Add files using upload-large-folder tool

config.json

@@ -0,0 +1,213 @@
+{
+  "architectures": [
+    "MiMoV2FlashForCausalLM"
+  ],
+  "auto_map": {
+    "AutoConfig": "configuration_mimo_v2_flash.MiMoV2FlashConfig",
+    "AutoModel": "modeling_mimo_v2_flash.MiMoV2FlashModel",
+    "AutoModelForCausalLM": "modeling_mimo_v2_flash.MiMoV2FlashForCausalLM"
+  },
+  "quantization_config": {
+    "activation_scheme": "dynamic",
+    "fmt": "e4m3",
+    "packed_modules_mapping": {},
+    "quant_method": "fp8",
+    "ignored_layers": [
+      "model.layers.0.self_attn.o_proj",
+      "model.layers.1.self_attn.o_proj",
+      "model.layers.2.self_attn.o_proj",
+      "model.layers.3.self_attn.o_proj",
+      "model.layers.4.self_attn.o_proj",
+      "model.layers.5.self_attn.o_proj",
+      "model.layers.6.self_attn.o_proj",
+      "model.layers.7.self_attn.o_proj",
+      "model.layers.8.self_attn.o_proj",
+      "model.layers.9.self_attn.o_proj",
+      "model.layers.10.self_attn.o_proj",
+      "model.layers.11.self_attn.o_proj",
+      "model.layers.12.self_attn.o_proj",
+      "model.layers.13.self_attn.o_proj",
+      "model.layers.14.self_attn.o_proj",
+      "model.layers.15.self_attn.o_proj",
+      "model.layers.16.self_attn.o_proj",
+      "model.layers.17.self_attn.o_proj",
+      "model.layers.18.self_attn.o_proj",
+      "model.layers.19.self_attn.o_proj",
+      "model.layers.20.self_attn.o_proj",
+      "model.layers.21.self_attn.o_proj",
+      "model.layers.22.self_attn.o_proj",
+      "model.layers.23.self_attn.o_proj",
+      "model.layers.24.self_attn.o_proj",
+      "model.layers.25.self_attn.o_proj",
+      "model.layers.26.self_attn.o_proj",
+      "model.layers.27.self_attn.o_proj",
+      "model.layers.28.self_attn.o_proj",
+      "model.layers.29.self_attn.o_proj",
+      "model.layers.30.self_attn.o_proj",
+      "model.layers.31.self_attn.o_proj",
+      "model.layers.32.self_attn.o_proj",
+      "model.layers.33.self_attn.o_proj",
+      "model.layers.34.self_attn.o_proj",
+      "model.layers.35.self_attn.o_proj",
+      "model.layers.36.self_attn.o_proj",
+      "model.layers.37.self_attn.o_proj",
+      "model.layers.38.self_attn.o_proj",
+      "model.layers.39.self_attn.o_proj",
+      "model.layers.40.self_attn.o_proj",
+      "model.layers.41.self_attn.o_proj",
+      "model.layers.42.self_attn.o_proj",
+      "model.layers.43.self_attn.o_proj",
+      "model.layers.44.self_attn.o_proj",
+      "model.layers.45.self_attn.o_proj",
+      "model.layers.46.self_attn.o_proj",
+      "model.layers.47.self_attn.o_proj",
+      "model.decoder.self_attn.o_proj"
+    ],
+    "weight_block_size": [
+      128,
+      128
+    ]
+  },
+  "attention_dropout": 0.0,
+  "attention_value_scale": 0.707,
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "intermediate_size": 16384,
+  "max_position_embeddings": 262144,
+  "model_type": "mimo_v2_flash",
+  "num_attention_heads": 64,
+  "head_dim": 192,
+  "num_hidden_layers": 48,
+  "num_key_value_heads": 4,
+  "layernorm_epsilon": 1e-05,
+  "rope_theta": 5000000,
+  "tie_word_embeddings": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.40.1",
+  "use_cache": true,
+  "vocab_size": 152576,
+  "partial_rotary_factor": 0.334,
+  "sliding_window": 128,
+  "swa_rope_theta": 10000,
+  "attention_bias": false,
+  "v_head_dim": 128,
+  "hybrid_layer_pattern": [
+    0,
+    1,
+    1,
+    1,
+    1,
+    0,
+    1,
+    1,
+    1,
+    1,
+    1,
+    0,
+    1,
+    1,
+    1,
+    1,
+    1,
+    0,
+    1,
+    1,
+    1,
+    1,
+    1,
+    0,
+    1,
+    1,
+    1,
+    1,
+    1,
+    0,
+    1,
+    1,
+    1,
+    1,
+    1,
+    0,
+    1,
+    1,
+    1,
+    1,
+    1,
+    0,
+    1,
+    1,
+    1,
+    1,
+    1,
+    0
+  ],
+  "add_swa_attention_sink_bias": true,
+  "add_full_attention_sink_bias": false,
+  "sliding_window_size": 128,
+  "attention_chunk_size": 128,
+  "moe_layer_freq": [
+    0,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1
+  ],
+  "moe_intermediate_size": 2048,
+  "n_routed_experts": 256,
+  "n_shared_experts": null,
+  "num_experts_per_tok": 8,
+  "norm_topk_prob": true,
+  "scoring_func": "sigmoid",
+  "n_group": 1,
+  "topk_group": 1,
+  "topk_method": "noaux_tc",
+  "routed_scaling_factor": null,
+  "swa_num_attention_heads": 64,
+  "swa_num_key_value_heads": 8,
+  "swa_head_dim": 192,
+  "swa_v_head_dim": 128
+}

configuration_mimo_v2_flash.py

@@ -0,0 +1,109 @@
+# coding=utf-8
+#
+# Copyright 2025 Xiaomi Corporation.
+# Copyright 2025 The HuggingFace Inc. team.
+#
+# 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.
+
+from transformers.configuration_utils import PretrainedConfig
+from transformers.modeling_rope_utils import rope_config_validation
+from transformers.utils import logging
+
+
+logger = logging.get_logger(__name__)
+
+
+class MiMoV2FlashConfig(PretrainedConfig):
+
+    model_type = ""
+    keys_to_ignore_at_inference = ["past_key_values"]
+
+    # Default tensor parallel plan for base model `Hybrid`
+    base_model_tp_plan = {
+        "layers.*.self_attn.q_proj": "colwise",
+        "layers.*.self_attn.k_proj": "colwise",
+        "layers.*.self_attn.v_proj": "colwise",
+        "layers.*.self_attn.o_proj": "rowwise",
+        "layers.*.mlp.gate_proj": "colwise",
+        "layers.*.mlp.up_proj": "colwise",
+        "layers.*.mlp.down_proj": "rowwise",
+    }
+    base_model_pp_plan = {
+        "embed_tokens": (["input_ids"], ["inputs_embeds"]),
+        "layers": (["hidden_states", "attention_mask"], ["hidden_states"]),
+        "norm": (["hidden_states"], ["hidden_states"]),
+    }
+
+    attribute_map = {
+        "num_local_experts": "n_routed_experts",
+    }
+
+    def __init__(
+        self,
+        vocab_size=151936,
+        hidden_size=4096,
+        intermediate_size=22016,
+        num_hidden_layers=32,
+        num_attention_heads=32,
+        num_key_value_heads=32,
+        hidden_act="silu",
+        max_position_embeddings=32768,
+        initializer_range=0.02,
+        layernorm_epsilon=1e-6,
+        use_cache=True,
+        tie_word_embeddings=False,
+        rope_theta=10000.0,
+        rope_scaling=None,
+        attention_dropout=0.0,
+        hybrid_block_size=None,
+        hybrid_layer_pattern=None,
+        partial_rotary_factor=1.0,
+        **kwargs,
+    ):
+        self.vocab_size = vocab_size
+        self.max_position_embeddings = max_position_embeddings
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+
+        # for backward compatibility
+        if num_key_value_heads is None:
+            num_key_value_heads = num_attention_heads
+
+        self.num_key_value_heads = num_key_value_heads
+        self.hidden_act = hidden_act
+        self.initializer_range = initializer_range
+        self.layernorm_epsilon = layernorm_epsilon
+        self.use_cache = use_cache
+        self.rope_theta = rope_theta
+        self.rope_scaling = rope_scaling
+        self.attention_dropout = attention_dropout
+
+        if hybrid_block_size is not None and hybrid_layer_pattern is None:
+            hybrid_layer_pattern = [0 if ((i + 1) % hybrid_block_size == 0) else 1 for i in range(num_hidden_layers)]
+        self.hybrid_block_size = hybrid_block_size
+        self.hybrid_layer_pattern = hybrid_layer_pattern
+
+        self.partial_rotary_factor = partial_rotary_factor
+
+        # Validate the correctness of rotary position embeddings parameters
+        # BC: if there is a 'type' field, move it to 'rope_type'.
+        if self.rope_scaling is not None and "type" in self.rope_scaling:
+            self.rope_scaling["rope_type"] = self.rope_scaling["type"]
+        rope_config_validation(self)
+
+        super().__init__(
+            tie_word_embeddings=tie_word_embeddings,
+            **kwargs,
+        )

generation_config.json

@@ -0,0 +1,7 @@
+{
+  "bos_token_id": 151643,
+  "do_sample": false,
+  "eos_token_id": 151643,
+  "max_new_tokens": 2048,
+  "transformers_version": "4.37.0"
+}

model_10_linear_fc2.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c86c4c2b77ffbae6bef2fc99864c5e3b295eac488c4464424b73a203475d8d8e
+size 2148072376

model_11_linear_fc1-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ee8ca7cb9f2a6f7ffdfba1ae49cf62ed1e6ba06301ce9bf9db2165e6184384e
+size 2500489784

model_11_linear_fc1-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f94dbf87c03eba4e348313d8fb0111d306ecfe592e827b81a375d457a49a29e0
+size 1795653952

model_12_linear_fc1-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5748a678e39f314baf1ab6dda46eaad76fb732f4fd30f95a947e5ca11b59e93b
+size 2500489784

model_22_linear_fc1-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2098e5b2387711b3448a7cad0f0eab6de2a6a448f09e4cf78b3d06f0e02a9e36
+size 2500489784

model_26_linear_fc1-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10f39076f966fd82459df03508fd0dccb1e00b11174943f2920f12cb9a5ad968
+size 1795653952

model_29_linear_fc2.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14c00e7dc128fa80c070c33ba04f54f38da51694f14ffe1debd55a9e6ea46813
+size 2148072376

model_2_linear_fc1-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a2bd21d87beedc0ffe07aa0d1e5c68c24cb8b157bc4e9da9431c054cae5c3ae0
+size 2500489184

model_36_linear_fc1-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3eccb80354a473eccd455f02f6dc407c3c5945f185c96bea860a3b35ad322ba0
+size 1795653952

model_38_linear_fc2.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e6a48be8d516030e63a0bfb5d34f848e7c6bd3db9b559f81ae0f129e67da0c26
+size 2148072376

model_3_linear_fc1-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f3ac9175ba9cdc60bf7e1dd0784b5370ea4cf0f421f70ed6daed70a55f335998
+size 2500489184

model_3_linear_fc1-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:87763c74c1cad9e2a7bffbd73814b3410b32d0008910ff26626c545fffa46960
+size 1795653520

model_40_linear_fc1-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5a6d5c7a278b685239de081e35c45ffe74db6dc70533919a6b4b3e040f42b7a7
+size 2500489784

model_40_linear_fc1-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:342f36afd9afdedcd5c1d52449d4dab6717ad4cf378d4611f87351d93f0a3110
+size 1795653952

model_41.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:11f9f838453744e4dbbe94c587abf0cde483f987b4b5e886037ab5102e8c3be7
+size 126910184

model_42_linear_fc1-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a4bfa9e7d36f996cc86d16513d1402b334815a68a14950ae7b7d74635d93aced
+size 1795653952

model_42_linear_fc2.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3eb2b03839fad606cab77af35204bcd92e3f346f4207dc44d04afad4ed6bf65b
+size 2148072376

model_43_linear_fc1-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d2fa36c60f050d296984b6696e0877d9054019c2062065115ce3845af0189f7f
+size 2500489784

model_44_linear_fc1-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:931cfb406ae5959f38e4ac48fde9981f4807c64a888ab0432ec715b1bbe6faec
+size 2500489784

model_44_linear_fc1-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a5e84fb450991b3a80214a80f8d8178cb541bb945bbffc33cb4a2363d72a52c8
+size 1795653952

model_45_linear_fc2.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1805e4735b2a3d6c784d3b8d84ba2dcfe9e8d69bf2a9f1e9457466a3fe49f69
+size 2148072376

model_46.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b21b85b45a46bc5794884b83e849e04641b65374d24005e99e329d92309e71a
+size 132154328

model_47.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c87154c960a93aa9671df4f43c5c29805c1675aed77b9b58624942a5ae7a5e70
+size 126910184

model_4_linear_fc1-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:47f55b72a29f9b6540073f7abe5f62924e587f9f28beb789abbac0ca994cb91a
+size 2500489184

model_4_linear_fc1-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:27dfa25ee9ea11459615e4b2e868e688fd8586bf860aff6608e21d0d4ad6ee91
+size 1795653520

model_5.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3daf97ed1ce75c0791133080f303be6eba57440ad05b912b3495dc6bd3f7941c
+size 126910168

model_6.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:656fda23fd3175c1837dec5dcc1439dcc04f5b26be5f90f470ebe90cd02b01d0
+size 132154312

model_6_linear_fc1-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8e9473226ad20a7fc3ebac8c56115dc92feeb3bcc271550733bd88addb132621
+size 2500489184

model_6_linear_fc1-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eecb9becd85b8996ccf5e9e683faea1a49b3d2337aabc47442a015ce038fd9cf
+size 1795653520

model_6_linear_fc2.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d1a47f1f6789763449e7aa906e385624494258e33ddcfd72d7e1f68ec3ea49c0
+size 2148071864

model_7.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:216079e8c45ece3d2470e08e3bb12eb8ef417d996f5cae6be9686ef6550ba15e
+size 132154312

model_7_linear_fc1-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:621a7bc26963664a6c15adfff7c194e3a04855f3b9dd6baaffa92e7875a63c7b
+size 1795653520

model_7_linear_fc2.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9adcd21128b0f5dd1cf520b3852e99782e1d369174b6a2df871f2ae8fdfde366
+size 2148071864

model_8.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6881f17d7ae28a576485d5e6ec3cf8aff63fe643aa05774a2c269f43ce32fc04
+size 132154312

model_8_linear_fc1-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:de18140dc3eafc13b4edb9db81e325630b89501c55928b365a64911ef0501dde
+size 1795653520

model_8_linear_fc2.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a9ac0504e80a5a44cd470ce07a6a59a0f4483340ad047072fa26543d4e6e1ed
+size 2148071864

model_9.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:964cad799947fbb467a4b35f90f3a155c1aaac69be4f0f8ea1deebd80826ea07
+size 132154312

model_9_linear_fc1-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d757e75acea1ad74d120917ada05052b35239f26ce3323dac2347c0759f15ef
+size 1795653520

model_9_linear_fc2.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e9cc7caf5643a98d68a44ddc35506d8526651a4ed866d97d65d0fd07430efbf
+size 2148071864

model_embedding.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:02a1bdc6589dff3bbe5ef9e2a77af02739c105c1e4d94ccc850b3e115b6e9e54
+size 1249902704

model_final.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1801cb96017ff809b18b896b8b2625336d9bf0f5f2937f200b1db4fa0626a7cd
+size 1249910976

modeling_mimo_v2_flash.py

@@ -0,0 +1,664 @@
+# coding=utf-8
+#
+# Copyright 2025 Xiaomi Corporation.
+# Copyright 2025 The HuggingFace Inc. team.
+#
+# 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.
+
+from typing import Callable, Optional, Tuple, Union
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+from transformers.generation import GenerationMixin
+from transformers.activations import ACT2FN
+from transformers.cache_utils import Cache, DynamicCache
+from transformers.integrations import use_kernel_forward_from_hub
+
+from transformers.modeling_outputs import (
+    BaseModelOutputWithPast,
+    CausalLMOutputWithPast,
+)
+
+from transformers.masking_utils import create_causal_mask, create_sliding_window_causal_mask
+from transformers.modeling_rope_utils import ROPE_INIT_FUNCTIONS, dynamic_rope_update
+from transformers.modeling_utils import ALL_ATTENTION_FUNCTIONS, PreTrainedModel
+from transformers.processing_utils import Unpack
+from transformers.utils import (
+    logging,
+)
+
+from transformers.modeling_outputs import MoeModelOutputWithPast
+from transformers.utils import TransformersKwargs, auto_docstring, can_return_tuple
+from .configuration_mimo_v2_flash import MiMoV2FlashConfig
+
+logger = logging.get_logger(__name__)
+
+
+def rotate_half(x):
+    """Rotates half the hidden dims of the input."""
+    x1 = x[..., : x.shape[-1] // 2]
+    x2 = x[..., x.shape[-1] // 2:]
+    return torch.cat((-x2, x1), dim=-1)
+
+
+def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, 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`, *optional*):
+            Deprecated and unused.
+        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.unsqueeze(unsqueeze_dim)
+    sin = sin.unsqueeze(unsqueeze_dim)
+    q_embed = (q * cos) + (rotate_half(q) * sin)
+    k_embed = (k * cos) + (rotate_half(k) * sin)
+    return q_embed, k_embed
+
+
+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)
+
+
+def eager_attention_forward(
+    module: nn.Module,
+    query: torch.Tensor,
+    key: torch.Tensor,
+    value: torch.Tensor,
+    attention_mask: Optional[torch.Tensor],
+    scaling: float,
+    dropout: float = 0.0,
+    sinks: Optional[torch.Tensor] = None,
+):
+    key_states = repeat_kv(key, module.num_key_value_groups)
+    value_states = repeat_kv(value, module.num_key_value_groups)
+    attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling
+    if attention_mask is not None:
+        causal_mask = attention_mask[:, :, :, : key_states.shape[-2]]
+        attn_weights = attn_weights + causal_mask
+
+    if sinks is not None:
+        sinks = module.attention_sink_bias.reshape(1, -1, 1, 1).expand(query.shape[0], -1, query.shape[-2], -1)
+        attn_weights = torch.cat([attn_weights, sinks], dim=-1)
+
+    attn_weights = attn_weights - attn_weights.max(dim=-1, keepdim=True).values
+    probs = F.softmax(attn_weights, dim=-1, dtype=attn_weights.dtype)
+
+    if sinks is not None:
+        probs = probs[..., :-1]  # we drop the sink here
+
+    attn_weights = nn.functional.dropout(probs, p=dropout, training=module.training)
+    attn_output = torch.matmul(attn_weights, value_states)
+    attn_output = attn_output.transpose(1, 2).contiguous()
+    return attn_output, attn_weights
+
+
+@use_kernel_forward_from_hub("RMSNorm")
+class MiMoV2RMSNorm(nn.Module):
+    def __init__(self, hidden_size, eps=1e-6):
+        """
+            MiMoV2RMSNorm 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)
+
+
+class MiMoV2MLP(nn.Module):
+    """MiMoV2MLP matching the gate, up, and down projection layers."""
+
+    def __init__(self, config: MiMoV2FlashConfig, intermediate_size=None):
+        super().__init__()
+        self.config = config
+        self.hidden_size = config.hidden_size
+        self.intermediate_size = config.intermediate_size if intermediate_size is None else 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, hidden_states):
+        down_proj = self.down_proj(self.act_fn(self.gate_proj(hidden_states)) * self.up_proj(hidden_states))
+        return down_proj
+
+
+class MiMoV2MoEGate(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.config = config
+        self.top_k = config.num_experts_per_tok
+        self.n_routed_experts = config.n_routed_experts
+        self.routed_scaling_factor = (
+            config.routed_scaling_factor
+            if config.routed_scaling_factor is not None
+            else 1.0
+        )
+        self.scoring_func = config.scoring_func
+        self.topk_method = config.topk_method
+        self.n_group = config.n_group
+        self.topk_group = config.topk_group
+
+        # topk selection algorithm
+        self.norm_topk_prob = config.norm_topk_prob
+        self.gating_dim = config.hidden_size
+        self.weight = nn.Parameter(
+            torch.empty((self.n_routed_experts, self.gating_dim))
+        )
+        if self.topk_method == "noaux_tc":
+            self.e_score_correction_bias = nn.Parameter(
+                torch.empty((self.n_routed_experts))
+            )
+
+    def forward(self, hidden_states):
+        bsz, seq_len, h = hidden_states.shape
+        ### compute gating score
+        hidden_states = hidden_states.view(-1, h)
+        logits = F.linear(
+            hidden_states.type(torch.float32), self.weight.type(torch.float32), None
+        )
+        if self.scoring_func == "sigmoid":
+            scores = logits.sigmoid()
+        else:
+            raise NotImplementedError(
+                f"insupportable scoring function for MoE gating: {self.scoring_func}"
+            )
+
+        ### select top-k experts
+        if self.topk_method == "noaux_tc":
+            assert not self.training
+            scores_for_choice = scores.view(bsz * seq_len, -1) + self.e_score_correction_bias.unsqueeze(0)
+            group_scores = (
+                scores_for_choice.view(bsz * seq_len, self.n_group, -1).topk(2, dim=-1)[0].sum(dim = -1)
+            )  # [n, n_group]
+            group_idx = torch.topk(
+                group_scores, k=self.topk_group, dim=-1, sorted=False
+            )[
+                1
+            ]  # [n, top_k_group]
+            group_mask = torch.zeros_like(group_scores)  # [n, n_group]
+            group_mask.scatter_(1, group_idx, 1)  # [n, n_group]
+            score_mask = (
+                group_mask.unsqueeze(-1)
+                .expand(
+                    bsz * seq_len, self.n_group, self.n_routed_experts // self.n_group
+                )
+                .reshape(bsz * seq_len, -1)
+            )  # [n, e]
+            tmp_scores = scores_for_choice.masked_fill(~score_mask.bool(), float("-inf"))  # [n, e]
+            _, topk_idx = torch.topk(
+                tmp_scores, k=self.top_k, dim=-1, sorted=False
+            )
+            topk_weight = scores.gather(1, topk_idx)
+        else:
+            raise NotImplementedError(
+                f"insupportable TopK function for MoE gating: {self.topk_method}"
+            )
+
+        ### norm gate to sum 1
+        if self.top_k > 1 and self.norm_topk_prob:
+            denominator = topk_weight.sum(dim=-1, keepdim=True) + 1e-20
+            topk_weight = topk_weight / denominator
+        topk_weight = topk_weight * self.routed_scaling_factor # must multiply the scaling factor
+
+        return topk_idx, topk_weight
+
+
+class MiMoV2MoE(nn.Module):
+    """
+    A mixed expert module containing shared experts.
+    """
+
+    def __init__(self, config):
+        super().__init__()
+        self.config = config
+        self.experts = nn.ModuleList(
+            [
+                MiMoV2MLP(config, intermediate_size=config.moe_intermediate_size)
+                for _ in range(config.n_routed_experts)
+            ]
+        )
+        self.gate = MiMoV2MoEGate(config)
+
+    def moe(self, hidden_states: torch.Tensor, topk_indices: torch.Tensor, topk_weights: torch.Tensor):
+        r"""
+        CALL FOR CONTRIBUTION! I don't have time to optimise this right now, but expert weights need to be fused
+        to not have to do a loop here (deepseek has 256 experts soooo yeah).
+        """
+        final_hidden_states = torch.zeros_like(hidden_states, dtype=topk_weights.dtype)
+        expert_mask = torch.nn.functional.one_hot(topk_indices, num_classes=len(self.experts))
+        expert_mask = expert_mask.permute(2, 0, 1)
+
+        for expert_idx in range(len(self.experts)):
+            expert = self.experts[expert_idx]
+            mask = expert_mask[expert_idx]
+            token_indices, weight_indices = torch.where(mask)
+
+            if token_indices.numel() > 0:
+                expert_weights = topk_weights[token_indices, weight_indices]
+                expert_input = hidden_states[token_indices]
+                expert_output = expert(expert_input)
+                weighted_output = expert_output * expert_weights.unsqueeze(-1)
+                final_hidden_states.index_add_(0, token_indices, weighted_output)
+
+        # in original deepseek, the output of the experts are gathered once we leave this module
+        # thus the moe module is itelsf an IsolatedParallel module
+        # and all expert are "local" meaning we shard but we don't gather
+        return final_hidden_states.type(hidden_states.dtype)
+    
+
+    def forward(self, hidden_states: torch.Tensor) -> tuple[torch.Tensor, torch.Tensor]:
+        orig_shape = hidden_states.shape
+        topk_indices, topk_weights = self.gate(hidden_states)
+        hidden_states = hidden_states.view(-1, hidden_states.shape[-1])
+        hidden_states = self.moe(hidden_states, topk_indices, topk_weights).view(*orig_shape)
+
+        return hidden_states
+
+
+class MiMoV2Attention(nn.Module):
+    """MiMoV2 Global Attention (pattern == 0) and Sliding Window Attention (pattern == 1)."""
+
+    def __init__(self, config: MiMoV2FlashConfig, is_swa: bool, layer_idx: int):
+        super().__init__()
+        self.config = config
+        self.layer_idx = layer_idx
+
+        if is_swa:
+            self.head_dim = config.swa_head_dim
+            self.v_head_dim = config.swa_v_head_dim
+            self.num_attention_heads = config.swa_num_attention_heads
+            self.num_key_value_heads = config.swa_num_key_value_heads
+        else:
+            self.head_dim = config.head_dim
+            self.v_head_dim = config.v_head_dim
+            self.num_attention_heads = config.num_attention_heads
+            self.num_key_value_heads = config.num_key_value_heads
+
+        self.rope_dim = int(self.head_dim * config.partial_rotary_factor)
+        self.num_key_value_groups = self.num_attention_heads // self.num_key_value_heads
+        self.attention_bias = config.attention_bias
+        self.attention_dropout: float = config.attention_dropout
+        self.scaling = self.head_dim ** -0.5
+
+        # These dimensions are for the attention layers
+        q_hidden_size = self.num_attention_heads * self.head_dim
+        k_hidden_size = self.num_key_value_heads * self.head_dim
+        v_hidden_size = self.num_key_value_heads * self.v_head_dim
+        o_hidden_size = self.num_attention_heads * self.v_head_dim
+
+        self.q_proj = nn.Linear(config.hidden_size, q_hidden_size, bias=self.attention_bias)
+        self.k_proj = nn.Linear(config.hidden_size, k_hidden_size, bias=self.attention_bias)
+        self.v_proj = nn.Linear(config.hidden_size, v_hidden_size, bias=self.attention_bias)
+        self.o_proj = nn.Linear(o_hidden_size, config.hidden_size, bias=False)
+
+        self.attention_sink_bias = (
+            torch.nn.Parameter(torch.empty(config.num_attention_heads), requires_grad=False)
+            if (config.add_full_attention_sink_bias and not is_swa) or (config.add_swa_attention_sink_bias and is_swa)
+            else None
+        )
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        position_embeddings: tuple[torch.Tensor, torch.Tensor],
+        attention_mask: Optional[torch.Tensor],
+        past_key_values: Optional[Cache] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        **kwargs: Unpack[TransformersKwargs],
+    ) -> tuple[torch.Tensor, torch.Tensor]:
+        input_shape = hidden_states.shape[:-1]
+        qk_hidden_shape = (*input_shape, -1, self.head_dim)
+        v_hidden_shape = (*input_shape, -1, self.v_head_dim)
+
+        query_states = self.q_proj(hidden_states).view(qk_hidden_shape).transpose(1, 2)
+        key_states = self.k_proj(hidden_states).view(qk_hidden_shape).transpose(1, 2)
+        value_states = self.v_proj(hidden_states).view(v_hidden_shape).transpose(1, 2)
+
+        cos, sin = position_embeddings
+
+        query_rope, query_nope = query_states.split([self.rope_dim, self.head_dim - self.rope_dim], dim=-1)
+        key_rope, key_nope = key_states.split([self.rope_dim, self.head_dim - self.rope_dim], dim=-1)
+
+        query_rope, key_rope = apply_rotary_pos_emb(query_rope, key_rope, cos, sin)
+
+        query_states = torch.cat([query_rope, query_nope], dim=-1)
+        key_states = torch.cat([key_rope, key_nope], dim=-1)
+
+        if past_key_values is not None:
+            # sin and cos are specific to RoPE models; cache_position needed for the static cache
+            cache_kwargs = {"sin": sin, "cos": cos, "cache_position": cache_position}
+            key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx, cache_kwargs)
+
+        attention_interface: Callable = eager_attention_forward
+        if self.config._attn_implementation != "eager":
+            attention_interface = ALL_ATTENTION_FUNCTIONS[self.config._attn_implementation]
+
+        attn_output, attn_weights = attention_interface(
+            self,
+            query_states,
+            key_states,
+            value_states,
+            attention_mask,
+            dropout=0.0 if not self.training else self.attention_dropout,
+            scaling=self.scaling,
+            position_ids=position_ids,
+            sinks=self.attention_sink_bias,
+        )
+
+        attn_output = attn_output.reshape(*input_shape, -1).contiguous()
+        attn_output = self.o_proj(attn_output)
+        return attn_output, attn_weights
+
+
+class MiMoV2DecoderLayer(nn.Module):
+    """
+    MiMoV2 Decoder Layer. It dynamically chooses the correct attention
+    module based on the layer index and the `hybrid_layer_pattern`.
+    """
+
+    def __init__(self, config: MiMoV2FlashConfig, layer_idx: int):
+        super().__init__()
+
+        # This is the key logic: choose the module based on the pattern
+        is_swa_layer = config.hybrid_layer_pattern[layer_idx] == 1
+        if is_swa_layer:
+            self.attention_type = "sliding_window_attention"
+            self.self_attn = MiMoV2Attention(config, True, layer_idx)
+        else:
+            self.attention_type = "full_attention"
+            self.self_attn = MiMoV2Attention(config, False, layer_idx)
+
+        self.mlp = (
+            MiMoV2MoE(config)
+            if (
+                    getattr(config, 'n_routed_experts', None) is not None
+                    and config.moe_layer_freq[layer_idx]
+            )
+            else MiMoV2MLP(config)
+        )
+
+        self.input_layernorm = MiMoV2RMSNorm(config.hidden_size, eps=config.layernorm_epsilon)
+        self.post_attention_layernorm = MiMoV2RMSNorm(config.hidden_size, eps=config.layernorm_epsilon)
+        self.hidden_size = config.hidden_size
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Cache] = None,
+        use_cache: Optional[bool] = False,
+        cache_position: Optional[torch.LongTensor] = None,
+        position_embeddings: Optional[tuple[torch.Tensor, torch.Tensor]] = None,
+        **kwargs: Unpack[TransformersKwargs],
+    ) -> torch.Tensor:
+        residual = hidden_states
+        hidden_states = self.input_layernorm(hidden_states)
+        # Self Attention
+        hidden_states, _ = self.self_attn(
+            hidden_states=hidden_states,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            use_cache=use_cache,
+            cache_position=cache_position,
+            position_embeddings=position_embeddings,
+            **kwargs,
+        )
+        hidden_states = residual + hidden_states
+
+        # MLP or MOE
+        residual = hidden_states
+        hidden_states = self.post_attention_layernorm(hidden_states)
+        hidden_states = self.mlp(hidden_states)
+        hidden_states = residual + hidden_states
+        return hidden_states
+
+class MiMoV2FlashRotaryEmbedding(nn.Module):
+    inv_freq: torch.Tensor  # fix linting for `register_buffer`
+
+    def __init__(self, config: MiMoV2FlashConfig, is_swa, device=None):
+        super().__init__()
+        # BC: "rope_type" was originally "type"
+        if hasattr(config, "rope_scaling") and isinstance(config.rope_scaling, dict):
+            self.rope_type = config.rope_scaling.get("rope_type", config.rope_scaling.get("type"))
+        else:
+            self.rope_type = "default"
+        self.max_seq_len_cached = config.max_position_embeddings
+        self.original_max_seq_len = config.max_position_embeddings
+
+        self.config = config
+        self.rope_init_fn = ROPE_INIT_FUNCTIONS[self.rope_type]
+
+        if is_swa:
+            self.config.rope_theta = config.swa_rope_theta
+            self.config.head_dim = config.swa_head_dim
+
+        inv_freq, self.attention_scaling = self.rope_init_fn(self.config, device)
+        self.register_buffer("inv_freq", inv_freq, persistent=False)
+        self.original_inv_freq = self.inv_freq
+
+    @torch.no_grad()
+    @dynamic_rope_update  # power user: used with advanced RoPE types (e.g. dynamic rope)
+    def forward(self, x, position_ids):
+        inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1).to(x.device)
+        position_ids_expanded = position_ids[:, None, :].float()
+
+        device_type = x.device.type if isinstance(x.device.type, str) and x.device.type != "mps" else "cpu"
+        with torch.autocast(device_type=device_type, enabled=False):  # Force float32
+            freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).transpose(1, 2)
+            emb = torch.cat((freqs, freqs), dim=-1)
+            cos = emb.cos() * self.attention_scaling
+            sin = emb.sin() * self.attention_scaling
+
+        return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)
+
+
+@auto_docstring
+class MiMoV2Model(PreTrainedModel):
+    """The main 'model' block, corresponding to `model.` in the weight map."""
+    config_class = MiMoV2FlashConfig
+
+    def __init__(self, config: MiMoV2FlashConfig):
+        super().__init__(config)
+        self.vocab_size = config.vocab_size
+
+        self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size)
+        self.layers = nn.ModuleList(
+            [MiMoV2DecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
+        )
+        self.norm = MiMoV2RMSNorm(config.hidden_size, eps=config.layernorm_epsilon)
+        self.rotary_emb = MiMoV2FlashRotaryEmbedding(config=config, is_swa=False)
+        self.swa_rotary_emb = MiMoV2FlashRotaryEmbedding(config=config, is_swa=True)
+
+        self.has_sliding_layers = any(
+            pattern == 1 for pattern in config.hybrid_layer_pattern
+        )
+
+        # For Huggingface DynamicCache compatibility
+        self.config.layer_types = [
+            "sliding_attention" if config.hybrid_layer_pattern[i] == 1 else "full_attention"
+            for i in range(config.num_hidden_layers)
+        ]
+
+    @auto_docstring
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Cache] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        use_cache: Optional[bool] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        **kwargs: Unpack[TransformersKwargs],
+    ) -> MoeModelOutputWithPast:
+        if (input_ids is None) ^ (inputs_embeds is not None):
+            raise ValueError("You must specify exactly one of input_ids or inputs_embeds")
+
+        if inputs_embeds is None:
+            inputs_embeds = self.embed_tokens(input_ids)
+
+        if use_cache and past_key_values is None:
+            past_key_values = DynamicCache(config=self.config)
+
+        if cache_position is None:
+            past_seen_tokens = past_key_values.get_seq_length() if past_key_values is not None else 0
+            cache_position = torch.arange(
+                past_seen_tokens, past_seen_tokens + inputs_embeds.shape[1], device=inputs_embeds.device
+            )
+
+        if position_ids is None:
+            position_ids = cache_position.unsqueeze(0)
+
+        # It may already have been prepared by e.g. `generate`
+        if not isinstance(causal_mask_mapping := attention_mask, dict):
+            # Prepare mask arguments
+            mask_kwargs = {
+                "config": self.config,
+                "input_embeds": inputs_embeds,
+                "attention_mask": attention_mask,
+                "cache_position": cache_position,
+                "past_key_values": past_key_values,
+                "position_ids": position_ids,
+            }
+            # Create the masks
+            causal_mask_mapping = {
+                "full_attention": create_causal_mask(**mask_kwargs),
+            }
+            # The sliding window alternating layers are not always activated depending on the config
+            if self.has_sliding_layers:
+                causal_mask_mapping["sliding_window_attention"] = create_sliding_window_causal_mask(**mask_kwargs)
+
+        hidden_states = inputs_embeds
+        position_embeddings = self.rotary_emb(hidden_states, position_ids)
+        swa_position_embeddings = self.swa_rotary_emb(hidden_states, position_ids)
+
+        for decoder_layer in self.layers[: self.config.num_hidden_layers]:
+            hidden_states = decoder_layer(
+                hidden_states,
+                attention_mask=causal_mask_mapping[decoder_layer.attention_type],
+                position_embeddings=(
+                    position_embeddings 
+                    if decoder_layer.attention_type == "full_attention" 
+                    else swa_position_embeddings
+                ),
+                position_ids=position_ids,
+                past_key_values=past_key_values,
+                use_cache=use_cache,
+                cache_position=cache_position,
+                **kwargs,
+            )
+
+        hidden_states = self.norm(hidden_states)
+        return BaseModelOutputWithPast(
+            last_hidden_state=hidden_states,
+            past_key_values=past_key_values if use_cache else None,
+        )
+
+
+@auto_docstring
+class MiMoV2FlashForCausalLM(PreTrainedModel,GenerationMixin):
+    _tied_weights_keys = {"lm_head.weight": "model.embed_tokens.weight"}
+    _tp_plan = {"lm_head": "colwise_rep"}
+    _pp_plan = {"lm_head": (["hidden_states"], ["logits"])}
+
+    config_class = MiMoV2FlashConfig
+    _keys_to_ignore_on_load_unexpected = [r"model.layers\.\d+\.self_attn\.rotary_emb\.inv_freq"]
+
+    def __init__(self, config: MiMoV2FlashConfig):
+        super().__init__(config)
+        self.model = MiMoV2Model(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()
+
+    @can_return_tuple
+    @auto_docstring
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Cache] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        logits_to_keep: Union[int, torch.Tensor] = 0,
+        **kwargs: Unpack[TransformersKwargs],
+    ) -> CausalLMOutputWithPast:
+        
+        outputs: BaseModelOutputWithPast = 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,
+            cache_position=cache_position,
+            **kwargs,
+        )
+
+        hidden_states = outputs.last_hidden_state
+        # Only compute necessary logits, and do not upcast them to float if we are not computing the loss
+        slice_indices = slice(-logits_to_keep, None) if isinstance(logits_to_keep, int) else logits_to_keep
+        logits = self.lm_head(hidden_states[:, slice_indices, :])
+
+        loss = None
+        if labels is not None:
+            loss = self.loss_function(logits=logits, labels=labels, vocab_size=self.config.vocab_size, **kwargs)
+
+        return CausalLMOutputWithPast(
+            loss=loss,
+            logits=logits,
+            past_key_values=outputs.past_key_values,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+__all__ = [
+    "MiMoV2FlashForCausalLM"
+]

tokenizer_config.json

@@ -0,0 +1,207 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "chat_template": "{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- messages[0]['content'] }}\n    {%- else %}\n        {{- 'You are a helpful assistant.' }}\n    {%- endif %}\n    {{- \"\\n\\n# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n    {%- for tool in tools %}\n        {{- \"\\n\" }}\n        {{- tool | tojson }}\n    {%- endfor %}\n    {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- '<|im_start|>system\\n' + messages[0]['content'] + '<|im_end|>\\n' }}\n    {%- else %}\n        {{- '<|im_start|>system\\nYou are a helpful assistant.<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- for message in messages %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) or (message.role == \"assistant\" and not message.tool_calls) %}\n        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {{- '<|im_start|>' + message.role }}\n        {%- if message.content %}\n            {{- '\\n' + message.content }}\n        {%- endif %}\n        {%- for tool_call in message.tool_calls %}\n            {%- if tool_call.function is defined %}\n                {%- set tool_call = tool_call.function %}\n            {%- endif %}\n            {{- '\\n<tool_call>\\n{\"name\": \"' }}\n            {{- tool_call.name }}\n            {{- '\", \"arguments\": ' }}\n            {{- tool_call.arguments | tojson }}\n            {{- '}\\n</tool_call>' }}\n        {%- endfor %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- message.content }}\n        {{- '\\n</tool_response>' }}\n        {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n            {{- '<|im_end|>\\n' }}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\\n' }}\n{%- endif %}\n",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "model_max_length": 262144,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}