init

hybrid_model_lv_mixed_moe/config.json

@@ -0,0 +1,41 @@
+{
+  "_name_or_path": "modelglue/which-4/ffn_mixed_moe",
+  "architectures": [
+    "MixtralForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "auto_map": {
+    "AutoConfig": "configuration_mixtral_model_level_hybrid.MixtralModelLevelHybridConfig",
+    "AutoModelForCausalLM": "modeling_mixtral_model_level_hybrid.MixtralModelLevelHybridForCausalLM"
+  },
+  "badwordsids": "[[29961], [14352], [24630], [29962], [11759], [15974], [5519], [25473], [18899], [25901], [7110], [9341], [13531], [518], [9310], [2636], [3366], [21069], [11970], [23098], [16733], [21298], [18173], [10846], [3816], [28513], [15625], [23192], [28166], [10062], [1385], [11724], [3108], [15555], [10834], [10370], [14330], [1822], [12436], [5262], [17094], [10725], [17077], [11424], [4197], [24406], [13359], [17531], [24566], [23076], [4514], [13192], [19942], [16261], [7072], [6024], [1402], [1839], [2033], [13970], [850], [5913], [28895], [5387], [8308], [24927], [5691], [12940], [19997], [18959], [11287], [16862], [4638], [22322], [29861], [21251], [14704], [17548], [12452], [17288], [23160], [24960], [8219], [18024], [5539], [7464], [27865], [29588], [20068], [19660], [27706], [22896], [24264], [12258], [2314], [4400], [5586], [12622], [6796], [7226], [21939], [18456], [14178], [21540], [21945], [14664], [16215], [10338], [17361], [7503], [13769], [26073], [9601], [26909], [7961], [8999], [20840], [16272], [21545], [3199], [10514], [5159], [22689], [6525], [20526], [27077], [18017]]",
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "intermediate_size": 11008,
+  "max_position_embeddings": 4096,
+  "model_type": "mixtral_block_level",
+  "num_attention_heads": 32,
+  "num_dense_layers": 16,
+  "num_experts_per_tok": 2,
+  "num_hidden_layers": 32,
+  "num_key_value_heads": 32,
+  "num_local_experts": 4,
+  "output_router_logits": false,
+  "pretraining_tp": 1,
+  "rms_norm_eps": 1e-05,
+  "rope_scaling": null,
+  "rope_theta": 10000.0,
+  "router_aux_loss_coef": 0.001,
+  "router_jitter_noise": 0.0,
+  "sequence_pooler": "mean",
+  "sliding_window": null,
+  "tie_word_embeddings": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.40.1",
+  "use_cache": false,
+  "vocab_size": 32000
+}

hybrid_model_lv_mixed_moe/configuration_mixtral_model_level_hybrid.py

@@ -0,0 +1,171 @@
+# coding=utf-8
+# Copyright 2023 Mixtral AI and the HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+""" Mixtral model configuration"""
+
+from transformers.configuration_utils import PretrainedConfig
+from transformers.utils import logging
+
+logger = logging.get_logger(__name__)
+
+
+class MixtralModelLevelHybridConfig(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`MixtralModel`]. It is used to instantiate an
+    Mixtral 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 Mixtral-7B-v0.1 or Mixtral-7B-Instruct-v0.1.
+
+    [mixtralai/Mixtral-8x7B](https://huggingface.co/mixtralai/Mixtral-8x7B)
+    [mixtralai/Mixtral-7B-Instruct-v0.1](https://huggingface.co/mixtralai/Mixtral-7B-Instruct-v0.1)
+
+    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 32000):
+            Vocabulary size of the Mixtral model. Defines the number of different tokens that can be represented by the
+            `inputs_ids` passed when calling [`MixtralModel`]
+        hidden_size (`int`, *optional*, defaults to 4096):
+            Dimension of the hidden representations.
+        intermediate_size (`int`, *optional*, defaults to 14336):
+            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*, defaults to 8):
+            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 `8`.
+        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*32`):
+            The maximum sequence length that this model might ever be used with. Mixtral's sliding window attention
+            allows sequence of up to 4096*32 tokens.
+        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-05):
+            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`.
+        pad_token_id (`int`, *optional*):
+            The id of the padding token.
+        bos_token_id (`int`, *optional*, defaults to 1):
+            The id of the "beginning-of-sequence" token.
+        eos_token_id (`int`, *optional*, defaults to 2):
+            The id of the "end-of-sequence" token.
+        tie_word_embeddings (`bool`, *optional*, defaults to `False`):
+            Whether the model's input and output word embeddings should be tied.
+        rope_theta (`float`, *optional*, defaults to 1000000.0):
+            The base period of the RoPE embeddings.
+        sliding_window (`int`, *optional*):
+            Sliding window attention window size. If not specified, will default to `4096`.
+        attention_dropout (`float`, *optional*, defaults to 0.0):
+            The dropout ratio for the attention probabilities.
+        num_experts_per_tok (`int`, *optional*, defaults to 2):
+            The number of experts to root per-token, can be also interpreted as the `top-p` routing
+            parameter
+        num_local_experts (`int`, *optional*, defaults to 8):
+            Number of experts per Sparse MLP layer.
+        output_router_logits (`bool`, *optional*, defaults to `False`):
+            Whether or not the router logits should be returned by the model. Enabeling this will also
+            allow the model to output the auxiliary loss. See [here]() for more details
+        router_aux_loss_coef (`float`, *optional*, defaults to 0.001):
+            The aux loss factor for the total loss.
+
+    ```python
+    >>> from transformers import MixtralModel, MixtralConfig
+
+    >>> # Initializing a Mixtral 7B style configuration
+    >>> configuration = MixtralConfig()
+
+    >>> # Initializing a model from the Mixtral 7B style configuration
+    >>> model = MixtralModel(configuration)
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+
+    model_type = "mixtral_block_level"
+    keys_to_ignore_at_inference = ["past_key_values"]
+
+    def __init__(
+            self,
+            vocab_size=32000,
+            hidden_size=4096,
+            intermediate_size=14336,
+            num_hidden_layers=32,
+            num_attention_heads=32,
+            num_key_value_heads=8,
+            hidden_act="silu",
+            max_position_embeddings=4096 * 32,
+            initializer_range=0.02,
+            rms_norm_eps=1e-5,
+            use_cache=True,
+            pad_token_id=None,
+            bos_token_id=1,
+            eos_token_id=2,
+            tie_word_embeddings=False,
+            rope_theta=1e6,
+            sliding_window=None,
+            attention_dropout=0.0,
+            num_experts_per_tok=2,
+            num_local_experts=8,
+            output_router_logits=False,
+            router_aux_loss_coef=0.001,
+            num_dense_layers=0,
+            sequence_pooler="mean",
+            **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
+        self.sliding_window = sliding_window
+
+        # 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.rope_theta = rope_theta
+        self.attention_dropout = attention_dropout
+
+        self.num_experts_per_tok = num_experts_per_tok
+        self.num_local_experts = num_local_experts
+        self.output_router_logits = output_router_logits
+        self.router_aux_loss_coef = router_aux_loss_coef
+        self.num_dense_layers = num_dense_layers
+        self.sequence_pooler = sequence_pooler
+        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,
+        )
+
+
+MixtralModelLevelHybridConfig.register_for_auto_class("AutoConfig")

hybrid_model_lv_mixed_moe/model-00001-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c1d875e0455beb95f81854296b81453b60b7ce91d3a06f79b6de802d0801681
+size 9993615456

hybrid_model_lv_mixed_moe/model-00002-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1f4c9f969554aae738862aecd4276adc1a1cc16108f90f0719c38bc413143fd4
+size 9919542664

hybrid_model_lv_mixed_moe/model-00003-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f0fbaa877dff3c1348cd8160389996df782efe5f7f8a6edf8f602055191bb48a
+size 9972655456

hybrid_model_lv_mixed_moe/model-00004-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61784c5bbefa0e5966b822a73e108f8def6ab1ecdfe6d72ab5c582e84ff3f7e7
+size 3019943056

hybrid_model_lv_mixed_moe/model.safetensors.index.json

@@ -0,0 +1 @@
+{"metadata": {"mergekit_version": "0.0.4"}, "weight_map": {"model.embed_tokens.weight": "model-00001-of-00004.safetensors", "model.norm.weight": "model-00001-of-00004.safetensors", "lm_head.weight": "model-00001-of-00004.safetensors", "model.layers.0.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.1.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.2.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.3.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.4.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.5.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.6.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.7.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.8.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.9.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.10.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.11.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.12.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.13.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.14.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.15.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.16.experts.0.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.16.experts.1.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.16.experts.2.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.16.experts.3.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.17.experts.0.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.17.experts.1.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.17.experts.2.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.17.experts.3.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.18.experts.0.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.18.experts.1.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.18.experts.2.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.18.experts.3.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.19.experts.0.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.19.experts.1.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.19.experts.2.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.19.experts.3.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.20.experts.0.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.20.experts.1.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.20.experts.2.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.20.experts.3.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.21.experts.0.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.21.experts.1.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.21.experts.2.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.21.experts.3.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.22.experts.0.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.22.experts.1.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.22.experts.2.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.22.experts.3.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.23.experts.0.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.23.experts.1.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.23.experts.2.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.23.experts.3.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.24.experts.0.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.24.experts.1.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.24.experts.2.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.24.experts.3.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.25.experts.0.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.25.experts.1.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.25.experts.2.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.25.experts.3.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.26.experts.0.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.26.experts.1.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.26.experts.2.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.26.experts.3.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.27.experts.0.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.27.experts.1.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.27.experts.2.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.27.experts.3.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.28.experts.0.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.28.experts.1.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.28.experts.2.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.28.experts.3.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.29.experts.0.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.29.experts.1.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.29.experts.2.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.29.experts.3.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.30.experts.0.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.30.experts.1.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.30.experts.2.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.30.experts.3.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.31.experts.0.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.31.experts.1.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.31.experts.2.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.31.experts.3.input_layernorm.weight": "model-00001-of-00004.safetensors", "model.layers.0.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.1.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.2.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.3.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.4.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.5.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.6.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.7.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.8.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.9.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.10.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.11.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.12.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.13.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.14.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.15.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.16.experts.0.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.16.experts.1.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.16.experts.2.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.16.experts.3.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.17.experts.0.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.17.experts.1.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.17.experts.2.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.17.experts.3.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.18.experts.0.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.18.experts.1.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.18.experts.2.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.18.experts.3.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.19.experts.0.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.19.experts.1.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.19.experts.2.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.19.experts.3.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.20.experts.0.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.20.experts.1.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.20.experts.2.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.20.experts.3.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.21.experts.0.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.21.experts.1.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.21.experts.2.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.21.experts.3.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.22.experts.0.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.22.experts.1.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.22.experts.2.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.22.experts.3.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.23.experts.0.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.23.experts.1.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.23.experts.2.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.23.experts.3.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.24.experts.0.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.24.experts.1.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.24.experts.2.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.24.experts.3.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.25.experts.0.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.25.experts.1.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.25.experts.2.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.25.experts.3.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.26.experts.0.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.26.experts.1.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.26.experts.2.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.26.experts.3.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.27.experts.0.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.27.experts.1.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.27.experts.2.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.27.experts.3.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.28.experts.0.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.28.experts.1.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.28.experts.2.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.28.experts.3.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.29.experts.0.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.29.experts.1.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.29.experts.2.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.29.experts.3.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.30.experts.0.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.30.experts.1.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.30.experts.2.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.30.experts.3.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.31.experts.0.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.31.experts.1.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.31.experts.2.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.31.experts.3.mlp.up_proj.weight": "model-00001-of-00004.safetensors", "model.layers.0.mlp.down_proj.weight": "model-00001-of-00004.safetensors", "model.layers.1.mlp.down_proj.weight": "model-00001-of-00004.safetensors", "model.layers.2.mlp.down_proj.weight": "model-00001-of-00004.safetensors", "model.layers.3.mlp.down_proj.weight": "model-00001-of-00004.safetensors", "model.layers.4.mlp.down_proj.weight": "model-00001-of-00004.safetensors", "model.layers.5.mlp.down_proj.weight": "model-00001-of-00004.safetensors", "model.layers.6.mlp.down_proj.weight": "model-00001-of-00004.safetensors", "model.layers.7.mlp.down_proj.weight": "model-00001-of-00004.safetensors", "model.layers.8.mlp.down_proj.weight": "model-00001-of-00004.safetensors", "model.layers.9.mlp.down_proj.weight": "model-00001-of-00004.safetensors", "model.layers.10.mlp.down_proj.weight": "model-00001-of-00004.safetensors", "model.layers.11.mlp.down_proj.weight": "model-00001-of-00004.safetensors", "model.layers.12.mlp.down_proj.weight": "model-00001-of-00004.safetensors", "model.layers.13.mlp.down_proj.weight": "model-00001-of-00004.safetensors", "model.layers.14.mlp.down_proj.weight": "model-00001-of-00004.safetensors", "model.layers.15.mlp.down_proj.weight": "model-00001-of-00004.safetensors", "model.layers.16.experts.0.mlp.down_proj.weight": "model-00001-of-00004.safetensors", "model.layers.16.experts.1.mlp.down_proj.weight": "model-00001-of-00004.safetensors", "model.layers.16.experts.2.mlp.down_proj.weight": "model-00001-of-00004.safetensors", "model.layers.16.experts.3.mlp.down_proj.weight": "model-00001-of-00004.safetensors", "model.layers.17.experts.0.mlp.down_proj.weight": "model-00001-of-00004.safetensors", "model.layers.17.experts.1.mlp.down_proj.weight": "model-00001-of-00004.safetensors", "model.layers.17.experts.2.mlp.down_proj.weight": "model-00001-of-00004.safetensors", "model.layers.17.experts.3.mlp.down_proj.weight": "model-00001-of-00004.safetensors", "model.layers.18.experts.0.mlp.down_proj.weight": "model-00001-of-00004.safetensors", "model.layers.18.experts.1.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.18.experts.2.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.18.experts.3.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.19.experts.0.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.19.experts.1.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.19.experts.2.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.19.experts.3.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.20.experts.0.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.20.experts.1.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.20.experts.2.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.20.experts.3.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.21.experts.0.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.21.experts.1.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.21.experts.2.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.21.experts.3.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.22.experts.0.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.22.experts.1.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.22.experts.2.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.22.experts.3.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.23.experts.0.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.23.experts.1.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.23.experts.2.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.23.experts.3.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.24.experts.0.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.24.experts.1.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.24.experts.2.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.24.experts.3.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.25.experts.0.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.25.experts.1.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.25.experts.2.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.25.experts.3.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.26.experts.0.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.26.experts.1.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.26.experts.2.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.26.experts.3.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.27.experts.0.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.27.experts.1.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.27.experts.2.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.27.experts.3.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.28.experts.0.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.28.experts.1.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.28.experts.2.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.28.experts.3.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.29.experts.0.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.29.experts.1.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.29.experts.2.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.29.experts.3.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.30.experts.0.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.30.experts.1.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.30.experts.2.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.30.experts.3.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.31.experts.0.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.31.experts.1.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.31.experts.2.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.31.experts.3.mlp.down_proj.weight": "model-00002-of-00004.safetensors", "model.layers.0.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.1.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.2.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.3.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.4.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.5.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.6.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.7.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.8.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.9.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.10.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.11.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.12.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.13.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.14.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.15.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.16.experts.0.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.16.experts.1.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.16.experts.2.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.16.experts.3.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.17.experts.0.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.17.experts.1.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.17.experts.2.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.17.experts.3.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.18.experts.0.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.18.experts.1.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.18.experts.2.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.18.experts.3.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.19.experts.0.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.19.experts.1.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.19.experts.2.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.19.experts.3.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.20.experts.0.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.20.experts.1.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.20.experts.2.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.20.experts.3.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.21.experts.0.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.21.experts.1.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.21.experts.2.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.21.experts.3.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.22.experts.0.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.22.experts.1.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.22.experts.2.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.22.experts.3.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.23.experts.0.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.23.experts.1.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.23.experts.2.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.23.experts.3.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.24.experts.0.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.24.experts.1.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.24.experts.2.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.24.experts.3.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.25.experts.0.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.25.experts.1.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.25.experts.2.mlp.gate_proj.weight": "model-00002-of-00004.safetensors", "model.layers.25.experts.3.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", "model.layers.26.experts.0.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", "model.layers.26.experts.1.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", "model.layers.26.experts.2.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", "model.layers.26.experts.3.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", "model.layers.27.experts.0.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", "model.layers.27.experts.1.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", "model.layers.27.experts.2.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", "model.layers.27.experts.3.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", "model.layers.28.experts.0.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", "model.layers.28.experts.1.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", "model.layers.28.experts.2.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", "model.layers.28.experts.3.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", "model.layers.29.experts.0.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", "model.layers.29.experts.1.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", "model.layers.29.experts.2.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", "model.layers.29.experts.3.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", "model.layers.30.experts.0.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", "model.layers.30.experts.1.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", "model.layers.30.experts.2.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", "model.layers.30.experts.3.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", "model.layers.31.experts.0.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", "model.layers.31.experts.1.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", "model.layers.31.experts.2.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", "model.layers.31.experts.3.mlp.gate_proj.weight": "model-00003-of-00004.safetensors", "model.layers.0.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.1.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.2.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.3.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.4.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.5.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.6.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.7.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.8.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.9.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.10.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.11.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.12.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.13.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.14.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.15.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.16.experts.0.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.16.experts.1.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.16.experts.2.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.16.experts.3.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.17.experts.0.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.17.experts.1.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.17.experts.2.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.17.experts.3.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.18.experts.0.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.18.experts.1.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.18.experts.2.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.18.experts.3.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.19.experts.0.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.19.experts.1.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.19.experts.2.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.19.experts.3.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.20.experts.0.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.20.experts.1.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.20.experts.2.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.20.experts.3.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.21.experts.0.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.21.experts.1.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.21.experts.2.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.21.experts.3.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.22.experts.0.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.22.experts.1.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.22.experts.2.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.22.experts.3.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.23.experts.0.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.23.experts.1.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.23.experts.2.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.23.experts.3.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.24.experts.0.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.24.experts.1.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.24.experts.2.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.24.experts.3.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.25.experts.0.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.25.experts.1.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.25.experts.2.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.25.experts.3.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.26.experts.0.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.26.experts.1.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.26.experts.2.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.26.experts.3.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.27.experts.0.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.27.experts.1.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.27.experts.2.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.27.experts.3.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.28.experts.0.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.28.experts.1.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.28.experts.2.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.28.experts.3.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.29.experts.0.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.29.experts.1.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.29.experts.2.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.29.experts.3.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.30.experts.0.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.30.experts.1.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.30.experts.2.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.30.experts.3.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.31.experts.0.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.31.experts.1.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.31.experts.2.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.31.experts.3.post_attention_layernorm.weight": "model-00003-of-00004.safetensors", "model.layers.0.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.1.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.2.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.3.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.4.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.5.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.6.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.7.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.8.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.9.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.10.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.11.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.12.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.13.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.14.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.15.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.16.experts.0.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.16.experts.1.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.16.experts.2.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.16.experts.3.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.17.experts.0.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.17.experts.1.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.17.experts.2.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.17.experts.3.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.18.experts.0.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.18.experts.1.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.18.experts.2.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.18.experts.3.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.19.experts.0.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.19.experts.1.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.19.experts.2.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.19.experts.3.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.20.experts.0.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.20.experts.1.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.20.experts.2.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.20.experts.3.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.21.experts.0.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.21.experts.1.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.21.experts.2.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.21.experts.3.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.22.experts.0.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.22.experts.1.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.22.experts.2.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.22.experts.3.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.23.experts.0.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.23.experts.1.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.23.experts.2.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.23.experts.3.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.24.experts.0.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.24.experts.1.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.24.experts.2.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.24.experts.3.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.25.experts.0.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.25.experts.1.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.25.experts.2.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.25.experts.3.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.26.experts.0.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.26.experts.1.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.26.experts.2.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.26.experts.3.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.27.experts.0.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.27.experts.1.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.27.experts.2.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.27.experts.3.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.28.experts.0.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.28.experts.1.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.28.experts.2.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.28.experts.3.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.29.experts.0.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.29.experts.1.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.29.experts.2.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.29.experts.3.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.30.experts.0.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.30.experts.1.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.30.experts.2.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.30.experts.3.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.31.experts.0.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.31.experts.1.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.31.experts.2.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.31.experts.3.self_attn.q_proj.weight": "model-00003-of-00004.safetensors", "model.layers.0.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.1.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.2.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.3.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.4.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.5.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.6.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.7.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.8.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.9.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.10.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.11.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.12.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.13.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.14.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.15.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.16.experts.0.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.16.experts.1.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.16.experts.2.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.16.experts.3.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.17.experts.0.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.17.experts.1.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.17.experts.2.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.17.experts.3.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.18.experts.0.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.18.experts.1.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.18.experts.2.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.18.experts.3.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.19.experts.0.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.19.experts.1.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.19.experts.2.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.19.experts.3.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.20.experts.0.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.20.experts.1.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.20.experts.2.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.20.experts.3.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.21.experts.0.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.21.experts.1.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.21.experts.2.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.21.experts.3.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.22.experts.0.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.22.experts.1.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.22.experts.2.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.22.experts.3.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.23.experts.0.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.23.experts.1.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.23.experts.2.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.23.experts.3.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.24.experts.0.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.24.experts.1.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.24.experts.2.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.24.experts.3.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.25.experts.0.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.25.experts.1.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.25.experts.2.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.25.experts.3.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.26.experts.0.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.26.experts.1.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.26.experts.2.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.26.experts.3.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.27.experts.0.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.27.experts.1.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.27.experts.2.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.27.experts.3.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.28.experts.0.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.28.experts.1.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.28.experts.2.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.28.experts.3.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.29.experts.0.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.29.experts.1.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.29.experts.2.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.29.experts.3.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.30.experts.0.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.30.experts.1.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.30.experts.2.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.30.experts.3.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.31.experts.0.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.31.experts.1.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.31.experts.2.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.31.experts.3.self_attn.k_proj.weight": "model-00003-of-00004.safetensors", "model.layers.0.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.1.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.2.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.3.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.4.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.5.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.6.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.7.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.8.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.9.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.10.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.11.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.12.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.13.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.14.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.15.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.16.experts.0.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.16.experts.1.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.16.experts.2.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.16.experts.3.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.17.experts.0.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.17.experts.1.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.17.experts.2.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.17.experts.3.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.18.experts.0.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.18.experts.1.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.18.experts.2.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.18.experts.3.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.19.experts.0.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.19.experts.1.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.19.experts.2.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.19.experts.3.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.20.experts.0.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.20.experts.1.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.20.experts.2.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.20.experts.3.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.21.experts.0.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.21.experts.1.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.21.experts.2.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.21.experts.3.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.22.experts.0.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.22.experts.1.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.22.experts.2.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.22.experts.3.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.23.experts.0.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.23.experts.1.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.23.experts.2.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.23.experts.3.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.24.experts.0.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.24.experts.1.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.24.experts.2.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.24.experts.3.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.25.experts.0.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.25.experts.1.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.25.experts.2.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.25.experts.3.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.26.experts.0.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.26.experts.1.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.26.experts.2.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.26.experts.3.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.27.experts.0.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.27.experts.1.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.27.experts.2.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.27.experts.3.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.28.experts.0.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.28.experts.1.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.28.experts.2.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.28.experts.3.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.29.experts.0.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.29.experts.1.self_attn.v_proj.weight": "model-00003-of-00004.safetensors", "model.layers.29.experts.2.self_attn.v_proj.weight": "model-00004-of-00004.safetensors", "model.layers.29.experts.3.self_attn.v_proj.weight": "model-00004-of-00004.safetensors", "model.layers.30.experts.0.self_attn.v_proj.weight": "model-00004-of-00004.safetensors", "model.layers.30.experts.1.self_attn.v_proj.weight": "model-00004-of-00004.safetensors", "model.layers.30.experts.2.self_attn.v_proj.weight": "model-00004-of-00004.safetensors", "model.layers.30.experts.3.self_attn.v_proj.weight": "model-00004-of-00004.safetensors", "model.layers.31.experts.0.self_attn.v_proj.weight": "model-00004-of-00004.safetensors", "model.layers.31.experts.1.self_attn.v_proj.weight": "model-00004-of-00004.safetensors", "model.layers.31.experts.2.self_attn.v_proj.weight": "model-00004-of-00004.safetensors", "model.layers.31.experts.3.self_attn.v_proj.weight": "model-00004-of-00004.safetensors", "model.layers.0.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.1.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.2.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.3.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.4.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.5.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.6.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.7.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.8.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.9.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.10.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.11.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.12.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.13.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.14.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.15.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.16.experts.0.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.16.experts.1.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.16.experts.2.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.16.experts.3.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.17.experts.0.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.17.experts.1.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.17.experts.2.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.17.experts.3.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.18.experts.0.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.18.experts.1.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.18.experts.2.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.18.experts.3.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.19.experts.0.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.19.experts.1.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.19.experts.2.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.19.experts.3.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.20.experts.0.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.20.experts.1.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.20.experts.2.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.20.experts.3.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.21.experts.0.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.21.experts.1.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.21.experts.2.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.21.experts.3.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.22.experts.0.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.22.experts.1.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.22.experts.2.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.22.experts.3.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.23.experts.0.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.23.experts.1.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.23.experts.2.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.23.experts.3.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.24.experts.0.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.24.experts.1.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.24.experts.2.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.24.experts.3.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.25.experts.0.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.25.experts.1.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.25.experts.2.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.25.experts.3.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.26.experts.0.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.26.experts.1.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.26.experts.2.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.26.experts.3.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.27.experts.0.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.27.experts.1.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.27.experts.2.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.27.experts.3.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.28.experts.0.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.28.experts.1.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.28.experts.2.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.28.experts.3.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.29.experts.0.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.29.experts.1.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.29.experts.2.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.29.experts.3.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.30.experts.0.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.30.experts.1.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.30.experts.2.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.30.experts.3.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.31.experts.0.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.31.experts.1.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.31.experts.2.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.layers.31.experts.3.self_attn.o_proj.weight": "model-00004-of-00004.safetensors", "model.gate.weight": "model-00004-of-00004.safetensors"}}

hybrid_model_lv_mixed_moe/modeling_mixtral_model_level_hybrid.py

@@ -0,0 +1,1332 @@
+# coding=utf-8
+# Copyright 2023 MixtralModelLevelHybrid AI and the HuggingFace Inc. team. All rights reserved.
+#
+# This code is based on EleutherAI's GPT-NeoX library and the GPT-NeoX
+# and OPT implementations in this library. It has been modified from its
+# original forms to accommodate minor architectural differences compared
+# to GPT-NeoX and OPT used by the Meta AI team that trained the model.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+""" PyTorch MixtralModelLevelHybrid model."""
+import inspect
+import math
+import warnings
+from typing import List, Optional, Tuple, Union
+
+import torch
+import torch.nn.functional as F
+import torch.utils.checkpoint
+from torch import nn
+from torch.nn import CrossEntropyLoss
+from transformers.activations import ACT2FN
+from transformers.cache_utils import Cache, DynamicCache
+from transformers.modeling_attn_mask_utils import (
+    _prepare_4d_causal_attention_mask,
+    _prepare_4d_causal_attention_mask_for_sdpa
+)
+from transformers.modeling_outputs import (
+    BaseModelOutputWithPast,
+    CausalLMOutputWithPast
+)
+from transformers.modeling_utils import PreTrainedModel
+from transformers.utils import (
+    add_start_docstrings,
+    add_start_docstrings_to_model_forward,
+    is_flash_attn_2_available,
+    is_flash_attn_greater_or_equal_2_10,
+    logging,
+)
+
+from .configuration_mixtral_model_level_hybrid import MixtralModelLevelHybridConfig
+
+if is_flash_attn_2_available():
+    from flash_attn import flash_attn_func, flash_attn_varlen_func
+    from flash_attn.bert_padding import index_first_axis, pad_input, unpad_input  # noqa
+
+    _flash_supports_window_size = "window_size" in list(inspect.signature(flash_attn_func).parameters)
+
+logger = logging.get_logger(__name__)
+
+_CONFIG_FOR_DOC = "MixtralModelLevelHybridConfig"
+
+
+# Copied from transformers.models.llama.modeling_llama._get_unpad_data
+def _get_unpad_data(attention_mask):
+    seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
+    indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
+    max_seqlen_in_batch = seqlens_in_batch.max().item()
+    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.int32), (1, 0))
+    return (
+        indices,
+        cu_seqlens,
+        max_seqlen_in_batch,
+    )
+
+
+# Copied from transformers.models.llama.modeling_llama.LlamaRMSNorm with Llama->MixtralModelLevelHybrid
+class MixtralModelLevelHybridRMSNorm(nn.Module):
+    def __init__(self, hidden_size, eps=1e-6):
+        """
+        MixtralModelLevelHybridRMSNorm 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)
+
+
+# copied from transformers.models.llama.modeling_llama.LlamaRotaryEmbedding with Llama->MixtralModelLevelHybrid
+# TODO @Arthur no longer copied from LLama after static cache
+class MixtralModelLevelHybridRotaryEmbedding(nn.Module):
+    def __init__(self, dim, max_position_embeddings=2048, base=10000, device=None):
+        super().__init__()
+
+        self.dim = dim
+        self.max_position_embeddings = max_position_embeddings
+        self.base = base
+        inv_freq = 1.0 / (self.base ** (torch.arange(0, self.dim, 2, dtype=torch.int64).float().to(device) / self.dim))
+        self.register_buffer("inv_freq", inv_freq, persistent=False)
+
+        # Build here to make `torch.jit.trace` work.
+        self._set_cos_sin_cache(
+            seq_len=max_position_embeddings, device=self.inv_freq.device, dtype=torch.get_default_dtype()
+        )
+
+    def _set_cos_sin_cache(self, seq_len, device, dtype):
+        self.max_seq_len_cached = seq_len
+        t = torch.arange(self.max_seq_len_cached, device=device, dtype=torch.int64).type_as(self.inv_freq)
+
+        freqs = torch.outer(t, self.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().to(dtype), persistent=False)
+        self.register_buffer("sin_cached", emb.sin().to(dtype), 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, dtype=x.dtype)
+
+        return (
+            self.cos_cached[:seq_len].to(dtype=x.dtype),
+            self.sin_cached[:seq_len].to(dtype=x.dtype),
+        )
+
+
+# Copied from transformers.models.llama.modeling_llama.rotate_half
+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)
+
+
+# copied from transformers.models.llama.modeling_llama.apply_rotary_pos_emb
+# TODO @Arthur no longer copied from LLama after static cache
+def apply_rotary_pos_emb(q, k, cos, sin, position_ids, unsqueeze_dim=1):
+    """Applies Rotary Position Embedding to the query and key tensors.
+
+    Args:
+        q (`torch.Tensor`): The query tensor.
+        k (`torch.Tensor`): The key tensor.
+        cos (`torch.Tensor`): The cosine part of the rotary embedding.
+        sin (`torch.Tensor`): The sine part of the rotary embedding.
+        position_ids (`torch.Tensor`):
+            The position indices of the tokens corresponding to the query and key tensors. For example, this can be
+            used to pass offsetted position ids when working with a KV-cache.
+        unsqueeze_dim (`int`, *optional*, defaults to 1):
+            The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and
+            sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note
+            that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and
+            k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes
+            cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have
+            the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.
+    Returns:
+        `tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding.
+    """
+    cos = cos[position_ids].unsqueeze(unsqueeze_dim)
+    sin = sin[position_ids].unsqueeze(unsqueeze_dim)
+    q_embed = (q * cos) + (rotate_half(q) * sin)
+    k_embed = (k * cos) + (rotate_half(k) * sin)
+    return q_embed, k_embed
+
+
+class MixtralModelLevelHybridMLP(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.config = config
+        self.hidden_size = config.hidden_size
+        self.intermediate_size = config.intermediate_size
+        self.gate_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False)
+        self.up_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False)
+        self.down_proj = nn.Linear(self.intermediate_size, self.hidden_size, bias=False)
+        self.act_fn = ACT2FN[config.hidden_act]
+
+    def forward(self, x):
+        return self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))
+
+
+# Copied from transformers.models.llama.modeling_llama.repeat_kv
+def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
+    """
+    This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch,
+    num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim)
+    """
+    batch, num_key_value_heads, slen, head_dim = hidden_states.shape
+    if n_rep == 1:
+        return hidden_states
+    hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)
+    return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)
+
+
+class MixtralModelLevelHybridAttention(nn.Module):
+    """
+    Multi-headed attention from 'Attention Is All You Need' paper. Modified to use sliding window attention: Longformer
+    and "Generating Long Sequences with Sparse Transformers".
+    """
+
+    def __init__(self, config: MixtralModelLevelHybridConfig, layer_idx: Optional[int] = None):
+        super().__init__()
+        self.config = config
+        self.layer_idx = layer_idx
+        if layer_idx is None:
+            logger.warning_once(
+                f"Instantiating {self.__class__.__name__} without passing a `layer_idx` is not recommended and will "
+                "lead to errors during the forward call if caching is used. Please make sure to provide a `layer_idx` "
+                "when creating this class."
+            )
+
+        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
+        self.rope_theta = config.rope_theta
+        self.is_causal = True
+        self.attention_dropout = config.attention_dropout
+
+        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 = MixtralModelLevelHybridRotaryEmbedding(
+            self.head_dim,
+            max_position_embeddings=self.max_position_embeddings,
+            base=self.rope_theta,
+        )
+
+    def _shape(self, tensor: torch.Tensor, seq_len: int, bsz: int):
+        return tensor.view(bsz, seq_len, self.num_heads, self.head_dim).transpose(1, 2).contiguous()
+
+    def forward(
+            self,
+            hidden_states: torch.Tensor,
+            attention_mask: Optional[torch.Tensor] = None,
+            position_ids: Optional[torch.LongTensor] = None,
+            past_key_value: Optional[Cache] = None,
+            output_attentions: bool = False,
+            use_cache: bool = False,
+            **kwargs,
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+        if "padding_mask" in kwargs:
+            warnings.warn(
+                "Passing `padding_mask` is deprecated and will be removed in v4.37. Please make sure use `attention_mask` instead.`"
+            )
+        bsz, q_len, _ = hidden_states.size()
+
+        query_states = self.q_proj(hidden_states)
+        key_states = self.k_proj(hidden_states)
+        value_states = self.v_proj(hidden_states)
+
+        query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+        key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+        value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+
+        kv_seq_len = key_states.shape[-2]
+        if past_key_value is not None:
+            if self.layer_idx is None:
+                raise ValueError(
+                    f"The cache structure has changed since version v4.36. If you are using {self.__class__.__name__} "
+                    "for auto-regressive decoding with k/v caching, please make sure to initialize the attention class "
+                    "with a layer index."
+                )
+            kv_seq_len += past_key_value.get_usable_length(kv_seq_len, self.layer_idx)
+        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)
+
+        if past_key_value is not None:
+            cache_kwargs = {"sin": sin, "cos": cos}  # Specific to RoPE models
+            key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
+
+        # repeat k/v heads if n_kv_heads < n_heads
+        key_states = repeat_kv(key_states, self.num_key_value_groups)
+        value_states = repeat_kv(value_states, self.num_key_value_groups)
+
+        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 {attention_mask.size()}"
+                )
+
+            attn_weights = attn_weights + attention_mask
+
+        # upcast attention to fp32
+        attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query_states.dtype)
+        attn_weights = nn.functional.dropout(attn_weights, p=self.attention_dropout, training=self.training)
+        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()}"
+            )
+
+        attn_output = attn_output.transpose(1, 2).contiguous()
+        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 MixtralModelLevelHybridFlashAttention2(MixtralModelLevelHybridAttention):
+    """
+    MixtralModelLevelHybrid flash attention module. This module inherits from `MixtralModelLevelHybridAttention` as the weights of the module stays
+    untouched. The only required change would be on the forward pass where it needs to correctly call the public API of
+    flash attention and deal with padding tokens in case the input contains any of them.
+    """
+
+    # Copied from transformers.models.llama.modeling_llama.LlamaFlashAttention2.__init__
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+
+        # TODO: Should be removed once Flash Attention for RoCm is bumped to 2.1.
+        # flash_attn<2.1 generates top-left aligned causal mask, while what is needed here is bottom-right alignement, that was made default for flash_attn>=2.1. This attribute is used to handle this difference. Reference: https://github.com/Dao-AILab/flash-attention/releases/tag/v2.1.0.
+        # Beware that with flash_attn<2.1, using q_seqlen != k_seqlen (except for the case q_seqlen == 1) produces a wrong mask (top-left).
+        self._flash_attn_uses_top_left_mask = not is_flash_attn_greater_or_equal_2_10()
+
+    def forward(
+            self,
+            hidden_states: torch.Tensor,
+            attention_mask: Optional[torch.Tensor] = None,
+            position_ids: Optional[torch.LongTensor] = None,
+            past_key_value: Optional[Cache] = None,
+            output_attentions: bool = False,
+            use_cache: bool = False,
+            **kwargs,
+    ):
+        if "padding_mask" in kwargs:
+            warnings.warn(
+                "Passing `padding_mask` is deprecated and will be removed in v4.37. Please make sure use `attention_mask` instead.`"
+            )
+
+            # overwrite attention_mask with padding_mask
+            attention_mask = kwargs.pop("padding_mask")
+        bsz, q_len, _ = hidden_states.size()
+
+        query_states = self.q_proj(hidden_states)
+        key_states = self.k_proj(hidden_states)
+        value_states = self.v_proj(hidden_states)
+
+        query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+        key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+        value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+
+        kv_seq_len = key_states.shape[-2]
+        if past_key_value is not None:
+            if self.layer_idx is None:
+                raise ValueError(
+                    f"The cache structure has changed since version v4.36. If you are using {self.__class__.__name__} "
+                    "for auto-regressive decoding with k/v caching, please make sure to initialize the attention class "
+                    "with a layer index."
+                )
+            kv_seq_len += past_key_value.get_usable_length(kv_seq_len, self.layer_idx)
+
+        # Because the input can be padded, the absolute sequence length depends on the max position id.
+        rotary_seq_len = max(kv_seq_len, position_ids[:, -1].max().item()) + 1
+        cos, sin = self.rotary_emb(value_states, seq_len=rotary_seq_len)
+
+        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids)
+
+        use_sliding_windows = (
+                _flash_supports_window_size
+                and getattr(self.config, "sliding_window", None) is not None
+                and kv_seq_len > self.config.sliding_window
+        )
+
+        if not _flash_supports_window_size:
+            logger.warning_once(
+                "The current flash attention version does not support sliding window attention, for a more memory efficient implementation"
+                " make sure to upgrade flash-attn library."
+            )
+
+        if past_key_value is not None:
+            # Activate slicing cache only if the config has a value `sliding_windows` attribute
+            cache_has_contents = past_key_value.get_seq_length(self.layer_idx) > 0
+            if (
+                    getattr(self.config, "sliding_window", None) is not None
+                    and kv_seq_len > self.config.sliding_window
+                    and cache_has_contents
+            ):
+                slicing_tokens = 1 - self.config.sliding_window
+
+                past_key = past_key_value[self.layer_idx][0]
+                past_value = past_key_value[self.layer_idx][1]
+
+                past_key = past_key[:, :, slicing_tokens:, :].contiguous()
+                past_value = past_value[:, :, slicing_tokens:, :].contiguous()
+
+                if past_key.shape[-2] != self.config.sliding_window - 1:
+                    raise ValueError(
+                        f"past key must have a shape of (`batch_size, num_heads, self.config.sliding_window-1, head_dim`), got"
+                        f" {past_key.shape}"
+                    )
+
+                if attention_mask is not None:
+                    attention_mask = attention_mask[:, slicing_tokens:]
+                    attention_mask = torch.cat([attention_mask, torch.ones_like(attention_mask[:, -1:])], dim=-1)
+
+            cache_kwargs = {"sin": sin, "cos": cos}  # Specific to RoPE models
+            key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
+
+        # repeat k/v heads if n_kv_heads < n_heads
+        key_states = repeat_kv(key_states, self.num_key_value_groups)
+        value_states = repeat_kv(value_states, self.num_key_value_groups)
+        dropout_rate = 0.0 if not self.training else self.attention_dropout
+
+        # In PEFT, usually we cast the layer norms in float32 for training stability reasons
+        # therefore the input hidden states gets silently casted in float32. Hence, we need
+        # cast them back in float16 just to be sure everything works as expected.
+        input_dtype = query_states.dtype
+        if input_dtype == torch.float32:
+            if torch.is_autocast_enabled():
+                target_dtype = torch.get_autocast_gpu_dtype()
+            # Handle the case where the model is quantized
+            elif hasattr(self.config, "_pre_quantization_dtype"):
+                target_dtype = self.config._pre_quantization_dtype
+            else:
+                target_dtype = self.q_proj.weight.dtype
+
+            logger.warning_once(
+                f"The input hidden states seems to be silently casted in float32, this might be related to"
+                f" the fact you have upcasted embedding or layer norm layers in float32. We will cast back the input in"
+                f" {target_dtype}."
+            )
+
+            query_states = query_states.to(target_dtype)
+            key_states = key_states.to(target_dtype)
+            value_states = value_states.to(target_dtype)
+
+        # Reashape to the expected shape for Flash Attention
+        query_states = query_states.transpose(1, 2)
+        key_states = key_states.transpose(1, 2)
+        value_states = value_states.transpose(1, 2)
+
+        attn_output = self._flash_attention_forward(
+            query_states,
+            key_states,
+            value_states,
+            attention_mask,
+            q_len,
+            dropout=dropout_rate,
+            use_sliding_windows=use_sliding_windows,
+        )
+
+        attn_output = attn_output.reshape(bsz, q_len, self.hidden_size).contiguous()
+        attn_output = self.o_proj(attn_output)
+
+        if not output_attentions:
+            attn_weights = None
+
+        return attn_output, attn_weights, past_key_value
+
+    def _flash_attention_forward(
+            self,
+            query_states,
+            key_states,
+            value_states,
+            attention_mask,
+            query_length,
+            dropout=0.0,
+            softmax_scale=None,
+            use_sliding_windows=False,
+    ):
+        """
+        Calls the forward method of Flash Attention - if the input hidden states contain at least one padding token
+        first unpad the input, then computes the attention scores and pad the final attention scores.
+
+        Args:
+            query_states (`torch.Tensor`):
+                Input query states to be passed to Flash Attention API
+            key_states (`torch.Tensor`):
+                Input key states to be passed to Flash Attention API
+            value_states (`torch.Tensor`):
+                Input value states to be passed to Flash Attention API
+            attention_mask (`torch.Tensor`):
+                The padding mask - corresponds to a tensor of size `(batch_size, seq_len)` where 0 stands for the
+                position of padding tokens and 1 for the position of non-padding tokens.
+            dropout (`int`, *optional*):
+                Attention dropout
+            softmax_scale (`float`, *optional*):
+                The scaling of QK^T before applying softmax. Default to 1 / sqrt(head_dim)
+            use_sliding_windows (`bool`, *optional*):
+                Whether to activate sliding window attention.
+        """
+        if not self._flash_attn_uses_top_left_mask:
+            causal = self.is_causal
+        else:
+            # TODO: Remove the `query_length != 1` check once Flash Attention for RoCm is bumped to 2.1. For details, please see the comment in LlamaFlashAttention2 __init__.
+            causal = self.is_causal and query_length != 1
+
+        # Contains at least one padding token in the sequence
+        if attention_mask is not None:
+            batch_size = query_states.shape[0]
+            query_states, key_states, value_states, indices_q, cu_seq_lens, max_seq_lens = self._upad_input(
+                query_states, key_states, value_states, attention_mask, query_length
+            )
+
+            cu_seqlens_q, cu_seqlens_k = cu_seq_lens
+            max_seqlen_in_batch_q, max_seqlen_in_batch_k = max_seq_lens
+
+            if not use_sliding_windows:
+                attn_output_unpad = flash_attn_varlen_func(
+                    query_states,
+                    key_states,
+                    value_states,
+                    cu_seqlens_q=cu_seqlens_q,
+                    cu_seqlens_k=cu_seqlens_k,
+                    max_seqlen_q=max_seqlen_in_batch_q,
+                    max_seqlen_k=max_seqlen_in_batch_k,
+                    dropout_p=dropout,
+                    softmax_scale=softmax_scale,
+                    causal=causal,
+                )
+            else:
+                attn_output_unpad = flash_attn_varlen_func(
+                    query_states,
+                    key_states,
+                    value_states,
+                    cu_seqlens_q=cu_seqlens_q,
+                    cu_seqlens_k=cu_seqlens_k,
+                    max_seqlen_q=max_seqlen_in_batch_q,
+                    max_seqlen_k=max_seqlen_in_batch_k,
+                    dropout_p=dropout,
+                    softmax_scale=softmax_scale,
+                    causal=causal,
+                    window_size=(self.config.sliding_window, self.config.sliding_window),
+                )
+
+            attn_output = pad_input(attn_output_unpad, indices_q, batch_size, query_length)
+        else:
+            if not use_sliding_windows:
+                attn_output = flash_attn_func(
+                    query_states,
+                    key_states,
+                    value_states,
+                    dropout,
+                    softmax_scale=softmax_scale,
+                    causal=causal,
+                )
+            else:
+                attn_output = flash_attn_func(
+                    query_states,
+                    key_states,
+                    value_states,
+                    dropout,
+                    softmax_scale=softmax_scale,
+                    causal=causal,
+                    window_size=(self.config.sliding_window, self.config.sliding_window),
+                )
+
+        return attn_output
+
+    def _upad_input(self, query_layer, key_layer, value_layer, attention_mask, query_length):
+        batch_size, kv_seq_len, num_heads, head_dim = key_layer.shape
+
+        # On the first iteration we need to properly re-create the padding mask
+        # by slicing it on the proper place
+        if kv_seq_len != attention_mask.shape[-1]:
+            attention_mask_num_tokens = attention_mask.shape[-1]
+            attention_mask = attention_mask[:, attention_mask_num_tokens - kv_seq_len:]
+
+        indices_k, cu_seqlens_k, max_seqlen_in_batch_k = _get_unpad_data(attention_mask)
+
+        key_layer = index_first_axis(key_layer.reshape(batch_size * kv_seq_len, num_heads, head_dim), indices_k)
+        value_layer = index_first_axis(value_layer.reshape(batch_size * kv_seq_len, num_heads, head_dim), indices_k)
+
+        if query_length == kv_seq_len:
+            query_layer = index_first_axis(
+                query_layer.reshape(batch_size * kv_seq_len, num_heads, head_dim), indices_k
+            )
+            cu_seqlens_q = cu_seqlens_k
+            max_seqlen_in_batch_q = max_seqlen_in_batch_k
+            indices_q = indices_k
+        elif query_length == 1:
+            max_seqlen_in_batch_q = 1
+            cu_seqlens_q = torch.arange(
+                batch_size + 1, dtype=torch.int32, device=query_layer.device
+            )  # There is a memcpy here, that is very bad.
+            indices_q = cu_seqlens_q[:-1]
+            query_layer = query_layer.squeeze(1)
+        else:
+            # The -q_len: slice assumes left padding.
+            attention_mask = attention_mask[:, -query_length:]
+            query_layer, indices_q, cu_seqlens_q, max_seqlen_in_batch_q = unpad_input(query_layer, attention_mask)
+
+        return (
+            query_layer,
+            key_layer,
+            value_layer,
+            indices_q,
+            (cu_seqlens_q, cu_seqlens_k),
+            (max_seqlen_in_batch_q, max_seqlen_in_batch_k),
+        )
+
+
+# copied from transformers.models.llama.modeling_llama.LlamaSdpaAttention with Llama->MixtralModelLevelHybrid
+# TODO @Arthur no longer copied from LLama after static cache
+class MixtralModelLevelHybridSdpaAttention(MixtralModelLevelHybridAttention):
+    """
+    MixtralModelLevelHybrid attention module using torch.nn.functional.scaled_dot_product_attention. This module inherits from
+    `MixtralModelLevelHybridAttention` as the weights of the module stays untouched. The only changes are on the forward pass to adapt to
+    SDPA API.
+    """
+
+    # Adapted from MixtralModelLevelHybridAttention.forward
+    def forward(
+            self,
+            hidden_states: torch.Tensor,
+            attention_mask: Optional[torch.Tensor] = None,
+            position_ids: Optional[torch.LongTensor] = None,
+            past_key_value: Optional[Cache] = None,
+            output_attentions: bool = False,
+            use_cache: bool = False,
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+        if output_attentions:
+            # TODO: Improve this warning with e.g. `model.config.attn_implementation = "manual"` once this is implemented.
+            logger.warning_once(
+                "MixtralModelLevelHybridModel is using MixtralModelLevelHybridSdpaAttention, but `torch.nn.functional.scaled_dot_product_attention` does not support `output_attentions=True`. Falling back to the manual attention implementation, "
+                'but specifying the manual implementation will be required from Transformers version v5.0.0 onwards. This warning can be removed using the argument `attn_implementation="eager"` when loading the model.'
+            )
+            return super().forward(
+                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,
+            )
+
+        bsz, q_len, _ = hidden_states.size()
+
+        query_states = self.q_proj(hidden_states)
+        key_states = self.k_proj(hidden_states)
+        value_states = self.v_proj(hidden_states)
+
+        query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+        key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+        value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+
+        kv_seq_len = key_states.shape[-2]
+        if past_key_value is not None:
+            kv_seq_len += past_key_value.get_usable_length(kv_seq_len, self.layer_idx)
+        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)
+
+        if past_key_value is not None:
+            cache_kwargs = {"sin": sin, "cos": cos}  # Specific to RoPE models
+            key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
+
+        key_states = repeat_kv(key_states, self.num_key_value_groups)
+        value_states = repeat_kv(value_states, self.num_key_value_groups)
+
+        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 {attention_mask.size()}"
+                )
+
+        # SDPA with memory-efficient backend is currently (torch==2.1.2) bugged with non-contiguous inputs with custom attn_mask,
+        # Reference: https://github.com/pytorch/pytorch/issues/112577.
+        if query_states.device.type == "cuda" and attention_mask is not None:
+            query_states = query_states.contiguous()
+            key_states = key_states.contiguous()
+            value_states = value_states.contiguous()
+
+        attn_output = torch.nn.functional.scaled_dot_product_attention(
+            query_states,
+            key_states,
+            value_states,
+            attn_mask=attention_mask,
+            dropout_p=self.attention_dropout if self.training else 0.0,
+            # The q_len > 1 is necessary to match with AttentionMaskConverter.to_causal_4d that does not create a causal mask in case q_len == 1.
+            is_causal=self.is_causal and attention_mask is None and q_len > 1,
+        )
+
+        attn_output = attn_output.transpose(1, 2).contiguous()
+        attn_output = attn_output.view(bsz, q_len, self.hidden_size)
+
+        attn_output = self.o_proj(attn_output)
+
+        return attn_output, None, past_key_value
+
+
+MIXTRAL_ATTENTION_CLASSES = {
+    "eager": MixtralModelLevelHybridAttention,
+    "flash_attention_2": MixtralModelLevelHybridFlashAttention2,
+    "sdpa": MixtralModelLevelHybridSdpaAttention,
+}
+
+
+class MixtralModelLevelHybridDenseDecoderLayer(nn.Module):
+    def __init__(self, config: MixtralModelLevelHybridConfig, layer_idx: int):
+        super().__init__()
+        self.hidden_size = config.hidden_size
+
+        self.self_attn = MIXTRAL_ATTENTION_CLASSES[config._attn_implementation](config, layer_idx)
+
+        self.mlp = MixtralModelLevelHybridMLP(config)
+        self.input_layernorm = MixtralModelLevelHybridRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.post_attention_layernorm = MixtralModelLevelHybridRMSNorm(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,
+            **kwargs,
+    ) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]:
+        if "padding_mask" in kwargs:
+            warnings.warn(
+                "Passing `padding_mask` is deprecated and will be removed in v4.37. Please make sure use `attention_mask` instead.`"
+            )
+
+        residual = hidden_states
+
+        hidden_states = self.input_layernorm(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.post_attention_layernorm(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
+
+
+class MixtralModelLevelHybridSparseDecoderLayer(nn.Module):
+    """This is a transformer model level hybrid sparse moe (i.e. DecoderLayer) with top-2 gating."""
+
+    def __init__(self, config: MixtralModelLevelHybridConfig, layer_idx: int):
+        super().__init__()
+        self.sequence_pooler = config.sequence_pooler
+        self.hidden_dim = config.hidden_size
+        self.num_experts = config.num_local_experts
+        self.top_k = config.num_experts_per_tok
+
+        self.experts = nn.ModuleList([
+            MixtralModelLevelHybridDenseDecoderLayer(config, layer_idx) for _ in range(self.num_experts)
+        ])
+
+    def forward(
+            self,
+            hidden_states: torch.Tensor,
+            router_logits: Optional[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,
+            output_router_logits: Optional[bool] = None,
+            use_cache: Optional[bool] = False,
+            **kwargs,
+    ) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]:
+        batch_size, sequence_length, hidden_dim = hidden_states.shape
+
+        # sentence routing
+        # design space: first token / mean without padding tokens / mean with padding tokens
+        if self.sequence_pooler == "mean":
+            sent_states = hidden_states.mean(dim=1, keepdim=True).view(-1, hidden_dim)
+        elif self.sequence_pooler == "first_token":
+            sent_states = hidden_states[:, 0].view(-1, hidden_dim)
+        else:
+            raise NotImplementedError(f"{self.sequence_pooler} not implemented yet.")
+
+        routing_weights = F.softmax(router_logits, dim=1, dtype=torch.float)
+        routing_weights, selected_experts = torch.topk(routing_weights, self.top_k, dim=-1)
+        routing_weights /= routing_weights.sum(dim=-1, keepdim=True)
+        routing_weights = routing_weights.unsqueeze(1).repeat(1, sequence_length, 1)
+        routing_weights = routing_weights.to(hidden_states.dtype)
+
+        final_hidden_states = torch.zeros(
+            (batch_size, sequence_length, hidden_dim), dtype=hidden_states.dtype, device=hidden_states.device
+        )
+        expert_mask = torch.nn.functional.one_hot(selected_experts, num_classes=self.num_experts).permute(2, 1, 0)
+
+        for expert_idx in range(self.num_experts):
+            expert_layer = self.experts[expert_idx]
+            idx, top_x = torch.where(expert_mask[expert_idx])
+
+            if top_x.shape[0] == 0:
+                continue
+            # in torch it is faster to index using lists than torch tensors
+            top_x_list = top_x.tolist()
+            idx_list = idx.tolist()
+
+            expert_hidden_states = hidden_states[top_x_list]
+            expert_position_ids = torch.arange(sequence_length).unsqueeze(0).repeat(expert_hidden_states.shape[0], 1)
+            expert_attention_mask = attention_mask[top_x_list] if attention_mask is not None else None
+
+            expert_state = expert_layer(
+                expert_hidden_states,
+                attention_mask=expert_attention_mask,
+                position_ids=expert_position_ids,
+                past_key_value=past_key_value,
+                output_attentions=output_attentions,
+                output_router_logits=output_router_logits,
+                use_cache=use_cache,
+            )[0]
+            current_hidden_states = expert_state * routing_weights[top_x_list, :, idx_list, None]
+
+            final_hidden_states.index_add_(0, top_x, current_hidden_states.to(hidden_states.dtype))
+
+        outputs = (final_hidden_states,)
+
+        # assert not use_cache, "use_cache is not supported yet for model level hybrid Sparse MoE"
+        # assert not output_attentions, "output_attentions is not supported yet for model level hybrid Sparse MoE"
+        # if output_attentions:
+        #     outputs += (self_attn_weights,)
+
+        # if use_cache:
+        #     outputs += (present_key_value,)
+
+        if output_router_logits:
+            outputs += (router_logits,)
+
+        return outputs
+
+
+MIXTRAL_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 ([`MixtralModelLevelHybridConfig`]):
+            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 MixtralModelLevelHybrid Model outputting raw hidden-states without any specific head on top.",
+    MIXTRAL_START_DOCSTRING,
+)
+class MixtralModelLevelHybridPreTrainedModel(PreTrainedModel):
+    config_class = MixtralModelLevelHybridConfig
+    base_model_prefix = "model"
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["MixtralModelLevelHybridDecoderLayer"]
+    _skip_keys_device_placement = "past_key_values"
+    _supports_flash_attn_2 = True
+    _supports_sdpa = True
+    _supports_cache_class = True
+
+    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_()
+
+
+MIXTRAL_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 (`Cache` or `tuple(tuple(torch.FloatTensor))`, *optional*):
+            Pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention
+            blocks) that can be used to speed up sequential decoding. This typically consists in the `past_key_values`
+            returned by the model at a previous stage of decoding, when `use_cache=True` or `config.use_cache=True`.
+
+            Two formats are allowed:
+            - a [`~cache_utils.Cache`] instance;
+            - 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)`). This is also known as the legacy
+            cache format.
+
+            The model will output the same cache format that is fed as input. If no `past_key_values` are passed, the
+            legacy cache format will be returned.
+
+            If `past_key_values` are used, the user can optionally input only the last `input_ids` (those that don't
+            have their past key value states given to this model) of shape `(batch_size, 1)` instead of all `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 MixtralModelLevelHybrid Model outputting raw hidden-states without any specific head on top.",
+    MIXTRAL_START_DOCSTRING,
+)
+class MixtralModelLevelHybridModel(MixtralModelLevelHybridPreTrainedModel):
+    """
+    Transformer decoder consisting of *config.num_hidden_layers* layers. Each layer is a [`MixtralModelLevelHybridDecoderLayer`]
+
+    Args:
+        config: MixtralModelLevelHybridConfig
+    """
+
+    def __init__(self, config: MixtralModelLevelHybridConfig):
+        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(
+            [MixtralModelLevelHybridSparseDecoderLayer(
+                config, layer_idx
+            ) if layer_idx >= config.num_dense_layers else MixtralModelLevelHybridDenseDecoderLayer(config, layer_idx)
+             for layer_idx in range(config.num_hidden_layers)]
+        )
+        self.num_dense_layers = config.num_dense_layers
+        self.sequence_pooler = config.sequence_pooler
+        self.gate = nn.Linear(config.hidden_size, config.num_local_experts, bias=False)
+        self._attn_implementation = config._attn_implementation
+        self.norm = MixtralModelLevelHybridRMSNorm(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
+
+    @add_start_docstrings_to_model_forward(MIXTRAL_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")
+
+        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
+
+        past_key_values_length = 0
+
+        if use_cache:
+            use_legacy_cache = not isinstance(past_key_values, Cache)
+            if use_legacy_cache:
+                past_key_values = DynamicCache.from_legacy_cache(past_key_values)
+            past_key_values_length = past_key_values.get_usable_length(seq_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 attention_mask is not None and self._attn_implementation == "flash_attention_2" and use_cache:
+            is_padding_right = attention_mask[:, -1].sum().item() != batch_size
+            if is_padding_right:
+                raise ValueError(
+                    "You are attempting to perform batched generation with padding_side='right'"
+                    " this may lead to unexpected behaviour for Flash Attention version of MixtralModelLevelHybrid. Make sure to "
+                    " call `tokenizer.padding_side  = 'left'` before tokenizing the input. "
+                )
+
+        if self._attn_implementation == "flash_attention_2":
+            # 2d mask is passed through the layers
+            attention_mask = attention_mask if (attention_mask is not None and 0 in attention_mask) else None
+        elif self._attn_implementation == "sdpa" and not output_attentions:
+            # output_attentions=True can not be supported when using SDPA, and we fall back on
+            # the manual implementation that requires a 4D causal mask in all cases.
+            attention_mask = _prepare_4d_causal_attention_mask_for_sdpa(
+                attention_mask,
+                (batch_size, seq_length),
+                inputs_embeds,
+                past_key_values_length,
+            )
+        else:
+            # 4d mask is passed through the layers
+            attention_mask = _prepare_4d_causal_attention_mask(
+                attention_mask,
+                (batch_size, seq_length),
+                inputs_embeds,
+                past_key_values_length,
+                sliding_window=self.config.sliding_window,
+            )
+
+        hidden_states = inputs_embeds
+
+        # decoder layers
+        all_hidden_states = () if output_hidden_states else None
+        all_self_attns = () if output_attentions else None
+        router_logits = None
+
+        for layer_idx, decoder_layer in enumerate(self.layers):
+            if output_hidden_states:
+                all_hidden_states += (hidden_states,)
+
+            if layer_idx == self.num_dense_layers:
+                # Calculate router logits
+                batch_size, sequence_length, hidden_dim = hidden_states.shape
+                if self.sequence_pooler == "mean":
+                    sent_states = hidden_states.mean(dim=1, keepdim=True).view(-1, hidden_dim)
+                elif self.sequence_pooler == "first_token":
+                    sent_states = hidden_states[:, 0].view(-1, hidden_dim)
+                else:
+                    raise NotImplementedError(f"{self.sequence_pooler} not implemented yet.")
+
+                router_logits = self.gate(sent_states)
+
+            if self.gradient_checkpointing and self.training:
+                raise NotImplementedError("Gradient checkpointing is not supported in MixtralModelLevelHybrid yet.")
+            else:
+                layer_outputs = decoder_layer(
+                    hidden_states,
+                    router_logits=router_logits,
+                    attention_mask=attention_mask,
+                    position_ids=position_ids,
+                    past_key_value=past_key_values,
+                    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 = 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 MixtralModelLevelHybridForCausalLM(MixtralModelLevelHybridPreTrainedModel):
+    """
+    Examples
+    --------
+    >>> model = MixtralModelLevelHybridForCausalLM(MixtralModelLevelHybridConfig(vocab_size=100, hidden_size=16, intermediate_size=32, num_dense_layers=0, num_attention_heads=4))
+    >>> outputs = model(input_ids=torch.randint(10, 50, (3,5)))
+    >>> outputs.logits.shape
+    """
+    _tied_weights_keys = ["lm_head.weight"]
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.model = MixtralModelLevelHybridModel(config)
+        self.vocab_size = config.vocab_size
+        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
+
+    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]:
+
+        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)
+        logits = logits.float()
+
+        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
+            shift_logits = shift_logits.view(-1, self.config.vocab_size)
+            shift_labels = shift_labels.view(-1)
+            # Ensure tensors are on the same device
+            shift_labels = shift_labels.to(shift_logits.device)
+            loss_fct = CrossEntropyLoss()
+            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
+    ):
+        # Omit tokens covered by past_key_values
+        if past_key_values is not None:
+            if isinstance(past_key_values, Cache):
+                cache_length = past_key_values.get_seq_length()
+                past_length = past_key_values.seen_tokens
+                max_cache_length = past_key_values.get_max_length()
+            else:
+                cache_length = past_length = past_key_values[0][0].shape[2]
+                max_cache_length = None
+
+            # Keep only the unprocessed tokens:
+            # 1 - If the length of the attention_mask exceeds the length of input_ids, then we are in a setting where
+            # some of the inputs are exclusively passed as part of the cache (e.g. when passing input_embeds as
+            # input)
+            if attention_mask is not None and attention_mask.shape[1] > input_ids.shape[1]:
+                input_ids = input_ids[:, -(attention_mask.shape[1] - past_length):]
+            # 2 - If the past_length is smaller than input_ids', then input_ids holds all input tokens. We can discard
+            # input_ids based on the past_length.
+            elif past_length < input_ids.shape[1]:
+                input_ids = input_ids[:, past_length:]
+            # 3 - Otherwise (past_length >= input_ids.shape[1]), let's assume input_ids only has unprocessed tokens.
+
+            # If we are about to go beyond the maximum cache length, we need to crop the input attention mask.
+            if (
+                    max_cache_length is not None
+                    and attention_mask is not None
+                    and cache_length + input_ids.shape[1] > max_cache_length
+            ):
+                attention_mask = attention_mask[:, -max_cache_length:]
+
+        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[:, -input_ids.shape[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

hybrid_model_lv_mixed_moe/special_tokens_map.json

@@ -0,0 +1,24 @@
+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": "<s>",
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

hybrid_model_lv_mixed_moe/tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
+size 499723

hybrid_model_lv_mixed_moe/tokenizer_config.json

@@ -0,0 +1,44 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "</s>",
+  "legacy": true,
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "<s>",
+  "padding_side": "left",
+  "sp_model_kwargs": {},
+  "spaces_between_special_tokens": false,
+  "tokenizer_class": "LlamaTokenizer",
+  "trust_remote_code": false,
+  "unk_token": "<unk>",
+  "use_default_system_prompt": true,
+  "use_fast": true
+}