    "model.layers.5.mlp.gate_proj.weight": "pytorch_model-00008-of-00077.bin",
+    "model.layers.5.mlp.up_proj.weight": "pytorch_model-00009-of-00077.bin",
+    "model.layers.5.self_attn.k_proj.weight": "pytorch_model-00008-of-00077.bin",
+    "model.layers.5.self_attn.o_proj.weight": "pytorch_model-00008-of-00077.bin",
+    "model.layers.5.self_attn.q_proj.weight": "pytorch_model-00008-of-00077.bin",
+    "model.layers.5.self_attn.v_proj.weight": "pytorch_model-00008-of-00077.bin",
+    "model.layers.50.ln1.weight": "pytorch_model-00065-of-00077.bin",
+    "model.layers.50.ln2.weight": "pytorch_model-00065-of-00077.bin",
+    "model.layers.50.mlp.down_proj.weight": "pytorch_model-00065-of-00077.bin",
+    "model.layers.50.mlp.gate_proj.weight": "pytorch_model-00065-of-00077.bin",
+    "model.layers.50.mlp.up_proj.weight": "pytorch_model-00065-of-00077.bin",
+    "model.layers.50.self_attn.k_proj.weight": "pytorch_model-00064-of-00077.bin",
+    "model.layers.50.self_attn.o_proj.weight": "pytorch_model-00064-of-00077.bin",
+    "model.layers.50.self_attn.q_proj.weight": "pytorch_model-00064-of-00077.bin",
+    "model.layers.50.self_attn.v_proj.weight": "pytorch_model-00064-of-00077.bin",
+    "model.layers.51.ln1.weight": "pytorch_model-00066-of-00077.bin",
+    "model.layers.51.ln2.weight": "pytorch_model-00066-of-00077.bin",
+    "model.layers.51.mlp.down_proj.weight": "pytorch_model-00066-of-00077.bin",
+    "model.layers.51.mlp.gate_proj.weight": "pytorch_model-00066-of-00077.bin",
+    "model.layers.51.mlp.up_proj.weight": "pytorch_model-00066-of-00077.bin",
+    "model.layers.51.self_attn.k_proj.weight": "pytorch_model-00065-of-00077.bin",
+    "model.layers.51.self_attn.o_proj.weight": "pytorch_model-00066-of-00077.bin",
+    "model.layers.51.self_attn.q_proj.weight": "pytorch_model-00065-of-00077.bin",
+    "model.layers.51.self_attn.v_proj.weight": "pytorch_model-00066-of-00077.bin",
+    "model.layers.52.ln1.weight": "pytorch_model-00068-of-00077.bin",
+    "model.layers.52.ln2.weight": "pytorch_model-00068-of-00077.bin",
+    "model.layers.52.mlp.down_proj.weight": "pytorch_model-00067-of-00077.bin",
+    "model.layers.52.mlp.gate_proj.weight": "pytorch_model-00067-of-00077.bin",
+    "model.layers.52.mlp.up_proj.weight": "pytorch_model-00068-of-00077.bin",
+    "model.layers.52.self_attn.k_proj.weight": "pytorch_model-00067-of-00077.bin",
+    "model.layers.52.self_attn.o_proj.weight": "pytorch_model-00067-of-00077.bin",
+    "model.layers.52.self_attn.q_proj.weight": "pytorch_model-00067-of-00077.bin",
+    "model.layers.52.self_attn.v_proj.weight": "pytorch_model-00067-of-00077.bin",
+    "model.layers.53.ln1.weight": "pytorch_model-00069-of-00077.bin",
+    "model.layers.53.ln2.weight": "pytorch_model-00069-of-00077.bin",
+    "model.layers.53.mlp.down_proj.weight": "pytorch_model-00069-of-00077.bin",
+    "model.layers.53.mlp.gate_proj.weight": "pytorch_model-00068-of-00077.bin",
+    "model.layers.53.mlp.up_proj.weight": "pytorch_model-00069-of-00077.bin",
+    "model.layers.53.self_attn.k_proj.weight": "pytorch_model-00068-of-00077.bin",
+    "model.layers.53.self_attn.o_proj.weight": "pytorch_model-00068-of-00077.bin",
+    "model.layers.53.self_attn.q_proj.weight": "pytorch_model-00068-of-00077.bin",
+    "model.layers.53.self_attn.v_proj.weight": "pytorch_model-00068-of-00077.bin",
+    "model.layers.54.ln1.weight": "pytorch_model-00070-of-00077.bin",
+    "model.layers.54.ln2.weight": "pytorch_model-00070-of-00077.bin",
+    "model.layers.54.mlp.down_proj.weight": "pytorch_model-00070-of-00077.bin",
+    "model.layers.54.mlp.gate_proj.weight": "pytorch_model-00070-of-00077.bin",
+    "model.layers.54.mlp.up_proj.weight": "pytorch_model-00070-of-00077.bin",
+    "model.layers.54.self_attn.k_proj.weight": "pytorch_model-00069-of-00077.bin",
+    "model.layers.54.self_attn.o_proj.weight": "pytorch_model-00069-of-00077.bin",
+    "model.layers.54.self_attn.q_proj.weight": "pytorch_model-00069-of-00077.bin",
+    "model.layers.54.self_attn.v_proj.weight": "pytorch_model-00069-of-00077.bin",
+    "model.layers.55.ln1.weight": "pytorch_model-00071-of-00077.bin",
+    "model.layers.55.ln2.weight": "pytorch_model-00071-of-00077.bin",
+    "model.layers.55.mlp.down_proj.weight": "pytorch_model-00071-of-00077.bin",
+    "model.layers.55.mlp.gate_proj.weight": "pytorch_model-00071-of-00077.bin",
+    "model.layers.55.mlp.up_proj.weight": "pytorch_model-00071-of-00077.bin",
+    "model.layers.55.self_attn.k_proj.weight": "pytorch_model-00070-of-00077.bin",
+    "model.layers.55.self_attn.o_proj.weight": "pytorch_model-00071-of-00077.bin",
+    "model.layers.55.self_attn.q_proj.weight": "pytorch_model-00070-of-00077.bin",
+    "model.layers.55.self_attn.v_proj.weight": "pytorch_model-00071-of-00077.bin",
+    "model.layers.56.ln1.weight": "pytorch_model-00073-of-00077.bin",
+    "model.layers.56.ln2.weight": "pytorch_model-00073-of-00077.bin",
+    "model.layers.56.mlp.down_proj.weight": "pytorch_model-00072-of-00077.bin",
+    "model.layers.56.mlp.gate_proj.weight": "pytorch_model-00072-of-00077.bin",
+    "model.layers.56.mlp.up_proj.weight": "pytorch_model-00073-of-00077.bin",
+    "model.layers.56.self_attn.k_proj.weight": "pytorch_model-00072-of-00077.bin",
+    "model.layers.56.self_attn.o_proj.weight": "pytorch_model-00072-of-00077.bin",
+    "model.layers.56.self_attn.q_proj.weight": "pytorch_model-00072-of-00077.bin",
+    "model.layers.56.self_attn.v_proj.weight": "pytorch_model-00072-of-00077.bin",
+    "model.layers.57.ln1.weight": "pytorch_model-00074-of-00077.bin",
+    "model.layers.57.ln2.weight": "pytorch_model-00074-of-00077.bin",
+    "model.layers.57.mlp.down_proj.weight": "pytorch_model-00074-of-00077.bin",
+    "model.layers.57.mlp.gate_proj.weight": "pytorch_model-00073-of-00077.bin",
+    "model.layers.57.mlp.up_proj.weight": "pytorch_model-00074-of-00077.bin",
+    "model.layers.57.self_attn.k_proj.weight": "pytorch_model-00073-of-00077.bin",
+    "model.layers.57.self_attn.o_proj.weight": "pytorch_model-00073-of-00077.bin",
+    "model.layers.57.self_attn.q_proj.weight": "pytorch_model-00073-of-00077.bin",
+    "model.layers.57.self_attn.v_proj.weight": "pytorch_model-00073-of-00077.bin",
+    "model.layers.58.ln1.weight": "pytorch_model-00075-of-00077.bin",
+    "model.layers.58.ln2.weight": "pytorch_model-00075-of-00077.bin",
+    "model.layers.58.mlp.down_proj.weight": "pytorch_model-00075-of-00077.bin",
+    "model.layers.58.mlp.gate_proj.weight": "pytorch_model-00075-of-00077.bin",
+    "model.layers.58.mlp.up_proj.weight": "pytorch_model-00075-of-00077.bin",
+    "model.layers.58.self_attn.k_proj.weight": "pytorch_model-00074-of-00077.bin",
+    "model.layers.58.self_attn.o_proj.weight": "pytorch_model-00074-of-00077.bin",
+    "model.layers.58.self_attn.q_proj.weight": "pytorch_model-00074-of-00077.bin",
+    "model.layers.58.self_attn.v_proj.weight": "pytorch_model-00074-of-00077.bin",
+    "model.layers.59.ln1.weight": "pytorch_model-00076-of-00077.bin",
+    "model.layers.59.ln2.weight": "pytorch_model-00076-of-00077.bin",
+    "model.layers.59.mlp.down_proj.weight": "pytorch_model-00076-of-00077.bin",
+    "model.layers.59.mlp.gate_proj.weight": "pytorch_model-00076-of-00077.bin",
+    "model.layers.59.mlp.up_proj.weight": "pytorch_model-00076-of-00077.bin",
+    "model.layers.59.self_attn.k_proj.weight": "pytorch_model-00075-of-00077.bin",
+    "model.layers.59.self_attn.o_proj.weight": "pytorch_model-00076-of-00077.bin",
+    "model.layers.59.self_attn.q_proj.weight": "pytorch_model-00075-of-00077.bin",
+    "model.layers.59.self_attn.v_proj.weight": "pytorch_model-00076-of-00077.bin",
+    "model.layers.6.ln1.weight": "pytorch_model-00010-of-00077.bin",
+    "model.layers.6.ln2.weight": "pytorch_model-00010-of-00077.bin",
+    "model.layers.6.mlp.down_proj.weight": "pytorch_model-00010-of-00077.bin",
+    "model.layers.6.mlp.gate_proj.weight": "pytorch_model-00010-of-00077.bin",
+    "model.layers.6.mlp.up_proj.weight": "pytorch_model-00010-of-00077.bin",
+    "model.layers.6.self_attn.k_proj.weight": "pytorch_model-00009-of-00077.bin",
+    "model.layers.6.self_attn.o_proj.weight": "pytorch_model-00009-of-00077.bin",
+    "model.layers.6.self_attn.q_proj.weight": "pytorch_model-00009-of-00077.bin",
+    "model.layers.6.self_attn.v_proj.weight": "pytorch_model-00009-of-00077.bin",
+    "model.layers.7.ln1.weight": "pytorch_model-00011-of-00077.bin",
+    "model.layers.7.ln2.weight": "pytorch_model-00011-of-00077.bin",
+    "model.layers.7.mlp.down_proj.weight": "pytorch_model-00011-of-00077.bin",
+    "model.layers.7.mlp.gate_proj.weight": "pytorch_model-00011-of-00077.bin",
+    "model.layers.7.mlp.up_proj.weight": "pytorch_model-00011-of-00077.bin",
+    "model.layers.7.self_attn.k_proj.weight": "pytorch_model-00010-of-00077.bin",
+    "model.layers.7.self_attn.o_proj.weight": "pytorch_model-00011-of-00077.bin",
+    "model.layers.7.self_attn.q_proj.weight": "pytorch_model-00010-of-00077.bin",
+    "model.layers.7.self_attn.v_proj.weight": "pytorch_model-00011-of-00077.bin",
+    "model.layers.8.ln1.weight": "pytorch_model-00013-of-00077.bin",
+    "model.layers.8.ln2.weight": "pytorch_model-00013-of-00077.bin",
+    "model.layers.8.mlp.down_proj.weight": "pytorch_model-00012-of-00077.bin",
+    "model.layers.8.mlp.gate_proj.weight": "pytorch_model-00012-of-00077.bin",
+    "model.layers.8.mlp.up_proj.weight": "pytorch_model-00013-of-00077.bin",
+    "model.layers.8.self_attn.k_proj.weight": "pytorch_model-00012-of-00077.bin",
+    "model.layers.8.self_attn.o_proj.weight": "pytorch_model-00012-of-00077.bin",
+    "model.layers.8.self_attn.q_proj.weight": "pytorch_model-00012-of-00077.bin",
+    "model.layers.8.self_attn.v_proj.weight": "pytorch_model-00012-of-00077.bin",
+    "model.layers.9.ln1.weight": "pytorch_model-00014-of-00077.bin",
+    "model.layers.9.ln2.weight": "pytorch_model-00014-of-00077.bin",
+    "model.layers.9.mlp.down_proj.weight": "pytorch_model-00014-of-00077.bin",
+    "model.layers.9.mlp.gate_proj.weight": "pytorch_model-00013-of-00077.bin",
+    "model.layers.9.mlp.up_proj.weight": "pytorch_model-00014-of-00077.bin",
+    "model.layers.9.self_attn.k_proj.weight": "pytorch_model-00013-of-00077.bin",
+    "model.layers.9.self_attn.o_proj.weight": "pytorch_model-00013-of-00077.bin",
+    "model.layers.9.self_attn.q_proj.weight": "pytorch_model-00013-of-00077.bin",
+    "model.layers.9.self_attn.v_proj.weight": "pytorch_model-00013-of-00077.bin",
+    "model.norm.weight": "pytorch_model-00076-of-00077.bin"
+  }
+}

special_tokens_map.json

@@ -0,0 +1,30 @@
+{
+  "bos_token": {
+    "content": "<|startoftext|>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenization_yi.py

@@ -0,0 +1,255 @@
+import os
+from shutil import copyfile
+from typing import Any, Dict, List, Optional, Tuple
+
+import sentencepiece as spm
+from transformers.tokenization_utils import AddedToken, PreTrainedTokenizer
+from transformers.utils import logging
+
+logger = logging.get_logger(__name__)
+
+VOCAB_FILES_NAMES = {"vocab_file": "tokenizer.model"}
+
+PRETRAINED_VOCAB_FILES_MAP = {
+    "vocab_file": {},
+    "tokenizer_file": {},
+}
+PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES = {}
+
+
+class YiTokenizer(PreTrainedTokenizer):
+    """
+    Construct a Yi tokenizer. Based on byte-level Byte-Pair-Encoding.
+
+    Args:
+        vocab_file (`str`):
+            Path to the vocabulary file.
+    """
+
+    vocab_files_names = VOCAB_FILES_NAMES
+    pretrained_vocab_files_map = PRETRAINED_VOCAB_FILES_MAP
+    max_model_input_sizes = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
+    model_input_names = ["input_ids", "attention_mask"]
+
+    def __init__(
+        self,
+        vocab_file,
+        unk_token="<unk>",
+        bos_token="<|startoftext|>",
+        eos_token="<|endoftext|>",
+        pad_token="<unk>",
+        sp_model_kwargs: Optional[Dict[str, Any]] = None,
+        add_bos_token=True,
+        add_eos_token=False,
+        clean_up_tokenization_spaces=False,
+        **kwargs,
+    ):
+        self.sp_model_kwargs = {} if sp_model_kwargs is None else sp_model_kwargs
+        bos_token = (
+            AddedToken(bos_token, lstrip=False, rstrip=False)
+            if isinstance(bos_token, str)
+            else bos_token
+        )
+        eos_token = (
+            AddedToken(eos_token, lstrip=False, rstrip=False)
+            if isinstance(eos_token, str)
+            else eos_token
+        )
+        unk_token = (
+            AddedToken(unk_token, lstrip=False, rstrip=False)
+            if isinstance(unk_token, str)
+            else unk_token
+        )
+        pad_token = (
+            AddedToken(pad_token, lstrip=False, rstrip=False)
+            if isinstance(pad_token, str)
+            else pad_token
+        )
+        self.vocab_file = vocab_file
+        self.add_bos_token = add_bos_token
+        self.add_eos_token = add_eos_token
+        self.sp_model = spm.SentencePieceProcessor(**self.sp_model_kwargs)
+        self.sp_model.Load(vocab_file)
+        super().__init__(
+            bos_token=bos_token,
+            eos_token=eos_token,
+            unk_token=unk_token,
+            pad_token=pad_token,
+            add_bos_token=add_bos_token,
+            add_eos_token=add_eos_token,
+            sp_model_kwargs=self.sp_model_kwargs,
+            clean_up_tokenization_spaces=clean_up_tokenization_spaces,
+            **kwargs,
+        )
+
+    def __getstate__(self):
+        state = self.__dict__.copy()
+        state["sp_model"] = None
+        return state
+
+    def __setstate__(self, d):
+        self.__dict__ = d
+        self.sp_model = spm.SentencePieceProcessor(**self.sp_model_kwargs)
+        self.sp_model.Load(self.vocab_file)
+
+    @property
+    def vocab_size(self):
+        """Returns vocab size"""
+        return self.sp_model.get_piece_size()
+
+    def get_vocab(self):
+        """Returns vocab as a dict"""
+        vocab = {self.convert_ids_to_tokens(i): i for i in range(self.vocab_size)}
+        vocab.update(self.added_tokens_encoder)
+        return vocab
+
+    def _tokenize(self, text):
+        """Returns a tokenized string."""
+        return self.sp_model.encode(text, out_type=str)
+
+    def _convert_token_to_id(self, token):
+        """Converts a token (str) in an id using the vocab."""
+        return self.sp_model.piece_to_id(token)
+
+    def _convert_id_to_token(self, index):
+        """Converts an index (integer) in a token (str) using the vocab."""
+        token = self.sp_model.IdToPiece(index)
+        return token
+
+    def convert_tokens_to_string(self, tokens):
+        """Converts a sequence of tokens (string) in a single string."""
+        current_sub_tokens = []
+        out_string = ""
+        prev_is_special = False
+        for i, token in enumerate(tokens):
+            # make sure that special tokens are not decoded using sentencepiece model
+            if token in self.all_special_tokens:
+                if not prev_is_special and i != 0:
+                    out_string += " "
+                out_string += self.sp_model.decode(current_sub_tokens) + token
+                prev_is_special = True
+                current_sub_tokens = []
+            else:
+                current_sub_tokens.append(token)
+                prev_is_special = False
+        out_string += self.sp_model.decode(current_sub_tokens)
+        return out_string
+
+    def save_vocabulary(
+        self, save_directory, filename_prefix: Optional[str] = None
+    ) -> Tuple[str]:
+        """
+        Save the vocabulary and special tokens file to a directory.
+
+        Args:
+            save_directory (`str`):
+                The directory in which to save the vocabulary.
+
+        Returns:
+            `Tuple(str)`: Paths to the files saved.
+        """
+        if not os.path.isdir(save_directory):
+            logger.error(f"Vocabulary path ({save_directory}) should be a directory")
+            return
+        out_vocab_file = os.path.join(
+            save_directory,
+            (filename_prefix + "-" if filename_prefix else "")
+            + VOCAB_FILES_NAMES["vocab_file"],
+        )
+
+        if os.path.abspath(self.vocab_file) != os.path.abspath(
+            out_vocab_file
+        ) and os.path.isfile(self.vocab_file):
+            copyfile(self.vocab_file, out_vocab_file)
+        elif not os.path.isfile(self.vocab_file):
+            with open(out_vocab_file, "wb") as fi:
+                content_spiece_model = self.sp_model.serialized_model_proto()
+                fi.write(content_spiece_model)
+
+        return (out_vocab_file,)
+
+    def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=None):
+        bos_token_id = [self.bos_token_id] if self.add_bos_token else []
+        eos_token_id = [self.eos_token_id] if self.add_eos_token else []
+
+        output = bos_token_id + token_ids_0 + eos_token_id
+
+        if token_ids_1 is not None:
+            output = output + bos_token_id + token_ids_1 + eos_token_id
+
+        return output
+
+    def get_special_tokens_mask(
+        self,
+        token_ids_0: List[int],
+        token_ids_1: Optional[List[int]] = None,
+        already_has_special_tokens: bool = False,
+    ) -> List[int]:
+        """
+        Retrieve sequence ids from a token list that has no special tokens added. This method is called when adding
+        special tokens using the tokenizer `prepare_for_model` method.
+
+        Args:
+            token_ids_0 (`List[int]`):
+                List of IDs.
+            token_ids_1 (`List[int]`, *optional*):
+                Optional second list of IDs for sequence pairs.
+            already_has_special_tokens (`bool`, *optional*, defaults to `False`):
+                Whether or not the token list is already formatted with special tokens for the model.
+
+        Returns:
+            `List[int]`: A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token.
+        """
+        if already_has_special_tokens:
+            return super().get_special_tokens_mask(
+                token_ids_0=token_ids_0,
+                token_ids_1=token_ids_1,
+                already_has_special_tokens=True,
+            )
+
+        bos_token_id = [1] if self.add_bos_token else []
+        eos_token_id = [1] if self.add_eos_token else []
+
+        if token_ids_1 is None:
+            return bos_token_id + ([0] * len(token_ids_0)) + eos_token_id
+        return (
+            bos_token_id
+            + ([0] * len(token_ids_0))
+            + eos_token_id
+            + bos_token_id
+            + ([0] * len(token_ids_1))
+            + eos_token_id
+        )
+
+    def create_token_type_ids_from_sequences(
+        self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None
+    ) -> List[int]:
+        """
+        Creates a mask from the two sequences passed to be used in a sequence-pair classification task. An ALBERT
+        sequence pair mask has the following format:
+
+        ```
+        0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1
+        | first sequence    | second sequence |
+        ```
+
+        if token_ids_1 is None, only returns the first portion of the mask (0s).
+
+        Args:
+            token_ids_0 (`List[int]`):
+                List of ids.
+            token_ids_1 (`List[int]`, *optional*):
+                Optional second list of IDs for sequence pairs.
+
+        Returns:
+            `List[int]`: List of [token type IDs](../glossary#token-type-ids) according to the given sequence(s).
+        """
+        bos_token_id = [self.bos_token_id] if self.add_bos_token else []
+        eos_token_id = [self.eos_token_id] if self.add_eos_token else []
+
+        output = [0] * len(bos_token_id + token_ids_0 + eos_token_id)
+
+        if token_ids_1 is not None:
+            output += [1] * len(bos_token_id + token_ids_1 + eos_token_id)
+
+        return output

tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:386c49cf943d71aa110361135338c50e38beeff0a66593480421f37b319e1a39
+size 1033105

tokenizer_config.json

@@ -0,0 +1,46 @@
+{
+  "add_bos_token": false,
+  "add_eos_token": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<|startoftext|>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "auto_map": {
+    "AutoTokenizer": [
+      "tokenization_yi.YiTokenizer",
+      null
+    ]
+  },
+  "bos_token": "<|startoftext|>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "model_max_length": 4096,
+  "pad_token": "<unk>",
+  "padding_side": "left",
+  "sp_model_kwargs": {},
+  "split_special_tokens": false,
+  "tokenizer_class": "YiTokenizer",
+  "unk_token": "<unk>"
+}