Add files using upload-large-folder tool

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,40 @@
+---
+license: apache-2.0
+library_name: mlx
+tags:
+- dllm
+- diffusion
+- llm
+- text_generation
+- mlx
+pipeline_tag: text-generation
+base_model: inclusionAI/LLaDA2.0-flash
+---
+
+# mlx-community/LLaDA2.0-flash-8bit
+
+This model [mlx-community/LLaDA2.0-flash-8bit](https://huggingface.co/mlx-community/LLaDA2.0-flash-8bit) was
+converted to MLX format from [inclusionAI/LLaDA2.0-flash](https://huggingface.co/inclusionAI/LLaDA2.0-flash)
+using mlx-lm version **0.28.4**.
+
+## Use with mlx
+
+```bash
+pip install mlx-lm
+```
+
+```python
+from mlx_lm import load, generate
+
+model, tokenizer = load("mlx-community/LLaDA2.0-flash-8bit")
+
+prompt = "hello"
+
+if tokenizer.chat_template is not None:
+    messages = [{"role": "user", "content": prompt}]
+    prompt = tokenizer.apply_chat_template(
+        messages, add_generation_prompt=True
+    )
+
+response = generate(model, tokenizer, prompt=prompt, verbose=True)
+```

chat_template.jinja

@@ -0,0 +1,86 @@
+{% set thinking_option = 'off' %}
+{{- '<role>SYSTEM</role>' }}
+{%- if messages[0].role == 'system' %}
+    {{- messages[0].content + '\n' }}
+{%- endif %}
+{%- if tools %}
+    {{- "# 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>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\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>\n" }}
+{%- endif %}
+{{- 'detailed thinking ' + thinking_option + '<|role_end|>' }}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if message.content is string %}
+        {%- set content = message.content %}
+    {%- else %}
+        {%- set content = '' %}
+    {%- endif %}
+    {%- if message.role == "user" %}
+        {{- '<role>HUMAN</role>' + message.content + '<|role_end|>' }}
+    {%- elif message.role == "system" and not loop.first %}
+        {{- '<role>SYSTEM</role>' + message.content + '<|role_end|>' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if reasoning_content %}
+                {{- '<role>ASSISTANT</role>' + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<role>ASSISTANT</role>' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<role>ASSISTANT</role>' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|role_end|>' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<role>OBSERVATION</role>' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|role_end|>' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<role>ASSISTANT</role>' }}
+{%- endif %}

config.json

@@ -0,0 +1,66 @@
+{
+    "architectures": [
+        "LLaDA2MoeModelLM"
+    ],
+    "attention_dropout": 0.0,
+    "auto_map": {
+        "AutoConfig": "configuration_llada2_moe.LLaDA2MoeConfig",
+        "AutoModel": "modeling_llada2_moe.LLaDA2MoeModel",
+        "AutoModelForCausalLM": "modeling_llada2_moe.LLaDA2MoeModelLM"
+    },
+    "embedding_dropout": 0.0,
+    "first_k_dense_replace": 1,
+    "head_dim": 128,
+    "hidden_act": "silu",
+    "hidden_size": 4096,
+    "initializer_range": 0.02,
+    "intermediate_size": 9216,
+    "max_position_embeddings": 32768,
+    "max_window_layers": 28,
+    "model_type": "llada2_moe",
+    "moe_intermediate_size": 1024,
+    "moe_router_enable_expert_bias": true,
+    "n_group": 8,
+    "norm_head": false,
+    "norm_softmax": false,
+    "norm_topk_prob": true,
+    "num_attention_heads": 32,
+    "num_experts": 256,
+    "num_experts_per_tok": 8,
+    "num_hidden_layers": 32,
+    "num_key_value_heads": 4,
+    "num_shared_experts": 1,
+    "output_dropout": 0.0,
+    "output_router_logits": false,
+    "pad_token_id": 156892,
+    "partial_rotary_factor": 0.5,
+    "quantization": {
+        "group_size": 64,
+        "bits": 8,
+        "mode": "affine"
+    },
+    "quantization_config": {
+        "group_size": 64,
+        "bits": 8,
+        "mode": "affine"
+    },
+    "rms_norm_eps": 1e-06,
+    "rope_scaling": null,
+    "rope_theta": 600000,
+    "rotary_dim": 64,
+    "routed_scaling_factor": 2.5,
+    "router_dtype": "fp32",
+    "score_function": "sigmoid",
+    "sliding_window": 4096,
+    "tie_word_embeddings": false,
+    "topk_group": 4,
+    "torch_dtype": "bfloat16",
+    "transformers_version": "4.51.0",
+    "use_bias": false,
+    "use_cache": false,
+    "use_qkv_bias": false,
+    "use_rmsnorm": true,
+    "use_sliding_window": false,
+    "using_split_qkv_in_self_attention": false,
+    "vocab_size": 157184
+}

configuration_llada2_moe.py

@@ -0,0 +1,90 @@
+"""LLaDA2 MoE model configuration"""
+
+from transformers.configuration_utils import PretrainedConfig
+
+
+class LLaDA2MoeConfig(PretrainedConfig):
+    model_type = "llada2_moe"
+
+    def __init__(
+        self,
+        vocab_size=30592,
+        hidden_size=1024,
+        intermediate_size=None,
+        num_hidden_layers=24,
+        num_attention_heads=16,
+        num_key_value_heads=0,
+        hidden_act="silu",
+        use_qkv_bias=False,  # llada2 only
+        use_qk_norm=True,
+        use_bias=True,  # llada2 only
+        rms_norm_eps=1e-05,
+        norm_head=False,  # llada2 only
+        tie_word_embeddings=False,  # PretrainedConfig key, here change default value.
+        embedding_dropout=0.1,
+        attention_dropout=0.1,
+        output_dropout=0.1,
+        initializer_range=0.02,
+        max_position_embeddings=16384,
+        rope_theta=10000.0,
+        use_cache=True,
+        use_sliding_window=False,
+        sliding_window=4096,
+        max_window_layers=28,
+        rope_scaling=None,
+        pad_token_id=126081,
+        num_experts=16,
+        num_shared_experts=0,
+        num_experts_per_tok=2,
+        n_group=8,
+        topk_group=4,
+        routed_scaling_factor=2.5,
+        moe_intermediate_size=None,
+        first_k_dense_replace=0,
+        head_dim=None,
+        output_router_logits=False,
+        partial_rotary_factor=0.5,
+        **kwargs,
+    ):
+        self.num_hidden_layers = num_hidden_layers
+        self.vocab_size = vocab_size
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_attention_heads = num_attention_heads
+        self.num_key_value_heads = num_key_value_heads
+        self.hidden_act = hidden_act
+        self.use_qkv_bias = use_qkv_bias
+        self.use_qk_norm = use_qk_norm
+        self.use_bias = use_bias
+        self.norm_head = norm_head
+        self.rms_norm_eps = rms_norm_eps
+        self.embedding_dropout = embedding_dropout
+        self.attention_dropout = attention_dropout
+        self.output_dropout = output_dropout
+        self.initializer_range = initializer_range
+        self.max_position_embeddings = max_position_embeddings
+        self.rope_theta = rope_theta
+        self.use_cache = use_cache
+        self.use_sliding_window = use_sliding_window
+        self.sliding_window = sliding_window
+        self.max_window_layers = max_window_layers
+        self.head_dim = head_dim or self.hidden_size // self.num_attention_heads
+        self.rope_scaling = rope_scaling
+
+        # MoE configs
+        self.num_experts = num_experts
+        self.num_shared_experts = num_shared_experts
+        self.num_experts_per_tok = num_experts_per_tok
+        self.n_group = n_group
+        self.topk_group = topk_group
+        self.moe_intermediate_size = moe_intermediate_size
+        self.first_k_dense_replace = first_k_dense_replace
+        self.output_router_logits = output_router_logits
+        self.routed_scaling_factor = routed_scaling_factor
+        self.partial_rotary_factor = partial_rotary_factor
+
+        super().__init__(
+            pad_token_id=pad_token_id, tie_word_embeddings=tie_word_embeddings, **kwargs
+        )
+
+

model-00001-of-00024.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:81d8a716917398bac45aa8a231c4dad500c2af49d584616584fe8fe89c993ff4
+size 4362773449

model-00002-of-00024.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bed2b84832e38a7dfb864e8769901da46ae3fedfc67fa183f5c9638dba228476
+size 4618998727

model-00003-of-00024.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:497c48aa28682bbf2546e0ac6c3b35fb99bd9a18eb4ab6233159eb57339ae2d3
+size 4618998763

model-00004-of-00024.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:337599f371db23ff763851dc7dae1a7d082c43ee3ddaa4ba11fa876a60ab0598
+size 4674593214

model-00005-of-00024.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b39e95b989985cbd5f85315e7db2ba8935c9d708831265f86642e96fa92afe0
+size 4618998697

model-00006-of-00024.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6a42883d72692e8d7f9047086b70144838f24b24a37d182e7bb453f08d9aedeb
+size 4618998767

model-00007-of-00024.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:326206884d71907a18ed11c11652e3cef549774734582659b0ce8a67e3bdcb59
+size 4674593174

model-00008-of-00024.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8659f68c6832b4cead65358e042759c96509ff76a472ab11885511304d5a1170
+size 4618998708

model-00009-of-00024.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f5248d7ce651c71d42a5dab475b7fe4f15d28d9725c7ebebacb068912736bb0
+size 4618998804

model-00010-of-00024.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:105cd444180550dfcb34425001f3176243f37d1fe12ae0e183ef6267760a7768
+size 4674593238

model-00011-of-00024.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fe2d8f6870b6a8dcc8f3e97d26024ce7e4894219f80b6149af922442527ca818
+size 4618998782

model-00012-of-00024.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4a726199657aa7ac937fd01d402b5363230d2894c06f8a43c26dffa1ad6156ca
+size 4618998800

model-00013-of-00024.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a4ea17f80b113f75d82dbcc7f37a2a1c3ded1b9321a3bee02718200a807a10a0
+size 4674593188

model-00014-of-00024.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c82a836335ed68be56d4c37d6262823f9cde587aae26f9c66f32447359e06008
+size 4618998806

model-00015-of-00024.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e430f1d9e74be3928fcc6e8e217affd7108b6b1959eade2b0444ca055aebd674
+size 4618998798

model-00016-of-00024.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:031694608ddceb3923a88e51437d8f105d5e2d5b488f2eb667840e2c08493ea3
+size 4674593206

model-00017-of-00024.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:73c44dca0942f1ae3b4aa332d8628c92f5751348916392a3998b8825fd3b0f3e
+size 4618998810

model-00018-of-00024.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2cf04ae01973af59e731fbaae7d186f5e3e491164dbf434f45f4beb5d8c17123
+size 4618998764

model-00019-of-00024.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:474c0c7ba36d099cc6bf4225a9ab02ad89f0e509a07cc0db810e21e58769bd99
+size 4674593262

model-00020-of-00024.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6183f1506a39104f8f0cd81bac1135cb0b5b7eb24dd33c24d9b513bb0b7566b1
+size 4618998810

model-00021-of-00024.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a15e741c5555a35747a7e0a81111180ff6bb308dd83287ec6044ab1953f47ec5
+size 4618998800

model-00022-of-00024.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:99ff71cd38e2e65f85a65bbbbd814f30f032418f120d60cb37e967f1f5dc9e6c
+size 4674593278

model-00023-of-00024.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3877aa6eb5e38a43e5a41f2c459790c465b873490853c34b1b2cbdab2d04bcf9
+size 4618998810

model-00024-of-00024.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:023737fa6736c2cef9d79f252eddc0622651cc825f2e49e7f8d271c8df0c065e
+size 2981260471

model.safetensors.index.json

@@ -0,0 +1,964 @@
+{
+    "metadata": {
+        "total_size": 109351050752,
+        "total_parameters": 102889705216
+    },
+    "weight_map": {
+        "lm_head.biases": "model-00024-of-00024.safetensors",
+        "lm_head.scales": "model-00024-of-00024.safetensors",
+        "lm_head.weight": "model-00024-of-00024.safetensors",
+        "model.layers.0.attention.dense.biases": "model-00001-of-00024.safetensors",
+        "model.layers.0.attention.dense.scales": "model-00001-of-00024.safetensors",
+        "model.layers.0.attention.dense.weight": "model-00001-of-00024.safetensors",
+        "model.layers.0.attention.key_layernorm.weight": "model-00001-of-00024.safetensors",
+        "model.layers.0.attention.query_key_value.biases": "model-00001-of-00024.safetensors",
+        "model.layers.0.attention.query_key_value.scales": "model-00001-of-00024.safetensors",
+        "model.layers.0.attention.query_key_value.weight": "model-00001-of-00024.safetensors",
+        "model.layers.0.attention.query_layernorm.weight": "model-00001-of-00024.safetensors",
+        "model.layers.0.input_layernorm.weight": "model-00001-of-00024.safetensors",
+        "model.layers.0.mlp.down_proj.biases": "model-00001-of-00024.safetensors",
+        "model.layers.0.mlp.down_proj.scales": "model-00001-of-00024.safetensors",
+        "model.layers.0.mlp.down_proj.weight": "model-00001-of-00024.safetensors",
+        "model.layers.0.mlp.gate_proj.biases": "model-00001-of-00024.safetensors",
+        "model.layers.0.mlp.gate_proj.scales": "model-00001-of-00024.safetensors",
+        "model.layers.0.mlp.gate_proj.weight": "model-00001-of-00024.safetensors",
+        "model.layers.0.mlp.up_proj.biases": "model-00001-of-00024.safetensors",
+        "model.layers.0.mlp.up_proj.scales": "model-00001-of-00024.safetensors",
+        "model.layers.0.mlp.up_proj.weight": "model-00001-of-00024.safetensors",
+        "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00024.safetensors",
+        "model.layers.1.attention.dense.biases": "model-00001-of-00024.safetensors",
+        "model.layers.1.attention.dense.scales": "model-00001-of-00024.safetensors",
+        "model.layers.1.attention.dense.weight": "model-00001-of-00024.safetensors",
+        "model.layers.1.attention.key_layernorm.weight": "model-00001-of-00024.safetensors",
+        "model.layers.1.attention.query_key_value.biases": "model-00001-of-00024.safetensors",
+        "model.layers.1.attention.query_key_value.scales": "model-00001-of-00024.safetensors",
+        "model.layers.1.attention.query_key_value.weight": "model-00001-of-00024.safetensors",
+        "model.layers.1.attention.query_layernorm.weight": "model-00001-of-00024.safetensors",
+        "model.layers.1.input_layernorm.weight": "model-00001-of-00024.safetensors",
+        "model.layers.1.mlp.gate.expert_bias": "model-00001-of-00024.safetensors",
+        "model.layers.1.mlp.gate.weight": "model-00001-of-00024.safetensors",
+        "model.layers.1.mlp.shared_experts.down_proj.biases": "model-00001-of-00024.safetensors",
+        "model.layers.1.mlp.shared_experts.down_proj.scales": "model-00001-of-00024.safetensors",
+        "model.layers.1.mlp.shared_experts.down_proj.weight": "model-00001-of-00024.safetensors",
+        "model.layers.1.mlp.shared_experts.gate_proj.biases": "model-00001-of-00024.safetensors",
+        "model.layers.1.mlp.shared_experts.gate_proj.scales": "model-00001-of-00024.safetensors",
+        "model.layers.1.mlp.shared_experts.gate_proj.weight": "model-00001-of-00024.safetensors",
+        "model.layers.1.mlp.shared_experts.up_proj.biases": "model-00001-of-00024.safetensors",
+        "model.layers.1.mlp.shared_experts.up_proj.scales": "model-00001-of-00024.safetensors",
+        "model.layers.1.mlp.shared_experts.up_proj.weight": "model-00001-of-00024.safetensors",
+        "model.layers.1.mlp.switch_mlp.down_proj.biases": "model-00001-of-00024.safetensors",
+        "model.layers.1.mlp.switch_mlp.down_proj.scales": "model-00001-of-00024.safetensors",
+        "model.layers.1.mlp.switch_mlp.down_proj.weight": "model-00001-of-00024.safetensors",
+        "model.layers.1.mlp.switch_mlp.gate_proj.biases": "model-00001-of-00024.safetensors",
+        "model.layers.1.mlp.switch_mlp.gate_proj.scales": "model-00001-of-00024.safetensors",
+        "model.layers.1.mlp.switch_mlp.gate_proj.weight": "model-00001-of-00024.safetensors",
+        "model.layers.1.mlp.switch_mlp.up_proj.biases": "model-00001-of-00024.safetensors",
+        "model.layers.1.mlp.switch_mlp.up_proj.scales": "model-00001-of-00024.safetensors",
+        "model.layers.1.mlp.switch_mlp.up_proj.weight": "model-00001-of-00024.safetensors",
+        "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00024.safetensors",
+        "model.layers.10.attention.dense.biases": "model-00007-of-00024.safetensors",
+        "model.layers.10.attention.dense.scales": "model-00007-of-00024.safetensors",
+        "model.layers.10.attention.dense.weight": "model-00007-of-00024.safetensors",
+        "model.layers.10.attention.key_layernorm.weight": "model-00007-of-00024.safetensors",
+        "model.layers.10.attention.query_key_value.biases": "model-00007-of-00024.safetensors",
+        "model.layers.10.attention.query_key_value.scales": "model-00007-of-00024.safetensors",
+        "model.layers.10.attention.query_key_value.weight": "model-00007-of-00024.safetensors",
+        "model.layers.10.attention.query_layernorm.weight": "model-00007-of-00024.safetensors",
+        "model.layers.10.input_layernorm.weight": "model-00008-of-00024.safetensors",
+        "model.layers.10.mlp.gate.expert_bias": "model-00008-of-00024.safetensors",
+        "model.layers.10.mlp.gate.weight": "model-00008-of-00024.safetensors",
+        "model.layers.10.mlp.shared_experts.down_proj.biases": "model-00008-of-00024.safetensors",
+        "model.layers.10.mlp.shared_experts.down_proj.scales": "model-00008-of-00024.safetensors",
+        "model.layers.10.mlp.shared_experts.down_proj.weight": "model-00008-of-00024.safetensors",
+        "model.layers.10.mlp.shared_experts.gate_proj.biases": "model-00008-of-00024.safetensors",
+        "model.layers.10.mlp.shared_experts.gate_proj.scales": "model-00008-of-00024.safetensors",
+        "model.layers.10.mlp.shared_experts.gate_proj.weight": "model-00008-of-00024.safetensors",
+        "model.layers.10.mlp.shared_experts.up_proj.biases": "model-00008-of-00024.safetensors",
+        "model.layers.10.mlp.shared_experts.up_proj.scales": "model-00008-of-00024.safetensors",
+        "model.layers.10.mlp.shared_experts.up_proj.weight": "model-00008-of-00024.safetensors",
+        "model.layers.10.mlp.switch_mlp.down_proj.biases": "model-00008-of-00024.safetensors",
+        "model.layers.10.mlp.switch_mlp.down_proj.scales": "model-00008-of-00024.safetensors",
+        "model.layers.10.mlp.switch_mlp.down_proj.weight": "model-00008-of-00024.safetensors",
+        "model.layers.10.mlp.switch_mlp.gate_proj.biases": "model-00008-of-00024.safetensors",
+        "model.layers.10.mlp.switch_mlp.gate_proj.scales": "model-00008-of-00024.safetensors",
+        "model.layers.10.mlp.switch_mlp.gate_proj.weight": "model-00008-of-00024.safetensors",
+        "model.layers.10.mlp.switch_mlp.up_proj.biases": "model-00008-of-00024.safetensors",
+        "model.layers.10.mlp.switch_mlp.up_proj.scales": "model-00008-of-00024.safetensors",
+        "model.layers.10.mlp.switch_mlp.up_proj.weight": "model-00008-of-00024.safetensors",
+        "model.layers.10.post_attention_layernorm.weight": "model-00008-of-00024.safetensors",
+        "model.layers.11.attention.dense.biases": "model-00008-of-00024.safetensors",
+        "model.layers.11.attention.dense.scales": "model-00008-of-00024.safetensors",
+        "model.layers.11.attention.dense.weight": "model-00008-of-00024.safetensors",
+        "model.layers.11.attention.key_layernorm.weight": "model-00008-of-00024.safetensors",
+        "model.layers.11.attention.query_key_value.biases": "model-00008-of-00024.safetensors",
+        "model.layers.11.attention.query_key_value.scales": "model-00008-of-00024.safetensors",
+        "model.layers.11.attention.query_key_value.weight": "model-00008-of-00024.safetensors",
+        "model.layers.11.attention.query_layernorm.weight": "model-00008-of-00024.safetensors",
+        "model.layers.11.input_layernorm.weight": "model-00009-of-00024.safetensors",
+        "model.layers.11.mlp.gate.expert_bias": "model-00009-of-00024.safetensors",
+        "model.layers.11.mlp.gate.weight": "model-00009-of-00024.safetensors",
+        "model.layers.11.mlp.shared_experts.down_proj.biases": "model-00009-of-00024.safetensors",
+        "model.layers.11.mlp.shared_experts.down_proj.scales": "model-00009-of-00024.safetensors",
+        "model.layers.11.mlp.shared_experts.down_proj.weight": "model-00009-of-00024.safetensors",
+        "model.layers.11.mlp.shared_experts.gate_proj.biases": "model-00009-of-00024.safetensors",
+        "model.layers.11.mlp.shared_experts.gate_proj.scales": "model-00009-of-00024.safetensors",
+        "model.layers.11.mlp.shared_experts.gate_proj.weight": "model-00009-of-00024.safetensors",
+        "model.layers.11.mlp.shared_experts.up_proj.biases": "model-00009-of-00024.safetensors",
+        "model.layers.11.mlp.shared_experts.up_proj.scales": "model-00009-of-00024.safetensors",
+        "model.layers.11.mlp.shared_experts.up_proj.weight": "model-00009-of-00024.safetensors",
+        "model.layers.11.mlp.switch_mlp.down_proj.biases": "model-00009-of-00024.safetensors",
+        "model.layers.11.mlp.switch_mlp.down_proj.scales": "model-00009-of-00024.safetensors",
+        "model.layers.11.mlp.switch_mlp.down_proj.weight": "model-00009-of-00024.safetensors",
+        "model.layers.11.mlp.switch_mlp.gate_proj.biases": "model-00008-of-00024.safetensors",
+        "model.layers.11.mlp.switch_mlp.gate_proj.scales": "model-00008-of-00024.safetensors",
+        "model.layers.11.mlp.switch_mlp.gate_proj.weight": "model-00008-of-00024.safetensors",
+        "model.layers.11.mlp.switch_mlp.up_proj.biases": "model-00009-of-00024.safetensors",
+        "model.layers.11.mlp.switch_mlp.up_proj.scales": "model-00009-of-00024.safetensors",
+        "model.layers.11.mlp.switch_mlp.up_proj.weight": "model-00009-of-00024.safetensors",
+        "model.layers.11.post_attention_layernorm.weight": "model-00009-of-00024.safetensors",
+        "model.layers.12.attention.dense.biases": "model-00009-of-00024.safetensors",
+        "model.layers.12.attention.dense.scales": "model-00009-of-00024.safetensors",
+        "model.layers.12.attention.dense.weight": "model-00009-of-00024.safetensors",
+        "model.layers.12.attention.key_layernorm.weight": "model-00009-of-00024.safetensors",
+        "model.layers.12.attention.query_key_value.biases": "model-00009-of-00024.safetensors",
+        "model.layers.12.attention.query_key_value.scales": "model-00009-of-00024.safetensors",
+        "model.layers.12.attention.query_key_value.weight": "model-00009-of-00024.safetensors",
+        "model.layers.12.attention.query_layernorm.weight": "model-00009-of-00024.safetensors",
+        "model.layers.12.input_layernorm.weight": "model-00010-of-00024.safetensors",
+        "model.layers.12.mlp.gate.expert_bias": "model-00010-of-00024.safetensors",
+        "model.layers.12.mlp.gate.weight": "model-00010-of-00024.safetensors",
+        "model.layers.12.mlp.shared_experts.down_proj.biases": "model-00010-of-00024.safetensors",
+        "model.layers.12.mlp.shared_experts.down_proj.scales": "model-00010-of-00024.safetensors",
+        "model.layers.12.mlp.shared_experts.down_proj.weight": "model-00010-of-00024.safetensors",
+        "model.layers.12.mlp.shared_experts.gate_proj.biases": "model-00010-of-00024.safetensors",
+        "model.layers.12.mlp.shared_experts.gate_proj.scales": "model-00010-of-00024.safetensors",
+        "model.layers.12.mlp.shared_experts.gate_proj.weight": "model-00010-of-00024.safetensors",
+        "model.layers.12.mlp.shared_experts.up_proj.biases": "model-00010-of-00024.safetensors",
+        "model.layers.12.mlp.shared_experts.up_proj.scales": "model-00010-of-00024.safetensors",
+        "model.layers.12.mlp.shared_experts.up_proj.weight": "model-00010-of-00024.safetensors",
+        "model.layers.12.mlp.switch_mlp.down_proj.biases": "model-00010-of-00024.safetensors",
+        "model.layers.12.mlp.switch_mlp.down_proj.scales": "model-00010-of-00024.safetensors",
+        "model.layers.12.mlp.switch_mlp.down_proj.weight": "model-00010-of-00024.safetensors",
+        "model.layers.12.mlp.switch_mlp.gate_proj.biases": "model-00009-of-00024.safetensors",
+        "model.layers.12.mlp.switch_mlp.gate_proj.scales": "model-00009-of-00024.safetensors",
+        "model.layers.12.mlp.switch_mlp.gate_proj.weight": "model-00009-of-00024.safetensors",
+        "model.layers.12.mlp.switch_mlp.up_proj.biases": "model-00009-of-00024.safetensors",
+        "model.layers.12.mlp.switch_mlp.up_proj.scales": "model-00009-of-00024.safetensors",
+        "model.layers.12.mlp.switch_mlp.up_proj.weight": "model-00009-of-00024.safetensors",
+        "model.layers.12.post_attention_layernorm.weight": "model-00010-of-00024.safetensors",
+        "model.layers.13.attention.dense.biases": "model-00010-of-00024.safetensors",
+        "model.layers.13.attention.dense.scales": "model-00010-of-00024.safetensors",
+        "model.layers.13.attention.dense.weight": "model-00010-of-00024.safetensors",
+        "model.layers.13.attention.key_layernorm.weight": "model-00010-of-00024.safetensors",
+        "model.layers.13.attention.query_key_value.biases": "model-00010-of-00024.safetensors",
+        "model.layers.13.attention.query_key_value.scales": "model-00010-of-00024.safetensors",
+        "model.layers.13.attention.query_key_value.weight": "model-00010-of-00024.safetensors",
+        "model.layers.13.attention.query_layernorm.weight": "model-00010-of-00024.safetensors",
+        "model.layers.13.input_layernorm.weight": "model-00010-of-00024.safetensors",
+        "model.layers.13.mlp.gate.expert_bias": "model-00010-of-00024.safetensors",
+        "model.layers.13.mlp.gate.weight": "model-00010-of-00024.safetensors",
+        "model.layers.13.mlp.shared_experts.down_proj.biases": "model-00010-of-00024.safetensors",
+        "model.layers.13.mlp.shared_experts.down_proj.scales": "model-00010-of-00024.safetensors",
+        "model.layers.13.mlp.shared_experts.down_proj.weight": "model-00010-of-00024.safetensors",
+        "model.layers.13.mlp.shared_experts.gate_proj.biases": "model-00010-of-00024.safetensors",
+        "model.layers.13.mlp.shared_experts.gate_proj.scales": "model-00010-of-00024.safetensors",
+        "model.layers.13.mlp.shared_experts.gate_proj.weight": "model-00010-of-00024.safetensors",
+        "model.layers.13.mlp.shared_experts.up_proj.biases": "model-00010-of-00024.safetensors",
+        "model.layers.13.mlp.shared_experts.up_proj.scales": "model-00010-of-00024.safetensors",
+        "model.layers.13.mlp.shared_experts.up_proj.weight": "model-00010-of-00024.safetensors",
+        "model.layers.13.mlp.switch_mlp.down_proj.biases": "model-00010-of-00024.safetensors",
+        "model.layers.13.mlp.switch_mlp.down_proj.scales": "model-00010-of-00024.safetensors",
+        "model.layers.13.mlp.switch_mlp.down_proj.weight": "model-00010-of-00024.safetensors",
+        "model.layers.13.mlp.switch_mlp.gate_proj.biases": "model-00010-of-00024.safetensors",
+        "model.layers.13.mlp.switch_mlp.gate_proj.scales": "model-00010-of-00024.safetensors",
+        "model.layers.13.mlp.switch_mlp.gate_proj.weight": "model-00010-of-00024.safetensors",
+        "model.layers.13.mlp.switch_mlp.up_proj.biases": "model-00010-of-00024.safetensors",
+        "model.layers.13.mlp.switch_mlp.up_proj.scales": "model-00010-of-00024.safetensors",
+        "model.layers.13.mlp.switch_mlp.up_proj.weight": "model-00010-of-00024.safetensors",
+        "model.layers.13.post_attention_layernorm.weight": "model-00010-of-00024.safetensors",
+        "model.layers.14.attention.dense.biases": "model-00010-of-00024.safetensors",
+        "model.layers.14.attention.dense.scales": "model-00010-of-00024.safetensors",
+        "model.layers.14.attention.dense.weight": "model-00010-of-00024.safetensors",
+        "model.layers.14.attention.key_layernorm.weight": "model-00010-of-00024.safetensors",
+        "model.layers.14.attention.query_key_value.biases": "model-00010-of-00024.safetensors",
+        "model.layers.14.attention.query_key_value.scales": "model-00010-of-00024.safetensors",
+        "model.layers.14.attention.query_key_value.weight": "model-00010-of-00024.safetensors",
+        "model.layers.14.attention.query_layernorm.weight": "model-00010-of-00024.safetensors",
+        "model.layers.14.input_layernorm.weight": "model-00011-of-00024.safetensors",
+        "model.layers.14.mlp.gate.expert_bias": "model-00011-of-00024.safetensors",
+        "model.layers.14.mlp.gate.weight": "model-00011-of-00024.safetensors",
+        "model.layers.14.mlp.shared_experts.down_proj.biases": "model-00011-of-00024.safetensors",
+        "model.layers.14.mlp.shared_experts.down_proj.scales": "model-00011-of-00024.safetensors",
+        "model.layers.14.mlp.shared_experts.down_proj.weight": "model-00011-of-00024.safetensors",
+        "model.layers.14.mlp.shared_experts.gate_proj.biases": "model-00011-of-00024.safetensors",
+        "model.layers.14.mlp.shared_experts.gate_proj.scales": "model-00011-of-00024.safetensors",
+        "model.layers.14.mlp.shared_experts.gate_proj.weight": "model-00011-of-00024.safetensors",
+        "model.layers.14.mlp.shared_experts.up_proj.biases": "model-00011-of-00024.safetensors",
+        "model.layers.14.mlp.shared_experts.up_proj.scales": "model-00011-of-00024.safetensors",
+        "model.layers.14.mlp.shared_experts.up_proj.weight": "model-00011-of-00024.safetensors",
+        "model.layers.14.mlp.switch_mlp.down_proj.biases": "model-00011-of-00024.safetensors",
+        "model.layers.14.mlp.switch_mlp.down_proj.scales": "model-00011-of-00024.safetensors",
+        "model.layers.14.mlp.switch_mlp.down_proj.weight": "model-00011-of-00024.safetensors",
+        "model.layers.14.mlp.switch_mlp.gate_proj.biases": "model-00011-of-00024.safetensors",
+        "model.layers.14.mlp.switch_mlp.gate_proj.scales": "model-00011-of-00024.safetensors",
+        "model.layers.14.mlp.switch_mlp.gate_proj.weight": "model-00011-of-00024.safetensors",
+        "model.layers.14.mlp.switch_mlp.up_proj.biases": "model-00011-of-00024.safetensors",
+        "model.layers.14.mlp.switch_mlp.up_proj.scales": "model-00011-of-00024.safetensors",
+        "model.layers.14.mlp.switch_mlp.up_proj.weight": "model-00011-of-00024.safetensors",
+        "model.layers.14.post_attention_layernorm.weight": "model-00011-of-00024.safetensors",
+        "model.layers.15.attention.dense.biases": "model-00011-of-00024.safetensors",
+        "model.layers.15.attention.dense.scales": "model-00011-of-00024.safetensors",
+        "model.layers.15.attention.dense.weight": "model-00011-of-00024.safetensors",
+        "model.layers.15.attention.key_layernorm.weight": "model-00011-of-00024.safetensors",
+        "model.layers.15.attention.query_key_value.biases": "model-00011-of-00024.safetensors",
+        "model.layers.15.attention.query_key_value.scales": "model-00011-of-00024.safetensors",
+        "model.layers.15.attention.query_key_value.weight": "model-00011-of-00024.safetensors",
+        "model.layers.15.attention.query_layernorm.weight": "model-00011-of-00024.safetensors",
+        "model.layers.15.input_layernorm.weight": "model-00012-of-00024.safetensors",
+        "model.layers.15.mlp.gate.expert_bias": "model-00012-of-00024.safetensors",
+        "model.layers.15.mlp.gate.weight": "model-00012-of-00024.safetensors",
+        "model.layers.15.mlp.shared_experts.down_proj.biases": "model-00012-of-00024.safetensors",
+        "model.layers.15.mlp.shared_experts.down_proj.scales": "model-00012-of-00024.safetensors",
+        "model.layers.15.mlp.shared_experts.down_proj.weight": "model-00012-of-00024.safetensors",
+        "model.layers.15.mlp.shared_experts.gate_proj.biases": "model-00012-of-00024.safetensors",
+        "model.layers.15.mlp.shared_experts.gate_proj.scales": "model-00012-of-00024.safetensors",
+        "model.layers.15.mlp.shared_experts.gate_proj.weight": "model-00012-of-00024.safetensors",
+        "model.layers.15.mlp.shared_experts.up_proj.biases": "model-00012-of-00024.safetensors",
+        "model.layers.15.mlp.shared_experts.up_proj.scales": "model-00012-of-00024.safetensors",
+        "model.layers.15.mlp.shared_experts.up_proj.weight": "model-00012-of-00024.safetensors",
+        "model.layers.15.mlp.switch_mlp.down_proj.biases": "model-00012-of-00024.safetensors",
+        "model.layers.15.mlp.switch_mlp.down_proj.scales": "model-00012-of-00024.safetensors",
+        "model.layers.15.mlp.switch_mlp.down_proj.weight": "model-00012-of-00024.safetensors",
+        "model.layers.15.mlp.switch_mlp.gate_proj.biases": "model-00011-of-00024.safetensors",
+        "model.layers.15.mlp.switch_mlp.gate_proj.scales": "model-00011-of-00024.safetensors",
+        "model.layers.15.mlp.switch_mlp.gate_proj.weight": "model-00011-of-00024.safetensors",
+        "model.layers.15.mlp.switch_mlp.up_proj.biases": "model-00012-of-00024.safetensors",
+        "model.layers.15.mlp.switch_mlp.up_proj.scales": "model-00012-of-00024.safetensors",
+        "model.layers.15.mlp.switch_mlp.up_proj.weight": "model-00012-of-00024.safetensors",
+        "model.layers.15.post_attention_layernorm.weight": "model-00012-of-00024.safetensors",
+        "model.layers.16.attention.dense.biases": "model-00012-of-00024.safetensors",
+        "model.layers.16.attention.dense.scales": "model-00012-of-00024.safetensors",
+        "model.layers.16.attention.dense.weight": "model-00012-of-00024.safetensors",
+        "model.layers.16.attention.key_layernorm.weight": "model-00012-of-00024.safetensors",
+        "model.layers.16.attention.query_key_value.biases": "model-00012-of-00024.safetensors",
+        "model.layers.16.attention.query_key_value.scales": "model-00012-of-00024.safetensors",
+        "model.layers.16.attention.query_key_value.weight": "model-00012-of-00024.safetensors",
+        "model.layers.16.attention.query_layernorm.weight": "model-00012-of-00024.safetensors",
+        "model.layers.16.input_layernorm.weight": "model-00013-of-00024.safetensors",
+        "model.layers.16.mlp.gate.expert_bias": "model-00013-of-00024.safetensors",
+        "model.layers.16.mlp.gate.weight": "model-00013-of-00024.safetensors",
+        "model.layers.16.mlp.shared_experts.down_proj.biases": "model-00013-of-00024.safetensors",
+        "model.layers.16.mlp.shared_experts.down_proj.scales": "model-00013-of-00024.safetensors",
+        "model.layers.16.mlp.shared_experts.down_proj.weight": "model-00013-of-00024.safetensors",
+        "model.layers.16.mlp.shared_experts.gate_proj.biases": "model-00013-of-00024.safetensors",
+        "model.layers.16.mlp.shared_experts.gate_proj.scales": "model-00013-of-00024.safetensors",
+        "model.layers.16.mlp.shared_experts.gate_proj.weight": "model-00013-of-00024.safetensors",
+        "model.layers.16.mlp.shared_experts.up_proj.biases": "model-00013-of-00024.safetensors",
+        "model.layers.16.mlp.shared_experts.up_proj.scales": "model-00013-of-00024.safetensors",
+        "model.layers.16.mlp.shared_experts.up_proj.weight": "model-00013-of-00024.safetensors",
+        "model.layers.16.mlp.switch_mlp.down_proj.biases": "model-00013-of-00024.safetensors",
+        "model.layers.16.mlp.switch_mlp.down_proj.scales": "model-00013-of-00024.safetensors",
+        "model.layers.16.mlp.switch_mlp.down_proj.weight": "model-00013-of-00024.safetensors",
+        "model.layers.16.mlp.switch_mlp.gate_proj.biases": "model-00012-of-00024.safetensors",
+        "model.layers.16.mlp.switch_mlp.gate_proj.scales": "model-00012-of-00024.safetensors",
+        "model.layers.16.mlp.switch_mlp.gate_proj.weight": "model-00012-of-00024.safetensors",
+        "model.layers.16.mlp.switch_mlp.up_proj.biases": "model-00012-of-00024.safetensors",
+        "model.layers.16.mlp.switch_mlp.up_proj.scales": "model-00012-of-00024.safetensors",
+        "model.layers.16.mlp.switch_mlp.up_proj.weight": "model-00012-of-00024.safetensors",
+        "model.layers.16.post_attention_layernorm.weight": "model-00013-of-00024.safetensors",
+        "model.layers.17.attention.dense.biases": "model-00013-of-00024.safetensors",
+        "model.layers.17.attention.dense.scales": "model-00013-of-00024.safetensors",
+        "model.layers.17.attention.dense.weight": "model-00013-of-00024.safetensors",
+        "model.layers.17.attention.key_layernorm.weight": "model-00013-of-00024.safetensors",
+        "model.layers.17.attention.query_key_value.biases": "model-00013-of-00024.safetensors",
+        "model.layers.17.attention.query_key_value.scales": "model-00013-of-00024.safetensors",
+        "model.layers.17.attention.query_key_value.weight": "model-00013-of-00024.safetensors",
+        "model.layers.17.attention.query_layernorm.weight": "model-00013-of-00024.safetensors",
+        "model.layers.17.input_layernorm.weight": "model-00013-of-00024.safetensors",
+        "model.layers.17.mlp.gate.expert_bias": "model-00013-of-00024.safetensors",
+        "model.layers.17.mlp.gate.weight": "model-00013-of-00024.safetensors",
+        "model.layers.17.mlp.shared_experts.down_proj.biases": "model-00013-of-00024.safetensors",
+        "model.layers.17.mlp.shared_experts.down_proj.scales": "model-00013-of-00024.safetensors",
+        "model.layers.17.mlp.shared_experts.down_proj.weight": "model-00013-of-00024.safetensors",
+        "model.layers.17.mlp.shared_experts.gate_proj.biases": "model-00013-of-00024.safetensors",
+        "model.layers.17.mlp.shared_experts.gate_proj.scales": "model-00013-of-00024.safetensors",
+        "model.layers.17.mlp.shared_experts.gate_proj.weight": "model-00013-of-00024.safetensors",
+        "model.layers.17.mlp.shared_experts.up_proj.biases": "model-00013-of-00024.safetensors",
+        "model.layers.17.mlp.shared_experts.up_proj.scales": "model-00013-of-00024.safetensors",
+        "model.layers.17.mlp.shared_experts.up_proj.weight": "model-00013-of-00024.safetensors",
+        "model.layers.17.mlp.switch_mlp.down_proj.biases": "model-00013-of-00024.safetensors",
+        "model.layers.17.mlp.switch_mlp.down_proj.scales": "model-00013-of-00024.safetensors",
+        "model.layers.17.mlp.switch_mlp.down_proj.weight": "model-00013-of-00024.safetensors",
+        "model.layers.17.mlp.switch_mlp.gate_proj.biases": "model-00013-of-00024.safetensors",
+        "model.layers.17.mlp.switch_mlp.gate_proj.scales": "model-00013-of-00024.safetensors",
+        "model.layers.17.mlp.switch_mlp.gate_proj.weight": "model-00013-of-00024.safetensors",
+        "model.layers.17.mlp.switch_mlp.up_proj.biases": "model-00013-of-00024.safetensors",
+        "model.layers.17.mlp.switch_mlp.up_proj.scales": "model-00013-of-00024.safetensors",
+        "model.layers.17.mlp.switch_mlp.up_proj.weight": "model-00013-of-00024.safetensors",
+        "model.layers.17.post_attention_layernorm.weight": "model-00013-of-00024.safetensors",
+        "model.layers.18.attention.dense.biases": "model-00013-of-00024.safetensors",
+        "model.layers.18.attention.dense.scales": "model-00013-of-00024.safetensors",
+        "model.layers.18.attention.dense.weight": "model-00013-of-00024.safetensors",
+        "model.layers.18.attention.key_layernorm.weight": "model-00013-of-00024.safetensors",
+        "model.layers.18.attention.query_key_value.biases": "model-00013-of-00024.safetensors",
+        "model.layers.18.attention.query_key_value.scales": "model-00013-of-00024.safetensors",
+        "model.layers.18.attention.query_key_value.weight": "model-00013-of-00024.safetensors",
+        "model.layers.18.attention.query_layernorm.weight": "model-00013-of-00024.safetensors",
+        "model.layers.18.input_layernorm.weight": "model-00014-of-00024.safetensors",
+        "model.layers.18.mlp.gate.expert_bias": "model-00014-of-00024.safetensors",
+        "model.layers.18.mlp.gate.weight": "model-00014-of-00024.safetensors",
+        "model.layers.18.mlp.shared_experts.down_proj.biases": "model-00014-of-00024.safetensors",
+        "model.layers.18.mlp.shared_experts.down_proj.scales": "model-00014-of-00024.safetensors",
+        "model.layers.18.mlp.shared_experts.down_proj.weight": "model-00014-of-00024.safetensors",
+        "model.layers.18.mlp.shared_experts.gate_proj.biases": "model-00014-of-00024.safetensors",
+        "model.layers.18.mlp.shared_experts.gate_proj.scales": "model-00014-of-00024.safetensors",
+        "model.layers.18.mlp.shared_experts.gate_proj.weight": "model-00014-of-00024.safetensors",
+        "model.layers.18.mlp.shared_experts.up_proj.biases": "model-00014-of-00024.safetensors",
+        "model.layers.18.mlp.shared_experts.up_proj.scales": "model-00014-of-00024.safetensors",
+        "model.layers.18.mlp.shared_experts.up_proj.weight": "model-00014-of-00024.safetensors",
+        "model.layers.18.mlp.switch_mlp.down_proj.biases": "model-00014-of-00024.safetensors",
+        "model.layers.18.mlp.switch_mlp.down_proj.scales": "model-00014-of-00024.safetensors",
+        "model.layers.18.mlp.switch_mlp.down_proj.weight": "model-00014-of-00024.safetensors",
+        "model.layers.18.mlp.switch_mlp.gate_proj.biases": "model-00014-of-00024.safetensors",
+        "model.layers.18.mlp.switch_mlp.gate_proj.scales": "model-00014-of-00024.safetensors",
+        "model.layers.18.mlp.switch_mlp.gate_proj.weight": "model-00014-of-00024.safetensors",
+        "model.layers.18.mlp.switch_mlp.up_proj.biases": "model-00014-of-00024.safetensors",
+        "model.layers.18.mlp.switch_mlp.up_proj.scales": "model-00014-of-00024.safetensors",
+        "model.layers.18.mlp.switch_mlp.up_proj.weight": "model-00014-of-00024.safetensors",
+        "model.layers.18.post_attention_layernorm.weight": "model-00014-of-00024.safetensors",
+        "model.layers.19.attention.dense.biases": "model-00014-of-00024.safetensors",
+        "model.layers.19.attention.dense.scales": "model-00014-of-00024.safetensors",
+        "model.layers.19.attention.dense.weight": "model-00014-of-00024.safetensors",
+        "model.layers.19.attention.key_layernorm.weight": "model-00014-of-00024.safetensors",
+        "model.layers.19.attention.query_key_value.biases": "model-00014-of-00024.safetensors",
+        "model.layers.19.attention.query_key_value.scales": "model-00014-of-00024.safetensors",
+        "model.layers.19.attention.query_key_value.weight": "model-00014-of-00024.safetensors",
+        "model.layers.19.attention.query_layernorm.weight": "model-00014-of-00024.safetensors",
+        "model.layers.19.input_layernorm.weight": "model-00015-of-00024.safetensors",
+        "model.layers.19.mlp.gate.expert_bias": "model-00015-of-00024.safetensors",
+        "model.layers.19.mlp.gate.weight": "model-00015-of-00024.safetensors",
+        "model.layers.19.mlp.shared_experts.down_proj.biases": "model-00015-of-00024.safetensors",
+        "model.layers.19.mlp.shared_experts.down_proj.scales": "model-00015-of-00024.safetensors",
+        "model.layers.19.mlp.shared_experts.down_proj.weight": "model-00015-of-00024.safetensors",
+        "model.layers.19.mlp.shared_experts.gate_proj.biases": "model-00015-of-00024.safetensors",
+        "model.layers.19.mlp.shared_experts.gate_proj.scales": "model-00015-of-00024.safetensors",
+        "model.layers.19.mlp.shared_experts.gate_proj.weight": "model-00015-of-00024.safetensors",
+        "model.layers.19.mlp.shared_experts.up_proj.biases": "model-00015-of-00024.safetensors",
+        "model.layers.19.mlp.shared_experts.up_proj.scales": "model-00015-of-00024.safetensors",
+        "model.layers.19.mlp.shared_experts.up_proj.weight": "model-00015-of-00024.safetensors",
+        "model.layers.19.mlp.switch_mlp.down_proj.biases": "model-00015-of-00024.safetensors",
+        "model.layers.19.mlp.switch_mlp.down_proj.scales": "model-00015-of-00024.safetensors",
+        "model.layers.19.mlp.switch_mlp.down_proj.weight": "model-00015-of-00024.safetensors",
+        "model.layers.19.mlp.switch_mlp.gate_proj.biases": "model-00014-of-00024.safetensors",
+        "model.layers.19.mlp.switch_mlp.gate_proj.scales": "model-00014-of-00024.safetensors",
+        "model.layers.19.mlp.switch_mlp.gate_proj.weight": "model-00014-of-00024.safetensors",
+        "model.layers.19.mlp.switch_mlp.up_proj.biases": "model-00015-of-00024.safetensors",
+        "model.layers.19.mlp.switch_mlp.up_proj.scales": "model-00015-of-00024.safetensors",
+        "model.layers.19.mlp.switch_mlp.up_proj.weight": "model-00015-of-00024.safetensors",
+        "model.layers.19.post_attention_layernorm.weight": "model-00015-of-00024.safetensors",
+        "model.layers.2.attention.dense.biases": "model-00001-of-00024.safetensors",
+        "model.layers.2.attention.dense.scales": "model-00001-of-00024.safetensors",
+        "model.layers.2.attention.dense.weight": "model-00001-of-00024.safetensors",
+        "model.layers.2.attention.key_layernorm.weight": "model-00001-of-00024.safetensors",
+        "model.layers.2.attention.query_key_value.biases": "model-00001-of-00024.safetensors",
+        "model.layers.2.attention.query_key_value.scales": "model-00001-of-00024.safetensors",
+        "model.layers.2.attention.query_key_value.weight": "model-00001-of-00024.safetensors",
+        "model.layers.2.attention.query_layernorm.weight": "model-00001-of-00024.safetensors",
+        "model.layers.2.input_layernorm.weight": "model-00002-of-00024.safetensors",
+        "model.layers.2.mlp.gate.expert_bias": "model-00002-of-00024.safetensors",
+        "model.layers.2.mlp.gate.weight": "model-00002-of-00024.safetensors",
+        "model.layers.2.mlp.shared_experts.down_proj.biases": "model-00002-of-00024.safetensors",
+        "model.layers.2.mlp.shared_experts.down_proj.scales": "model-00002-of-00024.safetensors",
+        "model.layers.2.mlp.shared_experts.down_proj.weight": "model-00002-of-00024.safetensors",
+        "model.layers.2.mlp.shared_experts.gate_proj.biases": "model-00002-of-00024.safetensors",
+        "model.layers.2.mlp.shared_experts.gate_proj.scales": "model-00002-of-00024.safetensors",
+        "model.layers.2.mlp.shared_experts.gate_proj.weight": "model-00002-of-00024.safetensors",
+        "model.layers.2.mlp.shared_experts.up_proj.biases": "model-00002-of-00024.safetensors",
+        "model.layers.2.mlp.shared_experts.up_proj.scales": "model-00002-of-00024.safetensors",
+        "model.layers.2.mlp.shared_experts.up_proj.weight": "model-00002-of-00024.safetensors",
+        "model.layers.2.mlp.switch_mlp.down_proj.biases": "model-00002-of-00024.safetensors",
+        "model.layers.2.mlp.switch_mlp.down_proj.scales": "model-00002-of-00024.safetensors",
+        "model.layers.2.mlp.switch_mlp.down_proj.weight": "model-00002-of-00024.safetensors",
+        "model.layers.2.mlp.switch_mlp.gate_proj.biases": "model-00002-of-00024.safetensors",
+        "model.layers.2.mlp.switch_mlp.gate_proj.scales": "model-00002-of-00024.safetensors",
+        "model.layers.2.mlp.switch_mlp.gate_proj.weight": "model-00002-of-00024.safetensors",
+        "model.layers.2.mlp.switch_mlp.up_proj.biases": "model-00002-of-00024.safetensors",
+        "model.layers.2.mlp.switch_mlp.up_proj.scales": "model-00002-of-00024.safetensors",
+        "model.layers.2.mlp.switch_mlp.up_proj.weight": "model-00002-of-00024.safetensors",
+        "model.layers.2.post_attention_layernorm.weight": "model-00002-of-00024.safetensors",
+        "model.layers.20.attention.dense.biases": "model-00015-of-00024.safetensors",
+        "model.layers.20.attention.dense.scales": "model-00015-of-00024.safetensors",
+        "model.layers.20.attention.dense.weight": "model-00015-of-00024.safetensors",
+        "model.layers.20.attention.key_layernorm.weight": "model-00015-of-00024.safetensors",
+        "model.layers.20.attention.query_key_value.biases": "model-00015-of-00024.safetensors",
+        "model.layers.20.attention.query_key_value.scales": "model-00015-of-00024.safetensors",
+        "model.layers.20.attention.query_key_value.weight": "model-00015-of-00024.safetensors",
+        "model.layers.20.attention.query_layernorm.weight": "model-00015-of-00024.safetensors",
+        "model.layers.20.input_layernorm.weight": "model-00016-of-00024.safetensors",
+        "model.layers.20.mlp.gate.expert_bias": "model-00016-of-00024.safetensors",
+        "model.layers.20.mlp.gate.weight": "model-00016-of-00024.safetensors",
+        "model.layers.20.mlp.shared_experts.down_proj.biases": "model-00016-of-00024.safetensors",
+        "model.layers.20.mlp.shared_experts.down_proj.scales": "model-00016-of-00024.safetensors",
+        "model.layers.20.mlp.shared_experts.down_proj.weight": "model-00016-of-00024.safetensors",
+        "model.layers.20.mlp.shared_experts.gate_proj.biases": "model-00016-of-00024.safetensors",
+        "model.layers.20.mlp.shared_experts.gate_proj.scales": "model-00016-of-00024.safetensors",
+        "model.layers.20.mlp.shared_experts.gate_proj.weight": "model-00016-of-00024.safetensors",
+        "model.layers.20.mlp.shared_experts.up_proj.biases": "model-00016-of-00024.safetensors",
+        "model.layers.20.mlp.shared_experts.up_proj.scales": "model-00016-of-00024.safetensors",
+        "model.layers.20.mlp.shared_experts.up_proj.weight": "model-00016-of-00024.safetensors",
+        "model.layers.20.mlp.switch_mlp.down_proj.biases": "model-00016-of-00024.safetensors",
+        "model.layers.20.mlp.switch_mlp.down_proj.scales": "model-00016-of-00024.safetensors",
+        "model.layers.20.mlp.switch_mlp.down_proj.weight": "model-00016-of-00024.safetensors",
+        "model.layers.20.mlp.switch_mlp.gate_proj.biases": "model-00015-of-00024.safetensors",
+        "model.layers.20.mlp.switch_mlp.gate_proj.scales": "model-00015-of-00024.safetensors",
+        "model.layers.20.mlp.switch_mlp.gate_proj.weight": "model-00015-of-00024.safetensors",
+        "model.layers.20.mlp.switch_mlp.up_proj.biases": "model-00015-of-00024.safetensors",
+        "model.layers.20.mlp.switch_mlp.up_proj.scales": "model-00015-of-00024.safetensors",
+        "model.layers.20.mlp.switch_mlp.up_proj.weight": "model-00015-of-00024.safetensors",
+        "model.layers.20.post_attention_layernorm.weight": "model-00016-of-00024.safetensors",
+        "model.layers.21.attention.dense.biases": "model-00016-of-00024.safetensors",
+        "model.layers.21.attention.dense.scales": "model-00016-of-00024.safetensors",
+        "model.layers.21.attention.dense.weight": "model-00016-of-00024.safetensors",
+        "model.layers.21.attention.key_layernorm.weight": "model-00016-of-00024.safetensors",
+        "model.layers.21.attention.query_key_value.biases": "model-00016-of-00024.safetensors",
+        "model.layers.21.attention.query_key_value.scales": "model-00016-of-00024.safetensors",
+        "model.layers.21.attention.query_key_value.weight": "model-00016-of-00024.safetensors",
+        "model.layers.21.attention.query_layernorm.weight": "model-00016-of-00024.safetensors",
+        "model.layers.21.input_layernorm.weight": "model-00016-of-00024.safetensors",
+        "model.layers.21.mlp.gate.expert_bias": "model-00016-of-00024.safetensors",
+        "model.layers.21.mlp.gate.weight": "model-00016-of-00024.safetensors",
+        "model.layers.21.mlp.shared_experts.down_proj.biases": "model-00016-of-00024.safetensors",
+        "model.layers.21.mlp.shared_experts.down_proj.scales": "model-00016-of-00024.safetensors",
+        "model.layers.21.mlp.shared_experts.down_proj.weight": "model-00016-of-00024.safetensors",
+        "model.layers.21.mlp.shared_experts.gate_proj.biases": "model-00016-of-00024.safetensors",
+        "model.layers.21.mlp.shared_experts.gate_proj.scales": "model-00016-of-00024.safetensors",
+        "model.layers.21.mlp.shared_experts.gate_proj.weight": "model-00016-of-00024.safetensors",
+        "model.layers.21.mlp.shared_experts.up_proj.biases": "model-00016-of-00024.safetensors",
+        "model.layers.21.mlp.shared_experts.up_proj.scales": "model-00016-of-00024.safetensors",
+        "model.layers.21.mlp.shared_experts.up_proj.weight": "model-00016-of-00024.safetensors",
+        "model.layers.21.mlp.switch_mlp.down_proj.biases": "model-00016-of-00024.safetensors",
+        "model.layers.21.mlp.switch_mlp.down_proj.scales": "model-00016-of-00024.safetensors",
+        "model.layers.21.mlp.switch_mlp.down_proj.weight": "model-00016-of-00024.safetensors",
+        "model.layers.21.mlp.switch_mlp.gate_proj.biases": "model-00016-of-00024.safetensors",
+        "model.layers.21.mlp.switch_mlp.gate_proj.scales": "model-00016-of-00024.safetensors",
+        "model.layers.21.mlp.switch_mlp.gate_proj.weight": "model-00016-of-00024.safetensors",
+        "model.layers.21.mlp.switch_mlp.up_proj.biases": "model-00016-of-00024.safetensors",
+        "model.layers.21.mlp.switch_mlp.up_proj.scales": "model-00016-of-00024.safetensors",
+        "model.layers.21.mlp.switch_mlp.up_proj.weight": "model-00016-of-00024.safetensors",
+        "model.layers.21.post_attention_layernorm.weight": "model-00016-of-00024.safetensors",
+        "model.layers.22.attention.dense.biases": "model-00016-of-00024.safetensors",
+        "model.layers.22.attention.dense.scales": "model-00016-of-00024.safetensors",
+        "model.layers.22.attention.dense.weight": "model-00016-of-00024.safetensors",
+        "model.layers.22.attention.key_layernorm.weight": "model-00016-of-00024.safetensors",
+        "model.layers.22.attention.query_key_value.biases": "model-00016-of-00024.safetensors",
+        "model.layers.22.attention.query_key_value.scales": "model-00016-of-00024.safetensors",
+        "model.layers.22.attention.query_key_value.weight": "model-00016-of-00024.safetensors",
+        "model.layers.22.attention.query_layernorm.weight": "model-00016-of-00024.safetensors",
+        "model.layers.22.input_layernorm.weight": "model-00017-of-00024.safetensors",
+        "model.layers.22.mlp.gate.expert_bias": "model-00017-of-00024.safetensors",
+        "model.layers.22.mlp.gate.weight": "model-00017-of-00024.safetensors",
+        "model.layers.22.mlp.shared_experts.down_proj.biases": "model-00017-of-00024.safetensors",
+        "model.layers.22.mlp.shared_experts.down_proj.scales": "model-00017-of-00024.safetensors",
+        "model.layers.22.mlp.shared_experts.down_proj.weight": "model-00017-of-00024.safetensors",
+        "model.layers.22.mlp.shared_experts.gate_proj.biases": "model-00017-of-00024.safetensors",
+        "model.layers.22.mlp.shared_experts.gate_proj.scales": "model-00017-of-00024.safetensors",
+        "model.layers.22.mlp.shared_experts.gate_proj.weight": "model-00017-of-00024.safetensors",
+        "model.layers.22.mlp.shared_experts.up_proj.biases": "model-00017-of-00024.safetensors",
+        "model.layers.22.mlp.shared_experts.up_proj.scales": "model-00017-of-00024.safetensors",
+        "model.layers.22.mlp.shared_experts.up_proj.weight": "model-00017-of-00024.safetensors",
+        "model.layers.22.mlp.switch_mlp.down_proj.biases": "model-00017-of-00024.safetensors",
+        "model.layers.22.mlp.switch_mlp.down_proj.scales": "model-00017-of-00024.safetensors",
+        "model.layers.22.mlp.switch_mlp.down_proj.weight": "model-00017-of-00024.safetensors",
+        "model.layers.22.mlp.switch_mlp.gate_proj.biases": "model-00017-of-00024.safetensors",
+        "model.layers.22.mlp.switch_mlp.gate_proj.scales": "model-00017-of-00024.safetensors",
+        "model.layers.22.mlp.switch_mlp.gate_proj.weight": "model-00017-of-00024.safetensors",
+        "model.layers.22.mlp.switch_mlp.up_proj.biases": "model-00017-of-00024.safetensors",
+        "model.layers.22.mlp.switch_mlp.up_proj.scales": "model-00017-of-00024.safetensors",
+        "model.layers.22.mlp.switch_mlp.up_proj.weight": "model-00017-of-00024.safetensors",
+        "model.layers.22.post_attention_layernorm.weight": "model-00017-of-00024.safetensors",
+        "model.layers.23.attention.dense.biases": "model-00017-of-00024.safetensors",
+        "model.layers.23.attention.dense.scales": "model-00017-of-00024.safetensors",
+        "model.layers.23.attention.dense.weight": "model-00017-of-00024.safetensors",
+        "model.layers.23.attention.key_layernorm.weight": "model-00017-of-00024.safetensors",
+        "model.layers.23.attention.query_key_value.biases": "model-00017-of-00024.safetensors",
+        "model.layers.23.attention.query_key_value.scales": "model-00017-of-00024.safetensors",
+        "model.layers.23.attention.query_key_value.weight": "model-00017-of-00024.safetensors",
+        "model.layers.23.attention.query_layernorm.weight": "model-00017-of-00024.safetensors",
+        "model.layers.23.input_layernorm.weight": "model-00018-of-00024.safetensors",
+        "model.layers.23.mlp.gate.expert_bias": "model-00018-of-00024.safetensors",
+        "model.layers.23.mlp.gate.weight": "model-00018-of-00024.safetensors",
+        "model.layers.23.mlp.shared_experts.down_proj.biases": "model-00018-of-00024.safetensors",
+        "model.layers.23.mlp.shared_experts.down_proj.scales": "model-00018-of-00024.safetensors",
+        "model.layers.23.mlp.shared_experts.down_proj.weight": "model-00018-of-00024.safetensors",
+        "model.layers.23.mlp.shared_experts.gate_proj.biases": "model-00018-of-00024.safetensors",
+        "model.layers.23.mlp.shared_experts.gate_proj.scales": "model-00018-of-00024.safetensors",
+        "model.layers.23.mlp.shared_experts.gate_proj.weight": "model-00018-of-00024.safetensors",
+        "model.layers.23.mlp.shared_experts.up_proj.biases": "model-00018-of-00024.safetensors",
+        "model.layers.23.mlp.shared_experts.up_proj.scales": "model-00018-of-00024.safetensors",
+        "model.layers.23.mlp.shared_experts.up_proj.weight": "model-00018-of-00024.safetensors",
+        "model.layers.23.mlp.switch_mlp.down_proj.biases": "model-00018-of-00024.safetensors",
+        "model.layers.23.mlp.switch_mlp.down_proj.scales": "model-00018-of-00024.safetensors",
+        "model.layers.23.mlp.switch_mlp.down_proj.weight": "model-00018-of-00024.safetensors",
+        "model.layers.23.mlp.switch_mlp.gate_proj.biases": "model-00017-of-00024.safetensors",
+        "model.layers.23.mlp.switch_mlp.gate_proj.scales": "model-00017-of-00024.safetensors",
+        "model.layers.23.mlp.switch_mlp.gate_proj.weight": "model-00017-of-00024.safetensors",
+        "model.layers.23.mlp.switch_mlp.up_proj.biases": "model-00018-of-00024.safetensors",
+        "model.layers.23.mlp.switch_mlp.up_proj.scales": "model-00018-of-00024.safetensors",
+        "model.layers.23.mlp.switch_mlp.up_proj.weight": "model-00018-of-00024.safetensors",
+        "model.layers.23.post_attention_layernorm.weight": "model-00018-of-00024.safetensors",
+        "model.layers.24.attention.dense.biases": "model-00018-of-00024.safetensors",
+        "model.layers.24.attention.dense.scales": "model-00018-of-00024.safetensors",
+        "model.layers.24.attention.dense.weight": "model-00018-of-00024.safetensors",
+        "model.layers.24.attention.key_layernorm.weight": "model-00018-of-00024.safetensors",
+        "model.layers.24.attention.query_key_value.biases": "model-00018-of-00024.safetensors",
+        "model.layers.24.attention.query_key_value.scales": "model-00018-of-00024.safetensors",
+        "model.layers.24.attention.query_key_value.weight": "model-00018-of-00024.safetensors",
+        "model.layers.24.attention.query_layernorm.weight": "model-00018-of-00024.safetensors",
+        "model.layers.24.input_layernorm.weight": "model-00019-of-00024.safetensors",
+        "model.layers.24.mlp.gate.expert_bias": "model-00019-of-00024.safetensors",
+        "model.layers.24.mlp.gate.weight": "model-00019-of-00024.safetensors",
+        "model.layers.24.mlp.shared_experts.down_proj.biases": "model-00019-of-00024.safetensors",
+        "model.layers.24.mlp.shared_experts.down_proj.scales": "model-00019-of-00024.safetensors",
+        "model.layers.24.mlp.shared_experts.down_proj.weight": "model-00019-of-00024.safetensors",
+        "model.layers.24.mlp.shared_experts.gate_proj.biases": "model-00019-of-00024.safetensors",
+        "model.layers.24.mlp.shared_experts.gate_proj.scales": "model-00019-of-00024.safetensors",
+        "model.layers.24.mlp.shared_experts.gate_proj.weight": "model-00019-of-00024.safetensors",
+        "model.layers.24.mlp.shared_experts.up_proj.biases": "model-00019-of-00024.safetensors",
+        "model.layers.24.mlp.shared_experts.up_proj.scales": "model-00019-of-00024.safetensors",
+        "model.layers.24.mlp.shared_experts.up_proj.weight": "model-00019-of-00024.safetensors",
+        "model.layers.24.mlp.switch_mlp.down_proj.biases": "model-00019-of-00024.safetensors",
+        "model.layers.24.mlp.switch_mlp.down_proj.scales": "model-00019-of-00024.safetensors",
+        "model.layers.24.mlp.switch_mlp.down_proj.weight": "model-00019-of-00024.safetensors",
+        "model.layers.24.mlp.switch_mlp.gate_proj.biases": "model-00018-of-00024.safetensors",
+        "model.layers.24.mlp.switch_mlp.gate_proj.scales": "model-00018-of-00024.safetensors",
+        "model.layers.24.mlp.switch_mlp.gate_proj.weight": "model-00018-of-00024.safetensors",
+        "model.layers.24.mlp.switch_mlp.up_proj.biases": "model-00018-of-00024.safetensors",
+        "model.layers.24.mlp.switch_mlp.up_proj.scales": "model-00018-of-00024.safetensors",
+        "model.layers.24.mlp.switch_mlp.up_proj.weight": "model-00018-of-00024.safetensors",
+        "model.layers.24.post_attention_layernorm.weight": "model-00019-of-00024.safetensors",
+        "model.layers.25.attention.dense.biases": "model-00019-of-00024.safetensors",
+        "model.layers.25.attention.dense.scales": "model-00019-of-00024.safetensors",
+        "model.layers.25.attention.dense.weight": "model-00019-of-00024.safetensors",
+        "model.layers.25.attention.key_layernorm.weight": "model-00019-of-00024.safetensors",
+        "model.layers.25.attention.query_key_value.biases": "model-00019-of-00024.safetensors",
+        "model.layers.25.attention.query_key_value.scales": "model-00019-of-00024.safetensors",
+        "model.layers.25.attention.query_key_value.weight": "model-00019-of-00024.safetensors",
+        "model.layers.25.attention.query_layernorm.weight": "model-00019-of-00024.safetensors",
+        "model.layers.25.input_layernorm.weight": "model-00019-of-00024.safetensors",
+        "model.layers.25.mlp.gate.expert_bias": "model-00019-of-00024.safetensors",
+        "model.layers.25.mlp.gate.weight": "model-00019-of-00024.safetensors",
+        "model.layers.25.mlp.shared_experts.down_proj.biases": "model-00019-of-00024.safetensors",
+        "model.layers.25.mlp.shared_experts.down_proj.scales": "model-00019-of-00024.safetensors",
+        "model.layers.25.mlp.shared_experts.down_proj.weight": "model-00019-of-00024.safetensors",
+        "model.layers.25.mlp.shared_experts.gate_proj.biases": "model-00019-of-00024.safetensors",
+        "model.layers.25.mlp.shared_experts.gate_proj.scales": "model-00019-of-00024.safetensors",
+        "model.layers.25.mlp.shared_experts.gate_proj.weight": "model-00019-of-00024.safetensors",
+        "model.layers.25.mlp.shared_experts.up_proj.biases": "model-00019-of-00024.safetensors",
+        "model.layers.25.mlp.shared_experts.up_proj.scales": "model-00019-of-00024.safetensors",
+        "model.layers.25.mlp.shared_experts.up_proj.weight": "model-00019-of-00024.safetensors",
+        "model.layers.25.mlp.switch_mlp.down_proj.biases": "model-00019-of-00024.safetensors",
+        "model.layers.25.mlp.switch_mlp.down_proj.scales": "model-00019-of-00024.safetensors",
+        "model.layers.25.mlp.switch_mlp.down_proj.weight": "model-00019-of-00024.safetensors",
+        "model.layers.25.mlp.switch_mlp.gate_proj.biases": "model-00019-of-00024.safetensors",
+        "model.layers.25.mlp.switch_mlp.gate_proj.scales": "model-00019-of-00024.safetensors",
+        "model.layers.25.mlp.switch_mlp.gate_proj.weight": "model-00019-of-00024.safetensors",
+        "model.layers.25.mlp.switch_mlp.up_proj.biases": "model-00019-of-00024.safetensors",
+        "model.layers.25.mlp.switch_mlp.up_proj.scales": "model-00019-of-00024.safetensors",
+        "model.layers.25.mlp.switch_mlp.up_proj.weight": "model-00019-of-00024.safetensors",
+        "model.layers.25.post_attention_layernorm.weight": "model-00019-of-00024.safetensors",
+        "model.layers.26.attention.dense.biases": "model-00019-of-00024.safetensors",
+        "model.layers.26.attention.dense.scales": "model-00019-of-00024.safetensors",
+        "model.layers.26.attention.dense.weight": "model-00019-of-00024.safetensors",
+        "model.layers.26.attention.key_layernorm.weight": "model-00019-of-00024.safetensors",
+        "model.layers.26.attention.query_key_value.biases": "model-00019-of-00024.safetensors",
+        "model.layers.26.attention.query_key_value.scales": "model-00019-of-00024.safetensors",
+        "model.layers.26.attention.query_key_value.weight": "model-00019-of-00024.safetensors",
+        "model.layers.26.attention.query_layernorm.weight": "model-00019-of-00024.safetensors",
+        "model.layers.26.input_layernorm.weight": "model-00020-of-00024.safetensors",
+        "model.layers.26.mlp.gate.expert_bias": "model-00020-of-00024.safetensors",
+        "model.layers.26.mlp.gate.weight": "model-00020-of-00024.safetensors",
+        "model.layers.26.mlp.shared_experts.down_proj.biases": "model-00020-of-00024.safetensors",
+        "model.layers.26.mlp.shared_experts.down_proj.scales": "model-00020-of-00024.safetensors",
+        "model.layers.26.mlp.shared_experts.down_proj.weight": "model-00020-of-00024.safetensors",
+        "model.layers.26.mlp.shared_experts.gate_proj.biases": "model-00020-of-00024.safetensors",
+        "model.layers.26.mlp.shared_experts.gate_proj.scales": "model-00020-of-00024.safetensors",
+        "model.layers.26.mlp.shared_experts.gate_proj.weight": "model-00020-of-00024.safetensors",
+        "model.layers.26.mlp.shared_experts.up_proj.biases": "model-00020-of-00024.safetensors",
+        "model.layers.26.mlp.shared_experts.up_proj.scales": "model-00020-of-00024.safetensors",
+        "model.layers.26.mlp.shared_experts.up_proj.weight": "model-00020-of-00024.safetensors",
+        "model.layers.26.mlp.switch_mlp.down_proj.biases": "model-00020-of-00024.safetensors",
+        "model.layers.26.mlp.switch_mlp.down_proj.scales": "model-00020-of-00024.safetensors",
+        "model.layers.26.mlp.switch_mlp.down_proj.weight": "model-00020-of-00024.safetensors",
+        "model.layers.26.mlp.switch_mlp.gate_proj.biases": "model-00020-of-00024.safetensors",
+        "model.layers.26.mlp.switch_mlp.gate_proj.scales": "model-00020-of-00024.safetensors",
+        "model.layers.26.mlp.switch_mlp.gate_proj.weight": "model-00020-of-00024.safetensors",
+        "model.layers.26.mlp.switch_mlp.up_proj.biases": "model-00020-of-00024.safetensors",
+        "model.layers.26.mlp.switch_mlp.up_proj.scales": "model-00020-of-00024.safetensors",
+        "model.layers.26.mlp.switch_mlp.up_proj.weight": "model-00020-of-00024.safetensors",
+        "model.layers.26.post_attention_layernorm.weight": "model-00020-of-00024.safetensors",
+        "model.layers.27.attention.dense.biases": "model-00020-of-00024.safetensors",
+        "model.layers.27.attention.dense.scales": "model-00020-of-00024.safetensors",
+        "model.layers.27.attention.dense.weight": "model-00020-of-00024.safetensors",
+        "model.layers.27.attention.key_layernorm.weight": "model-00020-of-00024.safetensors",
+        "model.layers.27.attention.query_key_value.biases": "model-00020-of-00024.safetensors",
+        "model.layers.27.attention.query_key_value.scales": "model-00020-of-00024.safetensors",
+        "model.layers.27.attention.query_key_value.weight": "model-00020-of-00024.safetensors",
+        "model.layers.27.attention.query_layernorm.weight": "model-00020-of-00024.safetensors",
+        "model.layers.27.input_layernorm.weight": "model-00021-of-00024.safetensors",
+        "model.layers.27.mlp.gate.expert_bias": "model-00021-of-00024.safetensors",
+        "model.layers.27.mlp.gate.weight": "model-00021-of-00024.safetensors",
+        "model.layers.27.mlp.shared_experts.down_proj.biases": "model-00021-of-00024.safetensors",
+        "model.layers.27.mlp.shared_experts.down_proj.scales": "model-00021-of-00024.safetensors",
+        "model.layers.27.mlp.shared_experts.down_proj.weight": "model-00021-of-00024.safetensors",
+        "model.layers.27.mlp.shared_experts.gate_proj.biases": "model-00021-of-00024.safetensors",
+        "model.layers.27.mlp.shared_experts.gate_proj.scales": "model-00021-of-00024.safetensors",
+        "model.layers.27.mlp.shared_experts.gate_proj.weight": "model-00021-of-00024.safetensors",
+        "model.layers.27.mlp.shared_experts.up_proj.biases": "model-00021-of-00024.safetensors",
+        "model.layers.27.mlp.shared_experts.up_proj.scales": "model-00021-of-00024.safetensors",
+        "model.layers.27.mlp.shared_experts.up_proj.weight": "model-00021-of-00024.safetensors",
+        "model.layers.27.mlp.switch_mlp.down_proj.biases": "model-00021-of-00024.safetensors",
+        "model.layers.27.mlp.switch_mlp.down_proj.scales": "model-00021-of-00024.safetensors",
+        "model.layers.27.mlp.switch_mlp.down_proj.weight": "model-00021-of-00024.safetensors",
+        "model.layers.27.mlp.switch_mlp.gate_proj.biases": "model-00020-of-00024.safetensors",
+        "model.layers.27.mlp.switch_mlp.gate_proj.scales": "model-00020-of-00024.safetensors",
+        "model.layers.27.mlp.switch_mlp.gate_proj.weight": "model-00020-of-00024.safetensors",
+        "model.layers.27.mlp.switch_mlp.up_proj.biases": "model-00021-of-00024.safetensors",
+        "model.layers.27.mlp.switch_mlp.up_proj.scales": "model-00021-of-00024.safetensors",
+        "model.layers.27.mlp.switch_mlp.up_proj.weight": "model-00021-of-00024.safetensors",
+        "model.layers.27.post_attention_layernorm.weight": "model-00021-of-00024.safetensors",
+        "model.layers.28.attention.dense.biases": "model-00021-of-00024.safetensors",
+        "model.layers.28.attention.dense.scales": "model-00021-of-00024.safetensors",
+        "model.layers.28.attention.dense.weight": "model-00021-of-00024.safetensors",
+        "model.layers.28.attention.key_layernorm.weight": "model-00021-of-00024.safetensors",
+        "model.layers.28.attention.query_key_value.biases": "model-00021-of-00024.safetensors",
+        "model.layers.28.attention.query_key_value.scales": "model-00021-of-00024.safetensors",
+        "model.layers.28.attention.query_key_value.weight": "model-00021-of-00024.safetensors",
+        "model.layers.28.attention.query_layernorm.weight": "model-00021-of-00024.safetensors",
+        "model.layers.28.input_layernorm.weight": "model-00022-of-00024.safetensors",
+        "model.layers.28.mlp.gate.expert_bias": "model-00022-of-00024.safetensors",
+        "model.layers.28.mlp.gate.weight": "model-00022-of-00024.safetensors",
+        "model.layers.28.mlp.shared_experts.down_proj.biases": "model-00022-of-00024.safetensors",
+        "model.layers.28.mlp.shared_experts.down_proj.scales": "model-00022-of-00024.safetensors",
+        "model.layers.28.mlp.shared_experts.down_proj.weight": "model-00022-of-00024.safetensors",
+        "model.layers.28.mlp.shared_experts.gate_proj.biases": "model-00022-of-00024.safetensors",
+        "model.layers.28.mlp.shared_experts.gate_proj.scales": "model-00022-of-00024.safetensors",
+        "model.layers.28.mlp.shared_experts.gate_proj.weight": "model-00022-of-00024.safetensors",
+        "model.layers.28.mlp.shared_experts.up_proj.biases": "model-00022-of-00024.safetensors",
+        "model.layers.28.mlp.shared_experts.up_proj.scales": "model-00022-of-00024.safetensors",
+        "model.layers.28.mlp.shared_experts.up_proj.weight": "model-00022-of-00024.safetensors",
+        "model.layers.28.mlp.switch_mlp.down_proj.biases": "model-00022-of-00024.safetensors",
+        "model.layers.28.mlp.switch_mlp.down_proj.scales": "model-00022-of-00024.safetensors",
+        "model.layers.28.mlp.switch_mlp.down_proj.weight": "model-00022-of-00024.safetensors",
+        "model.layers.28.mlp.switch_mlp.gate_proj.biases": "model-00021-of-00024.safetensors",
+        "model.layers.28.mlp.switch_mlp.gate_proj.scales": "model-00021-of-00024.safetensors",
+        "model.layers.28.mlp.switch_mlp.gate_proj.weight": "model-00021-of-00024.safetensors",
+        "model.layers.28.mlp.switch_mlp.up_proj.biases": "model-00021-of-00024.safetensors",
+        "model.layers.28.mlp.switch_mlp.up_proj.scales": "model-00021-of-00024.safetensors",
+        "model.layers.28.mlp.switch_mlp.up_proj.weight": "model-00021-of-00024.safetensors",
+        "model.layers.28.post_attention_layernorm.weight": "model-00022-of-00024.safetensors",
+        "model.layers.29.attention.dense.biases": "model-00022-of-00024.safetensors",
+        "model.layers.29.attention.dense.scales": "model-00022-of-00024.safetensors",
+        "model.layers.29.attention.dense.weight": "model-00022-of-00024.safetensors",
+        "model.layers.29.attention.key_layernorm.weight": "model-00022-of-00024.safetensors",
+        "model.layers.29.attention.query_key_value.biases": "model-00022-of-00024.safetensors",
+        "model.layers.29.attention.query_key_value.scales": "model-00022-of-00024.safetensors",
+        "model.layers.29.attention.query_key_value.weight": "model-00022-of-00024.safetensors",
+        "model.layers.29.attention.query_layernorm.weight": "model-00022-of-00024.safetensors",
+        "model.layers.29.input_layernorm.weight": "model-00022-of-00024.safetensors",
+        "model.layers.29.mlp.gate.expert_bias": "model-00022-of-00024.safetensors",
+        "model.layers.29.mlp.gate.weight": "model-00022-of-00024.safetensors",
+        "model.layers.29.mlp.shared_experts.down_proj.biases": "model-00022-of-00024.safetensors",
+        "model.layers.29.mlp.shared_experts.down_proj.scales": "model-00022-of-00024.safetensors",
+        "model.layers.29.mlp.shared_experts.down_proj.weight": "model-00022-of-00024.safetensors",
+        "model.layers.29.mlp.shared_experts.gate_proj.biases": "model-00022-of-00024.safetensors",
+        "model.layers.29.mlp.shared_experts.gate_proj.scales": "model-00022-of-00024.safetensors",
+        "model.layers.29.mlp.shared_experts.gate_proj.weight": "model-00022-of-00024.safetensors",
+        "model.layers.29.mlp.shared_experts.up_proj.biases": "model-00022-of-00024.safetensors",
+        "model.layers.29.mlp.shared_experts.up_proj.scales": "model-00022-of-00024.safetensors",
+        "model.layers.29.mlp.shared_experts.up_proj.weight": "model-00022-of-00024.safetensors",
+        "model.layers.29.mlp.switch_mlp.down_proj.biases": "model-00022-of-00024.safetensors",
+        "model.layers.29.mlp.switch_mlp.down_proj.scales": "model-00022-of-00024.safetensors",
+        "model.layers.29.mlp.switch_mlp.down_proj.weight": "model-00022-of-00024.safetensors",
+        "model.layers.29.mlp.switch_mlp.gate_proj.biases": "model-00022-of-00024.safetensors",
+        "model.layers.29.mlp.switch_mlp.gate_proj.scales": "model-00022-of-00024.safetensors",
+        "model.layers.29.mlp.switch_mlp.gate_proj.weight": "model-00022-of-00024.safetensors",
+        "model.layers.29.mlp.switch_mlp.up_proj.biases": "model-00022-of-00024.safetensors",
+        "model.layers.29.mlp.switch_mlp.up_proj.scales": "model-00022-of-00024.safetensors",
+        "model.layers.29.mlp.switch_mlp.up_proj.weight": "model-00022-of-00024.safetensors",
+        "model.layers.29.post_attention_layernorm.weight": "model-00022-of-00024.safetensors",
+        "model.layers.3.attention.dense.biases": "model-00002-of-00024.safetensors",
+        "model.layers.3.attention.dense.scales": "model-00002-of-00024.safetensors",
+        "model.layers.3.attention.dense.weight": "model-00002-of-00024.safetensors",
+        "model.layers.3.attention.key_layernorm.weight": "model-00002-of-00024.safetensors",
+        "model.layers.3.attention.query_key_value.biases": "model-00002-of-00024.safetensors",
+        "model.layers.3.attention.query_key_value.scales": "model-00002-of-00024.safetensors",
+        "model.layers.3.attention.query_key_value.weight": "model-00002-of-00024.safetensors",
+        "model.layers.3.attention.query_layernorm.weight": "model-00002-of-00024.safetensors",
+        "model.layers.3.input_layernorm.weight": "model-00003-of-00024.safetensors",
+        "model.layers.3.mlp.gate.expert_bias": "model-00003-of-00024.safetensors",
+        "model.layers.3.mlp.gate.weight": "model-00003-of-00024.safetensors",
+        "model.layers.3.mlp.shared_experts.down_proj.biases": "model-00003-of-00024.safetensors",
+        "model.layers.3.mlp.shared_experts.down_proj.scales": "model-00003-of-00024.safetensors",
+        "model.layers.3.mlp.shared_experts.down_proj.weight": "model-00003-of-00024.safetensors",
+        "model.layers.3.mlp.shared_experts.gate_proj.biases": "model-00003-of-00024.safetensors",
+        "model.layers.3.mlp.shared_experts.gate_proj.scales": "model-00003-of-00024.safetensors",
+        "model.layers.3.mlp.shared_experts.gate_proj.weight": "model-00003-of-00024.safetensors",
+        "model.layers.3.mlp.shared_experts.up_proj.biases": "model-00003-of-00024.safetensors",
+        "model.layers.3.mlp.shared_experts.up_proj.scales": "model-00003-of-00024.safetensors",
+        "model.layers.3.mlp.shared_experts.up_proj.weight": "model-00003-of-00024.safetensors",
+        "model.layers.3.mlp.switch_mlp.down_proj.biases": "model-00003-of-00024.safetensors",
+        "model.layers.3.mlp.switch_mlp.down_proj.scales": "model-00003-of-00024.safetensors",
+        "model.layers.3.mlp.switch_mlp.down_proj.weight": "model-00003-of-00024.safetensors",
+        "model.layers.3.mlp.switch_mlp.gate_proj.biases": "model-00002-of-00024.safetensors",
+        "model.layers.3.mlp.switch_mlp.gate_proj.scales": "model-00002-of-00024.safetensors",
+        "model.layers.3.mlp.switch_mlp.gate_proj.weight": "model-00002-of-00024.safetensors",
+        "model.layers.3.mlp.switch_mlp.up_proj.biases": "model-00003-of-00024.safetensors",
+        "model.layers.3.mlp.switch_mlp.up_proj.scales": "model-00003-of-00024.safetensors",
+        "model.layers.3.mlp.switch_mlp.up_proj.weight": "model-00003-of-00024.safetensors",
+        "model.layers.3.post_attention_layernorm.weight": "model-00003-of-00024.safetensors",
+        "model.layers.30.attention.dense.biases": "model-00022-of-00024.safetensors",
+        "model.layers.30.attention.dense.scales": "model-00022-of-00024.safetensors",
+        "model.layers.30.attention.dense.weight": "model-00022-of-00024.safetensors",
+        "model.layers.30.attention.key_layernorm.weight": "model-00022-of-00024.safetensors",
+        "model.layers.30.attention.query_key_value.biases": "model-00022-of-00024.safetensors",
+        "model.layers.30.attention.query_key_value.scales": "model-00022-of-00024.safetensors",
+        "model.layers.30.attention.query_key_value.weight": "model-00022-of-00024.safetensors",
+        "model.layers.30.attention.query_layernorm.weight": "model-00022-of-00024.safetensors",
+        "model.layers.30.input_layernorm.weight": "model-00023-of-00024.safetensors",
+        "model.layers.30.mlp.gate.expert_bias": "model-00023-of-00024.safetensors",
+        "model.layers.30.mlp.gate.weight": "model-00023-of-00024.safetensors",
+        "model.layers.30.mlp.shared_experts.down_proj.biases": "model-00023-of-00024.safetensors",
+        "model.layers.30.mlp.shared_experts.down_proj.scales": "model-00023-of-00024.safetensors",
+        "model.layers.30.mlp.shared_experts.down_proj.weight": "model-00023-of-00024.safetensors",
+        "model.layers.30.mlp.shared_experts.gate_proj.biases": "model-00023-of-00024.safetensors",
+        "model.layers.30.mlp.shared_experts.gate_proj.scales": "model-00023-of-00024.safetensors",
+        "model.layers.30.mlp.shared_experts.gate_proj.weight": "model-00023-of-00024.safetensors",
+        "model.layers.30.mlp.shared_experts.up_proj.biases": "model-00023-of-00024.safetensors",
+        "model.layers.30.mlp.shared_experts.up_proj.scales": "model-00023-of-00024.safetensors",
+        "model.layers.30.mlp.shared_experts.up_proj.weight": "model-00023-of-00024.safetensors",
+        "model.layers.30.mlp.switch_mlp.down_proj.biases": "model-00023-of-00024.safetensors",
+        "model.layers.30.mlp.switch_mlp.down_proj.scales": "model-00023-of-00024.safetensors",
+        "model.layers.30.mlp.switch_mlp.down_proj.weight": "model-00023-of-00024.safetensors",
+        "model.layers.30.mlp.switch_mlp.gate_proj.biases": "model-00023-of-00024.safetensors",
+        "model.layers.30.mlp.switch_mlp.gate_proj.scales": "model-00023-of-00024.safetensors",
+        "model.layers.30.mlp.switch_mlp.gate_proj.weight": "model-00023-of-00024.safetensors",
+        "model.layers.30.mlp.switch_mlp.up_proj.biases": "model-00023-of-00024.safetensors",
+        "model.layers.30.mlp.switch_mlp.up_proj.scales": "model-00023-of-00024.safetensors",
+        "model.layers.30.mlp.switch_mlp.up_proj.weight": "model-00023-of-00024.safetensors",
+        "model.layers.30.post_attention_layernorm.weight": "model-00023-of-00024.safetensors",
+        "model.layers.31.attention.dense.biases": "model-00023-of-00024.safetensors",
+        "model.layers.31.attention.dense.scales": "model-00023-of-00024.safetensors",
+        "model.layers.31.attention.dense.weight": "model-00023-of-00024.safetensors",
+        "model.layers.31.attention.key_layernorm.weight": "model-00023-of-00024.safetensors",
+        "model.layers.31.attention.query_key_value.biases": "model-00023-of-00024.safetensors",
+        "model.layers.31.attention.query_key_value.scales": "model-00023-of-00024.safetensors",
+        "model.layers.31.attention.query_key_value.weight": "model-00023-of-00024.safetensors",
+        "model.layers.31.attention.query_layernorm.weight": "model-00023-of-00024.safetensors",
+        "model.layers.31.input_layernorm.weight": "model-00024-of-00024.safetensors",
+        "model.layers.31.mlp.gate.expert_bias": "model-00024-of-00024.safetensors",
+        "model.layers.31.mlp.gate.weight": "model-00024-of-00024.safetensors",
+        "model.layers.31.mlp.shared_experts.down_proj.biases": "model-00024-of-00024.safetensors",
+        "model.layers.31.mlp.shared_experts.down_proj.scales": "model-00024-of-00024.safetensors",
+        "model.layers.31.mlp.shared_experts.down_proj.weight": "model-00024-of-00024.safetensors",
+        "model.layers.31.mlp.shared_experts.gate_proj.biases": "model-00024-of-00024.safetensors",
+        "model.layers.31.mlp.shared_experts.gate_proj.scales": "model-00024-of-00024.safetensors",
+        "model.layers.31.mlp.shared_experts.gate_proj.weight": "model-00024-of-00024.safetensors",
+        "model.layers.31.mlp.shared_experts.up_proj.biases": "model-00024-of-00024.safetensors",
+        "model.layers.31.mlp.shared_experts.up_proj.scales": "model-00024-of-00024.safetensors",
+        "model.layers.31.mlp.shared_experts.up_proj.weight": "model-00024-of-00024.safetensors",
+        "model.layers.31.mlp.switch_mlp.down_proj.biases": "model-00024-of-00024.safetensors",
+        "model.layers.31.mlp.switch_mlp.down_proj.scales": "model-00024-of-00024.safetensors",
+        "model.layers.31.mlp.switch_mlp.down_proj.weight": "model-00024-of-00024.safetensors",
+        "model.layers.31.mlp.switch_mlp.gate_proj.biases": "model-00023-of-00024.safetensors",
+        "model.layers.31.mlp.switch_mlp.gate_proj.scales": "model-00023-of-00024.safetensors",
+        "model.layers.31.mlp.switch_mlp.gate_proj.weight": "model-00023-of-00024.safetensors",
+        "model.layers.31.mlp.switch_mlp.up_proj.biases": "model-00024-of-00024.safetensors",
+        "model.layers.31.mlp.switch_mlp.up_proj.scales": "model-00024-of-00024.safetensors",
+        "model.layers.31.mlp.switch_mlp.up_proj.weight": "model-00024-of-00024.safetensors",
+        "model.layers.31.post_attention_layernorm.weight": "model-00024-of-00024.safetensors",
+        "model.layers.4.attention.dense.biases": "model-00003-of-00024.safetensors",
+        "model.layers.4.attention.dense.scales": "model-00003-of-00024.safetensors",
+        "model.layers.4.attention.dense.weight": "model-00003-of-00024.safetensors",
+        "model.layers.4.attention.key_layernorm.weight": "model-00003-of-00024.safetensors",
+        "model.layers.4.attention.query_key_value.biases": "model-00003-of-00024.safetensors",
+        "model.layers.4.attention.query_key_value.scales": "model-00003-of-00024.safetensors",
+        "model.layers.4.attention.query_key_value.weight": "model-00003-of-00024.safetensors",
+        "model.layers.4.attention.query_layernorm.weight": "model-00003-of-00024.safetensors",
+        "model.layers.4.input_layernorm.weight": "model-00004-of-00024.safetensors",
+        "model.layers.4.mlp.gate.expert_bias": "model-00004-of-00024.safetensors",
+        "model.layers.4.mlp.gate.weight": "model-00004-of-00024.safetensors",
+        "model.layers.4.mlp.shared_experts.down_proj.biases": "model-00004-of-00024.safetensors",
+        "model.layers.4.mlp.shared_experts.down_proj.scales": "model-00004-of-00024.safetensors",
+        "model.layers.4.mlp.shared_experts.down_proj.weight": "model-00004-of-00024.safetensors",
+        "model.layers.4.mlp.shared_experts.gate_proj.biases": "model-00004-of-00024.safetensors",
+        "model.layers.4.mlp.shared_experts.gate_proj.scales": "model-00004-of-00024.safetensors",
+        "model.layers.4.mlp.shared_experts.gate_proj.weight": "model-00004-of-00024.safetensors",
+        "model.layers.4.mlp.shared_experts.up_proj.biases": "model-00004-of-00024.safetensors",
+        "model.layers.4.mlp.shared_experts.up_proj.scales": "model-00004-of-00024.safetensors",
+        "model.layers.4.mlp.shared_experts.up_proj.weight": "model-00004-of-00024.safetensors",
+        "model.layers.4.mlp.switch_mlp.down_proj.biases": "model-00004-of-00024.safetensors",
+        "model.layers.4.mlp.switch_mlp.down_proj.scales": "model-00004-of-00024.safetensors",
+        "model.layers.4.mlp.switch_mlp.down_proj.weight": "model-00004-of-00024.safetensors",
+        "model.layers.4.mlp.switch_mlp.gate_proj.biases": "model-00003-of-00024.safetensors",
+        "model.layers.4.mlp.switch_mlp.gate_proj.scales": "model-00003-of-00024.safetensors",
+        "model.layers.4.mlp.switch_mlp.gate_proj.weight": "model-00003-of-00024.safetensors",
+        "model.layers.4.mlp.switch_mlp.up_proj.biases": "model-00003-of-00024.safetensors",
+        "model.layers.4.mlp.switch_mlp.up_proj.scales": "model-00003-of-00024.safetensors",
+        "model.layers.4.mlp.switch_mlp.up_proj.weight": "model-00003-of-00024.safetensors",
+        "model.layers.4.post_attention_layernorm.weight": "model-00004-of-00024.safetensors",
+        "model.layers.5.attention.dense.biases": "model-00004-of-00024.safetensors",
+        "model.layers.5.attention.dense.scales": "model-00004-of-00024.safetensors",
+        "model.layers.5.attention.dense.weight": "model-00004-of-00024.safetensors",
+        "model.layers.5.attention.key_layernorm.weight": "model-00004-of-00024.safetensors",
+        "model.layers.5.attention.query_key_value.biases": "model-00004-of-00024.safetensors",
+        "model.layers.5.attention.query_key_value.scales": "model-00004-of-00024.safetensors",
+        "model.layers.5.attention.query_key_value.weight": "model-00004-of-00024.safetensors",
+        "model.layers.5.attention.query_layernorm.weight": "model-00004-of-00024.safetensors",
+        "model.layers.5.input_layernorm.weight": "model-00004-of-00024.safetensors",
+        "model.layers.5.mlp.gate.expert_bias": "model-00004-of-00024.safetensors",
+        "model.layers.5.mlp.gate.weight": "model-00004-of-00024.safetensors",
+        "model.layers.5.mlp.shared_experts.down_proj.biases": "model-00004-of-00024.safetensors",
+        "model.layers.5.mlp.shared_experts.down_proj.scales": "model-00004-of-00024.safetensors",
+        "model.layers.5.mlp.shared_experts.down_proj.weight": "model-00004-of-00024.safetensors",
+        "model.layers.5.mlp.shared_experts.gate_proj.biases": "model-00004-of-00024.safetensors",
+        "model.layers.5.mlp.shared_experts.gate_proj.scales": "model-00004-of-00024.safetensors",
+        "model.layers.5.mlp.shared_experts.gate_proj.weight": "model-00004-of-00024.safetensors",
+        "model.layers.5.mlp.shared_experts.up_proj.biases": "model-00004-of-00024.safetensors",
+        "model.layers.5.mlp.shared_experts.up_proj.scales": "model-00004-of-00024.safetensors",
+        "model.layers.5.mlp.shared_experts.up_proj.weight": "model-00004-of-00024.safetensors",
+        "model.layers.5.mlp.switch_mlp.down_proj.biases": "model-00004-of-00024.safetensors",
+        "model.layers.5.mlp.switch_mlp.down_proj.scales": "model-00004-of-00024.safetensors",
+        "model.layers.5.mlp.switch_mlp.down_proj.weight": "model-00004-of-00024.safetensors",
+        "model.layers.5.mlp.switch_mlp.gate_proj.biases": "model-00004-of-00024.safetensors",
+        "model.layers.5.mlp.switch_mlp.gate_proj.scales": "model-00004-of-00024.safetensors",
+        "model.layers.5.mlp.switch_mlp.gate_proj.weight": "model-00004-of-00024.safetensors",
+        "model.layers.5.mlp.switch_mlp.up_proj.biases": "model-00004-of-00024.safetensors",
+        "model.layers.5.mlp.switch_mlp.up_proj.scales": "model-00004-of-00024.safetensors",
+        "model.layers.5.mlp.switch_mlp.up_proj.weight": "model-00004-of-00024.safetensors",
+        "model.layers.5.post_attention_layernorm.weight": "model-00004-of-00024.safetensors",
+        "model.layers.6.attention.dense.biases": "model-00004-of-00024.safetensors",
+        "model.layers.6.attention.dense.scales": "model-00004-of-00024.safetensors",
+        "model.layers.6.attention.dense.weight": "model-00004-of-00024.safetensors",
+        "model.layers.6.attention.key_layernorm.weight": "model-00004-of-00024.safetensors",
+        "model.layers.6.attention.query_key_value.biases": "model-00004-of-00024.safetensors",
+        "model.layers.6.attention.query_key_value.scales": "model-00004-of-00024.safetensors",
+        "model.layers.6.attention.query_key_value.weight": "model-00004-of-00024.safetensors",
+        "model.layers.6.attention.query_layernorm.weight": "model-00004-of-00024.safetensors",
+        "model.layers.6.input_layernorm.weight": "model-00005-of-00024.safetensors",
+        "model.layers.6.mlp.gate.expert_bias": "model-00005-of-00024.safetensors",
+        "model.layers.6.mlp.gate.weight": "model-00005-of-00024.safetensors",
+        "model.layers.6.mlp.shared_experts.down_proj.biases": "model-00005-of-00024.safetensors",
+        "model.layers.6.mlp.shared_experts.down_proj.scales": "model-00005-of-00024.safetensors",
+        "model.layers.6.mlp.shared_experts.down_proj.weight": "model-00005-of-00024.safetensors",
+        "model.layers.6.mlp.shared_experts.gate_proj.biases": "model-00005-of-00024.safetensors",
+        "model.layers.6.mlp.shared_experts.gate_proj.scales": "model-00005-of-00024.safetensors",
+        "model.layers.6.mlp.shared_experts.gate_proj.weight": "model-00005-of-00024.safetensors",
+        "model.layers.6.mlp.shared_experts.up_proj.biases": "model-00005-of-00024.safetensors",
+        "model.layers.6.mlp.shared_experts.up_proj.scales": "model-00005-of-00024.safetensors",
+        "model.layers.6.mlp.shared_experts.up_proj.weight": "model-00005-of-00024.safetensors",
+        "model.layers.6.mlp.switch_mlp.down_proj.biases": "model-00005-of-00024.safetensors",
+        "model.layers.6.mlp.switch_mlp.down_proj.scales": "model-00005-of-00024.safetensors",
+        "model.layers.6.mlp.switch_mlp.down_proj.weight": "model-00005-of-00024.safetensors",
+        "model.layers.6.mlp.switch_mlp.gate_proj.biases": "model-00005-of-00024.safetensors",
+        "model.layers.6.mlp.switch_mlp.gate_proj.scales": "model-00005-of-00024.safetensors",
+        "model.layers.6.mlp.switch_mlp.gate_proj.weight": "model-00005-of-00024.safetensors",
+        "model.layers.6.mlp.switch_mlp.up_proj.biases": "model-00005-of-00024.safetensors",
+        "model.layers.6.mlp.switch_mlp.up_proj.scales": "model-00005-of-00024.safetensors",
+        "model.layers.6.mlp.switch_mlp.up_proj.weight": "model-00005-of-00024.safetensors",
+        "model.layers.6.post_attention_layernorm.weight": "model-00005-of-00024.safetensors",
+        "model.layers.7.attention.dense.biases": "model-00005-of-00024.safetensors",
+        "model.layers.7.attention.dense.scales": "model-00005-of-00024.safetensors",
+        "model.layers.7.attention.dense.weight": "model-00005-of-00024.safetensors",
+        "model.layers.7.attention.key_layernorm.weight": "model-00005-of-00024.safetensors",
+        "model.layers.7.attention.query_key_value.biases": "model-00005-of-00024.safetensors",
+        "model.layers.7.attention.query_key_value.scales": "model-00005-of-00024.safetensors",
+        "model.layers.7.attention.query_key_value.weight": "model-00005-of-00024.safetensors",
+        "model.layers.7.attention.query_layernorm.weight": "model-00005-of-00024.safetensors",
+        "model.layers.7.input_layernorm.weight": "model-00006-of-00024.safetensors",
+        "model.layers.7.mlp.gate.expert_bias": "model-00006-of-00024.safetensors",
+        "model.layers.7.mlp.gate.weight": "model-00006-of-00024.safetensors",
+        "model.layers.7.mlp.shared_experts.down_proj.biases": "model-00006-of-00024.safetensors",
+        "model.layers.7.mlp.shared_experts.down_proj.scales": "model-00006-of-00024.safetensors",
+        "model.layers.7.mlp.shared_experts.down_proj.weight": "model-00006-of-00024.safetensors",
+        "model.layers.7.mlp.shared_experts.gate_proj.biases": "model-00006-of-00024.safetensors",
+        "model.layers.7.mlp.shared_experts.gate_proj.scales": "model-00006-of-00024.safetensors",
+        "model.layers.7.mlp.shared_experts.gate_proj.weight": "model-00006-of-00024.safetensors",
+        "model.layers.7.mlp.shared_experts.up_proj.biases": "model-00006-of-00024.safetensors",
+        "model.layers.7.mlp.shared_experts.up_proj.scales": "model-00006-of-00024.safetensors",
+        "model.layers.7.mlp.shared_experts.up_proj.weight": "model-00006-of-00024.safetensors",
+        "model.layers.7.mlp.switch_mlp.down_proj.biases": "model-00006-of-00024.safetensors",
+        "model.layers.7.mlp.switch_mlp.down_proj.scales": "model-00006-of-00024.safetensors",
+        "model.layers.7.mlp.switch_mlp.down_proj.weight": "model-00006-of-00024.safetensors",
+        "model.layers.7.mlp.switch_mlp.gate_proj.biases": "model-00005-of-00024.safetensors",
+        "model.layers.7.mlp.switch_mlp.gate_proj.scales": "model-00005-of-00024.safetensors",
+        "model.layers.7.mlp.switch_mlp.gate_proj.weight": "model-00005-of-00024.safetensors",
+        "model.layers.7.mlp.switch_mlp.up_proj.biases": "model-00006-of-00024.safetensors",
+        "model.layers.7.mlp.switch_mlp.up_proj.scales": "model-00006-of-00024.safetensors",
+        "model.layers.7.mlp.switch_mlp.up_proj.weight": "model-00006-of-00024.safetensors",
+        "model.layers.7.post_attention_layernorm.weight": "model-00006-of-00024.safetensors",
+        "model.layers.8.attention.dense.biases": "model-00006-of-00024.safetensors",
+        "model.layers.8.attention.dense.scales": "model-00006-of-00024.safetensors",
+        "model.layers.8.attention.dense.weight": "model-00006-of-00024.safetensors",
+        "model.layers.8.attention.key_layernorm.weight": "model-00006-of-00024.safetensors",
+        "model.layers.8.attention.query_key_value.biases": "model-00006-of-00024.safetensors",
+        "model.layers.8.attention.query_key_value.scales": "model-00006-of-00024.safetensors",
+        "model.layers.8.attention.query_key_value.weight": "model-00006-of-00024.safetensors",
+        "model.layers.8.attention.query_layernorm.weight": "model-00006-of-00024.safetensors",
+        "model.layers.8.input_layernorm.weight": "model-00007-of-00024.safetensors",
+        "model.layers.8.mlp.gate.expert_bias": "model-00007-of-00024.safetensors",
+        "model.layers.8.mlp.gate.weight": "model-00007-of-00024.safetensors",
+        "model.layers.8.mlp.shared_experts.down_proj.biases": "model-00007-of-00024.safetensors",
+        "model.layers.8.mlp.shared_experts.down_proj.scales": "model-00007-of-00024.safetensors",
+        "model.layers.8.mlp.shared_experts.down_proj.weight": "model-00007-of-00024.safetensors",
+        "model.layers.8.mlp.shared_experts.gate_proj.biases": "model-00007-of-00024.safetensors",
+        "model.layers.8.mlp.shared_experts.gate_proj.scales": "model-00007-of-00024.safetensors",
+        "model.layers.8.mlp.shared_experts.gate_proj.weight": "model-00007-of-00024.safetensors",
+        "model.layers.8.mlp.shared_experts.up_proj.biases": "model-00007-of-00024.safetensors",
+        "model.layers.8.mlp.shared_experts.up_proj.scales": "model-00007-of-00024.safetensors",
+        "model.layers.8.mlp.shared_experts.up_proj.weight": "model-00007-of-00024.safetensors",
+        "model.layers.8.mlp.switch_mlp.down_proj.biases": "model-00007-of-00024.safetensors",
+        "model.layers.8.mlp.switch_mlp.down_proj.scales": "model-00007-of-00024.safetensors",
+        "model.layers.8.mlp.switch_mlp.down_proj.weight": "model-00007-of-00024.safetensors",
+        "model.layers.8.mlp.switch_mlp.gate_proj.biases": "model-00006-of-00024.safetensors",
+        "model.layers.8.mlp.switch_mlp.gate_proj.scales": "model-00006-of-00024.safetensors",
+        "model.layers.8.mlp.switch_mlp.gate_proj.weight": "model-00006-of-00024.safetensors",
+        "model.layers.8.mlp.switch_mlp.up_proj.biases": "model-00006-of-00024.safetensors",
+        "model.layers.8.mlp.switch_mlp.up_proj.scales": "model-00006-of-00024.safetensors",
+        "model.layers.8.mlp.switch_mlp.up_proj.weight": "model-00006-of-00024.safetensors",
+        "model.layers.8.post_attention_layernorm.weight": "model-00007-of-00024.safetensors",
+        "model.layers.9.attention.dense.biases": "model-00007-of-00024.safetensors",
+        "model.layers.9.attention.dense.scales": "model-00007-of-00024.safetensors",
+        "model.layers.9.attention.dense.weight": "model-00007-of-00024.safetensors",
+        "model.layers.9.attention.key_layernorm.weight": "model-00007-of-00024.safetensors",
+        "model.layers.9.attention.query_key_value.biases": "model-00007-of-00024.safetensors",
+        "model.layers.9.attention.query_key_value.scales": "model-00007-of-00024.safetensors",
+        "model.layers.9.attention.query_key_value.weight": "model-00007-of-00024.safetensors",
+        "model.layers.9.attention.query_layernorm.weight": "model-00007-of-00024.safetensors",
+        "model.layers.9.input_layernorm.weight": "model-00007-of-00024.safetensors",
+        "model.layers.9.mlp.gate.expert_bias": "model-00007-of-00024.safetensors",
+        "model.layers.9.mlp.gate.weight": "model-00007-of-00024.safetensors",
+        "model.layers.9.mlp.shared_experts.down_proj.biases": "model-00007-of-00024.safetensors",
+        "model.layers.9.mlp.shared_experts.down_proj.scales": "model-00007-of-00024.safetensors",
+        "model.layers.9.mlp.shared_experts.down_proj.weight": "model-00007-of-00024.safetensors",
+        "model.layers.9.mlp.shared_experts.gate_proj.biases": "model-00007-of-00024.safetensors",
+        "model.layers.9.mlp.shared_experts.gate_proj.scales": "model-00007-of-00024.safetensors",
+        "model.layers.9.mlp.shared_experts.gate_proj.weight": "model-00007-of-00024.safetensors",
+        "model.layers.9.mlp.shared_experts.up_proj.biases": "model-00007-of-00024.safetensors",
+        "model.layers.9.mlp.shared_experts.up_proj.scales": "model-00007-of-00024.safetensors",
+        "model.layers.9.mlp.shared_experts.up_proj.weight": "model-00007-of-00024.safetensors",
+        "model.layers.9.mlp.switch_mlp.down_proj.biases": "model-00007-of-00024.safetensors",
+        "model.layers.9.mlp.switch_mlp.down_proj.scales": "model-00007-of-00024.safetensors",
+        "model.layers.9.mlp.switch_mlp.down_proj.weight": "model-00007-of-00024.safetensors",
+        "model.layers.9.mlp.switch_mlp.gate_proj.biases": "model-00007-of-00024.safetensors",
+        "model.layers.9.mlp.switch_mlp.gate_proj.scales": "model-00007-of-00024.safetensors",
+        "model.layers.9.mlp.switch_mlp.gate_proj.weight": "model-00007-of-00024.safetensors",
+        "model.layers.9.mlp.switch_mlp.up_proj.biases": "model-00007-of-00024.safetensors",
+        "model.layers.9.mlp.switch_mlp.up_proj.scales": "model-00007-of-00024.safetensors",
+        "model.layers.9.mlp.switch_mlp.up_proj.weight": "model-00007-of-00024.safetensors",
+        "model.layers.9.post_attention_layernorm.weight": "model-00007-of-00024.safetensors",
+        "model.norm.weight": "model-00024-of-00024.safetensors",
+        "model.word_embeddings.biases": "model-00001-of-00024.safetensors",
+        "model.word_embeddings.scales": "model-00001-of-00024.safetensors",
+        "model.word_embeddings.weight": "model-00001-of-00024.safetensors"
+    }
+}

modeling_llada2_moe.py

@@ -0,0 +1,1407 @@
+# Copyright 2025 Antgroup 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 LLaDA2MoE model."""
+
+import math
+from typing import List, Callable, Optional, Tuple, Union
+
+import torch
+import torch.nn.functional as F
+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 (
+    MoeModelOutputWithPast,
+    MoeCausalLMOutputWithPast,
+)
+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.pytorch_utils import (
+    ALL_LAYERNORM_LAYERS,
+    is_torch_greater_or_equal_than_1_13,
+)
+from transformers.utils import (
+    TransformersKwargs,
+    add_start_docstrings,
+    add_start_docstrings_to_model_forward,
+    logging,
+    replace_return_docstrings,
+)
+from transformers.utils.import_utils import is_torch_fx_available
+from .configuration_llada2_moe import LLaDA2MoeConfig
+from transformers.generation.utils import GenerationMixin
+
+
+# This makes `_prepare_4d_causal_attention_mask` a leaf function in the FX graph.
+# It means that the function will not be traced through and simply appear as a node in the graph.
+if is_torch_fx_available():
+    if not is_torch_greater_or_equal_than_1_13:
+        import torch.fx
+
+    _prepare_4d_causal_attention_mask = torch.fx.wrap(_prepare_4d_causal_attention_mask)
+
+
+logger = logging.get_logger(__name__)
+
+_CONFIG_FOR_DOC = "LLaDA2MoeConfig"
+
+
+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.torch.int32), (1, 0)
+    )
+    return (
+        indices,
+        cu_seqlens,
+        max_seqlen_in_batch,
+    )
+
+
+class LLaDA2MoeRMSNorm(nn.Module):
+    def __init__(self, hidden_size, eps=1e-6):
+        """
+        LLaDA2MoeRMSNorm is equivalent to T5LayerNorm
+        """
+        super().__init__()
+        self.weight = nn.Parameter(torch.ones(hidden_size))
+        self.variance_epsilon = eps
+
+    def forward(self, hidden_states):
+        input_dtype = hidden_states.dtype
+        hidden_states = hidden_states.to(torch.float32)
+        variance = hidden_states.pow(2).mean(-1, keepdim=True)
+        hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)
+        return self.weight * hidden_states.to(input_dtype)
+
+
+ALL_LAYERNORM_LAYERS.append(LLaDA2MoeRMSNorm)
+
+
+class LLaDA2MoeRotaryEmbedding(nn.Module):
+    def __init__(self, config: LLaDA2MoeConfig, device=None):
+        super().__init__()
+        # BC: "rope_type" was originally "type"
+        if hasattr(config, "rope_scaling") and config.rope_scaling is not None:
+            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]
+
+        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)
+
+
+# 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
+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`):
+            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 the query and key tensors rotated using the Rotary Position Embedding.
+    """
+    cos = cos.unsqueeze(unsqueeze_dim)
+    sin = sin.unsqueeze(unsqueeze_dim)
+
+    # Keep half or full tensor for later concatenation
+    rotary_dim = cos.shape[-1]
+    q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]
+    k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]
+
+    # Apply rotary embeddings on the first half or full tensor
+    q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)
+    k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)
+
+    # Concatenate back to full shape
+    q_embed = torch.cat([q_embed, q_pass], dim=-1)
+    k_embed = torch.cat([k_embed, k_pass], dim=-1)
+    return q_embed, k_embed
+
+
+class LLaDA2MoeMLP(nn.Module):
+    def __init__(self, config: LLaDA2MoeConfig, intermediate_size: int):
+        super().__init__()
+        self.config = config
+        self.hidden_size = config.hidden_size
+        self.intermediate_size = 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))
+
+
+class LLaDA2MoeGate(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.config = config
+        self.top_k = config.num_experts_per_tok
+        self.num_experts = config.num_experts
+
+        self.n_group = config.n_group
+        self.topk_group = config.topk_group
+
+        # topk selection algorithm
+        self.gating_dim = config.hidden_size
+        self.weight = nn.Parameter(torch.empty((self.num_experts, self.gating_dim)))
+        self.routed_scaling_factor = config.routed_scaling_factor
+
+        self.register_buffer("expert_bias", torch.zeros(self.num_experts))
+        self.reset_parameters()
+
+    def reset_parameters(self) -> None:
+        import torch.nn.init as init
+
+        init.kaiming_uniform_(self.weight, a=math.sqrt(5))
+
+    def group_limited_topk(
+        self,
+        scores: torch.Tensor,
+    ):
+        num_tokens, _ = scores.size()
+        # Organize the experts into groups
+        group_scores = (
+            scores.view(num_tokens, self.n_group, -1).topk(2, dim=-1)[0].sum(dim=-1)
+        )
+        group_idx = torch.topk(group_scores, k=self.topk_group, dim=-1, sorted=False)[1]
+        group_mask = torch.zeros_like(group_scores)
+        group_mask.scatter_(1, group_idx, 1)
+
+        # Mask the experts based on selection groups
+        score_mask = (
+            group_mask.unsqueeze(-1)
+            .expand(num_tokens, self.n_group, self.num_experts // self.n_group)
+            .reshape(num_tokens, -1)
+        )
+
+        masked_scores = scores.masked_fill(~score_mask.bool(), float("-inf"))
+        probs, top_indices = torch.topk(masked_scores, k=self.top_k, dim=-1)
+
+        return probs, top_indices
+
+    def forward(self, hidden_states):
+        # compute gating score
+        hidden_states = hidden_states.view(-1, hidden_states.shape[-1])
+        logits = F.linear(
+            hidden_states.type(torch.float32), self.weight.type(torch.float32)
+        )
+
+        scores = torch.sigmoid(logits.float()).type_as(logits)
+
+        scores_for_routing = scores + self.expert_bias
+        _, topk_idx = self.group_limited_topk(scores_for_routing)
+
+        scores = torch.gather(scores, dim=1, index=topk_idx).type_as(logits)
+
+        topk_weight = (
+            scores / (scores.sum(dim=-1, keepdim=True) + 1e-20)
+            if self.top_k > 1
+            else scores
+        )
+        topk_weight = topk_weight * self.routed_scaling_factor
+
+        return topk_idx, topk_weight, logits
+
+
+class LLaDA2MoeSparseMoeBlock(nn.Module):
+    """
+    A mixed expert module containing shared experts.
+    """
+
+    def __init__(self, config: LLaDA2MoeConfig):
+        super().__init__()
+        self.config = config
+        self.num_experts_per_tok = config.num_experts_per_tok
+        self._setup_experts()
+        self.gate = LLaDA2MoeGate(config)
+        if config.num_shared_experts is not None:
+            self.shared_experts = LLaDA2MoeMLP(
+                config=config,
+                intermediate_size=config.moe_intermediate_size
+                * config.num_shared_experts,
+            )
+
+    def _setup_experts(self):
+        self.experts = nn.ModuleList(
+            [
+                LLaDA2MoeMLP(
+                    config=self.config,
+                    intermediate_size=self.config.moe_intermediate_size,
+                )
+                for _ in range(self.config.num_experts)
+            ]
+        )
+
+    def forward(self, hidden_states):
+        identity = hidden_states
+        bsz, seq_len, h = hidden_states.shape
+        topk_idx, topk_weight, router_logits = self.gate(hidden_states)
+        hidden_states = hidden_states.view(-1, hidden_states.shape[-1])
+        flat_topk_idx = topk_idx.view(-1)
+        if self.training:
+            hidden_states = hidden_states.repeat_interleave(
+                self.num_experts_per_tok, dim=0
+            )
+            y = torch.empty_like(hidden_states)
+            for i, expert in enumerate(self.experts):
+                y[flat_topk_idx == i] = expert(hidden_states[flat_topk_idx == i])
+            y = (y.view(*topk_weight.shape, -1) * topk_weight.unsqueeze(-1)).sum(dim=1)
+            y = y.to(hidden_states.dtype).view(bsz, seq_len, h)
+        else:
+            y = self.moe_infer(hidden_states, topk_idx, topk_weight).view(
+                bsz, seq_len, h
+            )
+        if self.config.num_shared_experts is not None:
+            y = y + self.shared_experts(identity)
+        return y, (
+            router_logits.view(bsz, seq_len, -1),
+            topk_idx.view(bsz, seq_len, -1),
+        )
+
+    @torch.no_grad()
+    def moe_infer(self, x, topk_ids, topk_weight):
+        cnts = topk_ids.new_zeros((topk_ids.shape[0], len(self.experts)))
+        cnts.scatter_(1, topk_ids, 1)
+        tokens_per_expert = cnts.sum(dim=0)
+        idxs = topk_ids.view(-1).argsort()
+        sorted_tokens = x[idxs // topk_ids.shape[1]]
+        tokens_per_expert = tokens_per_expert.cpu().numpy()
+        outputs = []
+        start_idx = 0
+        for i, num_tokens_tensor in enumerate(tokens_per_expert):
+            num_tokens = num_tokens_tensor.item()
+            if num_tokens == 0:
+                continue
+            end_idx = start_idx + num_tokens
+            expert = self.experts[i]
+            tokens_for_this_expert = sorted_tokens[start_idx:end_idx]
+            expert_out = expert(tokens_for_this_expert)
+            outputs.append(expert_out.to(x.device))
+            start_idx = end_idx
+
+        outs = torch.cat(outputs, dim=0) if len(outputs) else sorted_tokens.new_empty(0)
+        new_x = torch.empty_like(outs)
+        new_x[idxs] = outs
+        final_out = (
+            new_x.view(*topk_ids.shape, -1)
+            .type(topk_weight.dtype)
+            .mul_(topk_weight.unsqueeze(dim=-1))
+            .sum(dim=1)
+            .type(new_x.dtype)
+        )
+        return final_out
+
+
+# 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)
+
+
+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,
+    **kwargs: Unpack[TransformersKwargs],
+):
+    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:
+        attn_weights = attn_weights + attention_mask[:, :, :, : key_states.shape[-2]]
+
+    # upcast attention to fp32
+    attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(
+        query.dtype
+    )
+    attn_weights = nn.functional.dropout(
+        attn_weights, 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
+
+
+# Copied from transformers.models.llama.modeling_llama.LlamaAttention with Llama->LLaDA2Moe
+class LLaDA2MoeAttention(nn.Module):
+    """Multi-headed attention from 'Attention Is All You Need' paper"""
+
+    def __init__(self, config: LLaDA2MoeConfig, 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 `layer_idx` is not recommended and will "
+                "to errors during the forward call, if caching is used. Please make sure to provide a `layer_idx` "
+                "when creating this class."
+            )
+        self.attention_dropout = config.attention_dropout
+        self.hidden_size = config.hidden_size
+        self.num_heads = config.num_attention_heads
+        self.head_dim = config.head_dim or self.hidden_size // self.num_heads
+        partial_rotary_factor = (
+            config.partial_rotary_factor
+            if hasattr(config, "partial_rotary_factor")
+            else 1.0
+        )
+        self.rope_dim = int(self.head_dim * partial_rotary_factor)
+        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.scaling = self.head_dim**-0.5
+        self.is_causal = False
+
+        self.query_key_value = nn.Linear(
+            self.hidden_size,
+            (self.num_heads + 2 * self.num_key_value_heads) * self.head_dim,
+            bias=config.use_qkv_bias,
+        )
+
+        if self.config.use_qk_norm:
+            self.query_layernorm = LLaDA2MoeRMSNorm(
+                self.head_dim, eps=config.rms_norm_eps
+            )
+            self.key_layernorm = LLaDA2MoeRMSNorm(
+                self.head_dim, eps=config.rms_norm_eps
+            )
+        self.dense = nn.Linear(
+            self.num_heads * self.head_dim, self.hidden_size, bias=config.use_bias
+        )
+        self.sliding_window = getattr(config, "sliding_window", None)
+
+    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,
+        position_embeddings: Optional[
+            Tuple[torch.Tensor, torch.Tensor]
+        ] = None,  # necessary, but kept here for BC
+        **kwargs,
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+        input_shape = hidden_states.shape[:-1]
+
+        bsz, q_len, _ = hidden_states.size()
+
+        qkv = self.query_key_value(hidden_states)
+        qkv = qkv.view(
+            bsz, q_len, self.num_heads + 2 * self.num_key_value_heads, self.head_dim
+        )
+
+        query_states, key_states, value_states = qkv.split(
+            [self.num_heads, self.num_key_value_heads, self.num_key_value_heads], dim=-2
+        )
+        query_states = query_states.transpose(1, 2)
+        key_states = key_states.transpose(1, 2)
+        value_states = value_states.transpose(1, 2)
+
+        if self.config.use_qk_norm:
+            query_states = self.query_layernorm(query_states)
+            key_states = self.key_layernorm(key_states)
+
+        cos, sin = position_embeddings
+        query_states, key_states = apply_rotary_pos_emb(
+            query_states, key_states, cos, sin
+        )
+
+        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."
+                )
+            cache_kwargs = {"sin": sin, "cos": cos}
+            key_states, value_states = past_key_value.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,
+            sliding_window=self.sliding_window,  # diff with Llama
+            **kwargs,
+        )
+
+        attn_output = attn_output.reshape(*input_shape, -1).contiguous()
+        attn_output = self.dense(attn_output)
+
+        return attn_output, attn_weights, past_key_value
+
+
+class LLaDA2MoeDecoderLayer(nn.Module):
+    def __init__(self, config: LLaDA2MoeConfig, layer_idx: int):
+        super().__init__()
+        self.hidden_size = config.hidden_size
+
+        self.attention = LLaDA2MoeAttention(config=config, layer_idx=layer_idx)
+
+        self.mlp = (
+            LLaDA2MoeSparseMoeBlock(config)
+            if (
+                config.num_experts is not None
+                and layer_idx >= config.first_k_dense_replace
+            )
+            else LLaDA2MoeMLP(config=config, intermediate_size=config.intermediate_size)
+        )
+        self.input_layernorm = LLaDA2MoeRMSNorm(
+            config.hidden_size, eps=config.rms_norm_eps
+        )
+        self.post_attention_layernorm = LLaDA2MoeRMSNorm(
+            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,
+        output_router_logits: Optional[bool] = False,
+        use_cache: Optional[bool] = False,
+        position_embeddings: Optional[
+            Tuple[torch.Tensor, torch.Tensor]
+        ] = None,  # necessary, but kept here for BC
+        **kwargs,
+    ) -> Tuple[
+        torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]
+    ]:
+        """
+        Args:
+            hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
+            attention_mask (`torch.FloatTensor`, *optional*):
+                attention mask of size `(batch_size, sequence_length)` if flash attention is used or `(batch_size, 1,
+                query_sequence_length, key_sequence_length)` if default attention is used.
+            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]`.
+            past_key_value (`Tuple(torch.FloatTensor)`, *optional*):
+                cached past key and value projection states
+            output_attentions (`bool`, *optional*):
+                Whether to return the attentions tensors of all attention layers. See `attentions` under
+                returned tensors for more detail.
+            output_router_logits (`bool`, *optional*):
+                Whether or not to return the logits of all the routers. They are useful for computing the router loss,
+                and should not be returned during inference.
+            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`).
+        """
+        residual = hidden_states
+
+        hidden_states = self.input_layernorm(hidden_states)
+
+        # Self Attention
+        hidden_states, self_attn_weights, present_key_value = self.attention(
+            hidden_states=hidden_states,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_value=past_key_value,
+            output_attentions=output_attentions,
+            position_embeddings=position_embeddings,
+            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)
+        if isinstance(hidden_states, tuple):
+            hidden_states, router_logits = hidden_states
+        else:
+            router_logits = None
+        hidden_states = residual + hidden_states.to(residual.device)
+
+        outputs = (hidden_states,)
+
+        if output_attentions:
+            outputs += (self_attn_weights,)
+
+        if use_cache:
+            outputs += (present_key_value,)
+
+        if output_router_logits:
+            outputs += (router_logits,)
+
+        return outputs
+
+
+LLADA2MOE_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 ([`LLaDA2MoeConfig`]):
+            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 LLaDA2Moe Model outputting raw hidden-states without any specific head on top.",
+    LLADA2MOE_START_DOCSTRING,
+)
+class LLaDA2MoePreTrainedModel(PreTrainedModel):
+    config_class = LLaDA2MoeConfig
+    base_model_prefix = "model"
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["LLaDA2MoeDecoderLayer"]
+    _skip_keys_device_placement = ["past_key_values"]
+    _supports_flash_attn_2 = False
+    _supports_sdpa = True
+    _supports_flex_attn = 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_()
+
+
+LLADA2MOE_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 `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 LLaDA2Moe Model outputting raw hidden-states without any specific head on top.",
+    LLADA2MOE_START_DOCSTRING,
+)
+class LLaDA2MoeModel(LLaDA2MoePreTrainedModel):
+    """
+    Transformer decoder consisting of *config.num_hidden_layers* layers. Each layer is a [`LLaDA2MoeDecoderLayer`]
+    Args:
+        config: LLaDA2MoeConfig
+    """
+
+    def __init__(self, config: LLaDA2MoeConfig):
+        super().__init__(config)
+        self.padding_idx = config.pad_token_id
+        self.vocab_size = config.vocab_size
+
+        self.word_embeddings = nn.Embedding(
+            config.vocab_size, config.hidden_size, self.padding_idx
+        )
+        self.layers = nn.ModuleList(
+            [
+                LLaDA2MoeDecoderLayer(config, layer_idx)
+                for layer_idx in range(config.num_hidden_layers)
+            ]
+        )
+        self._use_sdpa = config._attn_implementation == "sdpa"
+        self._use_flex_attention = config._attn_implementation == "flex_attention"
+        self.norm = LLaDA2MoeRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.rotary_emb = LLaDA2MoeRotaryEmbedding(config=config)
+        self.gradient_checkpointing = False
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.word_embeddings
+
+    def set_input_embeddings(self, value):
+        self.word_embeddings = value
+
+    @add_start_docstrings_to_model_forward(LLADA2MOE_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,
+        output_router_logits: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        **kwargs,
+    ) -> Union[Tuple, MoeModelOutputWithPast]:
+        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
+        )
+        output_router_logits = (
+            output_router_logits
+            if output_router_logits is not None
+            else self.config.output_router_logits
+        )
+        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 input_ids and inputs_embeds at the same time"
+            )
+        elif input_ids is not None:
+            batch_size, seq_length = input_ids.shape[:2]
+        elif inputs_embeds is not None:
+            batch_size, seq_length = inputs_embeds.shape[:2]
+        else:
+            raise ValueError("You have to specify either input_ids or 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`transformers."
+                )
+                use_cache = False
+
+        if use_cache and past_key_values is None:
+            past_key_values = DynamicCache()
+
+        if inputs_embeds is None:
+            inputs_embeds = self.word_embeddings(input_ids)
+
+        past_seen_tokens = (
+            past_key_values.get_seq_length() if past_key_values is not None else 0
+        )
+
+        if position_ids is None:
+            position_ids = torch.arange(
+                past_seen_tokens,
+                past_seen_tokens + inputs_embeds.shape[1],
+                device=inputs_embeds.device,
+            )
+            position_ids = position_ids.unsqueeze(0)
+
+        if self._use_flex_attention:
+            if attention_mask is not None and isinstance(attention_mask, torch.Tensor):
+                attention_mask = _prepare_4d_causal_attention_mask_for_sdpa(
+                    attention_mask,
+                    (batch_size, seq_length),
+                    inputs_embeds,
+                    past_seen_tokens,
+                )
+        elif self._use_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_seen_tokens,
+            )
+        else:
+            # 4d mask is passed through the layers
+            attention_mask = _prepare_4d_causal_attention_mask(
+                attention_mask,
+                (batch_size, seq_length),
+                inputs_embeds,
+                past_seen_tokens,
+            )
+
+        # embed positions
+        hidden_states = inputs_embeds
+
+        # create position embeddings to be shared across the decoder layers
+        position_embeddings = self.rotary_emb(hidden_states, position_ids)
+
+        # decoder layers
+        all_hidden_states = () if output_hidden_states else None
+        all_self_attns = () if output_attentions else None
+        all_router_logits = () if output_router_logits else None
+        next_decoder_cache = None
+
+        for decoder_layer in self.layers:
+            if output_hidden_states:
+                all_hidden_states += (hidden_states,)
+
+            if self.gradient_checkpointing and self.training:
+                layer_outputs = self._gradient_checkpointing_func(
+                    decoder_layer.__call__,
+                    hidden_states,
+                    attention_mask,
+                    position_ids,
+                    past_key_values,
+                    output_attentions,
+                    output_router_logits,
+                    use_cache,
+                    position_embeddings,
+                )
+            else:
+                layer_outputs = decoder_layer(
+                    hidden_states,
+                    attention_mask=attention_mask,
+                    position_ids=position_ids,
+                    past_key_value=past_key_values,
+                    output_attentions=output_attentions,
+                    output_router_logits=output_router_logits,
+                    use_cache=use_cache,
+                    position_embeddings=position_embeddings,
+                )
+            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],)
+
+            if output_router_logits and layer_outputs[-1] is not None:
+                all_router_logits += (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 use_cache:
+            next_cache = next_decoder_cache
+        if not return_dict:
+            return tuple(
+                v
+                for v in [
+                    hidden_states,
+                    next_cache,
+                    all_hidden_states,
+                    all_self_attns,
+                    all_router_logits,
+                ]
+                if v is not None
+            )
+        return MoeModelOutputWithPast(
+            last_hidden_state=hidden_states,
+            past_key_values=next_cache,
+            hidden_states=all_hidden_states,
+            attentions=all_self_attns,
+            router_logits=all_router_logits,
+        )
+
+
+class LLaDA2MoeModelLM(LLaDA2MoePreTrainedModel, GenerationMixin):
+    _tied_weights_keys = ["lm_head.weight"]
+
+    def __init__(self, config: LLaDA2MoeConfig):
+        super().__init__(config)
+        self.model = LLaDA2MoeModel(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.word_embeddings
+
+    def set_input_embeddings(self, value):
+        self.model.word_embeddings = value
+
+    def get_output_embeddings(self):
+        return self.lm_head
+
+    def set_output_embeddings(self, new_embeddings):
+        self.lm_head = new_embeddings
+
+    def set_decoder(self, decoder):
+        self.model = decoder
+
+    def get_decoder(self):
+        return self.model
+
+    @add_start_docstrings_to_model_forward(LLADA2MOE_INPUTS_DOCSTRING)
+    @replace_return_docstrings(
+        output_type=MoeCausalLMOutputWithPast, config_class=_CONFIG_FOR_DOC
+    )
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[List[torch.FloatTensor]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        output_router_logits: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        **kwargs,
+    ) -> Union[Tuple, MoeCausalLMOutputWithPast]:
+        r"""
+        Args:
+            labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+                Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
+                config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
+                (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`.
+        Returns:
+        Example:
+        ```python
+        >>> from transformers import AutoTokenizer
+        >>> model = LLaDA2MoeForCausalLM.from_pretrained(PATH_TO_CONVERTED_WEIGHTS)
+        >>> tokenizer = AutoTokenizer.from_pretrained(PATH_TO_CONVERTED_TOKENIZER)
+        >>> prompt = "Hey, are you conscious? Can you talk to me?"
+        >>> inputs = tokenizer(prompt, return_tensors="pt")
+        >>> # Generate
+        >>> generate_ids = model.generate(inputs.input_ids, max_length=30)
+        >>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
+        "Hey, are you conscious? Can you talk to me?\nI'm not conscious, but I can talk to you."
+        ```"""
+        output_attentions = (
+            output_attentions
+            if output_attentions is not None
+            else self.config.output_attentions
+        )
+        output_hidden_states = (
+            output_hidden_states
+            if output_hidden_states is not None
+            else self.config.output_hidden_states
+        )
+        output_router_logits = (
+            output_router_logits
+            if output_router_logits is not None
+            else self.config.output_router_logits
+        )
+        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,
+            output_router_logits=output_router_logits,
+            return_dict=return_dict,
+            **kwargs,
+        )
+
+        loss = None
+        aux_loss = None
+        hidden_states = outputs[0]
+
+        logits = self.lm_head(hidden_states)
+        logits = logits.float()
+
+        if labels is not None:
+            # LLaDA2.0 will use same label position logits
+            shift_logits = logits
+            shift_labels = labels
+            # Flatten the tokens
+            loss_fct = CrossEntropyLoss()
+            shift_logits = shift_logits.view(-1, self.config.vocab_size)
+            shift_labels = shift_labels.view(-1)
+            # Enable model parallelism
+            shift_labels = shift_labels.to(shift_logits.device)
+            loss = loss_fct(shift_logits, shift_labels)
+
+        if not return_dict:
+            output = (logits,) + outputs[1:]
+            if output_router_logits:
+                output = (aux_loss,) + output
+            return (loss,) + output if loss is not None else output
+
+        return MoeCausalLMOutputWithPast(
+            loss=loss,
+            aux_loss=aux_loss,
+            logits=logits,
+            past_key_values=outputs.past_key_values,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+            router_logits=outputs.router_logits,
+        )
+
+    def prepare_inputs_for_generation(
+        self,
+        input_ids,
+        past_key_values=None,
+        attention_mask=None,
+        inputs_embeds=None,
+        token_type_ids=None,
+        **kwargs,
+    ):
+        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()
+                    if hasattr(past_key_values, "get_max_length")
+                    else past_key_values.get_max_cache_shape()
+                )
+            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 exclusivelly 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
+
+    @staticmethod
+    def _top_k_logits(logits, k):
+        if k is None or k <= 0:
+            return logits
+        else:
+            values, _ = torch.topk(logits, k)
+            min_values = values[..., -1, None]
+            return torch.where(
+                logits < min_values, torch.full_like(logits, float("-inf")), logits
+            )
+
+    @staticmethod
+    def _top_p_logits(logits, p):
+        if p is None or p >= 1.0:
+            return logits
+        sorted_logits, sorted_indices = torch.sort(logits, descending=True)
+        cumulative_probs = torch.cumsum(F.softmax(sorted_logits, dim=-1), dim=-1)
+        sorted_mask = cumulative_probs > p
+        sorted_mask[..., 1:] = sorted_mask[..., :-1].clone()
+        sorted_mask[..., 0] = False
+        mask_indices = torch.scatter(
+            torch.full_like(logits, False, dtype=torch.bool),
+            -1,
+            sorted_indices,
+            sorted_mask,
+        )
+        return logits.masked_fill(mask_indices, float("-inf"))
+
+    def _sample_with_temperature_topk_topp(
+        self, logits, temperature=1.0, top_k=0, top_p=1.0
+    ):
+        orig_shape = logits.shape[:-1]
+        vocab_size = logits.shape[-1]
+        logits = logits.reshape(-1, vocab_size)
+        if temperature > 0 and temperature != 1.0:
+            logits = logits / temperature
+        logits = self._top_k_logits(logits, top_k)
+        logits = self._top_p_logits(logits, top_p)
+        probs = F.softmax(logits, dim=-1)
+        token = torch.multinomial(probs, num_samples=1)
+        token_prob = torch.gather(probs, -1, token)
+        return token.view(*orig_shape), token_prob.view(*orig_shape)
+
+    @staticmethod
+    def _get_num_transfer_tokens(block_length, steps):
+        if steps == 0:
+            return torch.tensor([], dtype=torch.int64)
+        base = block_length // steps
+        remainder = block_length % steps
+        num_transfer_tokens = torch.full((steps,), base, dtype=torch.int64)
+        num_transfer_tokens[:remainder] += 1
+        return num_transfer_tokens
+
+    @torch.no_grad()
+    def generate(
+        self,
+        inputs: Optional[torch.Tensor] = None,
+        temperature: int = 0.0,
+        block_length: int = 32,
+        steps: int = 32,
+        gen_length: int = 2048,
+        top_p: Optional[int] = None,
+        top_k: Optional[int] = None,
+        eos_early_stop: bool = False,
+        minimal_topk: int = 1,
+        threshold: float = 0.95,
+        eos_id: int = 156892,
+        mask_id: int = 156895,
+    ):
+        r"""
+        Generates tokens using a block-wise, iterative refinement strategy.
+        This method operates differently from standard autoregressive generation. It first creates a template of the
+        full desired length, filled with a special `mask_id`. It then processes this template in segments (`blocks`)
+        and iteratively "denoises" or "refines" the `mask_id` tokens into actual tokens over a series of `steps` for
+        each block. A custom block-diagonal causal attention mask ensures that generation within a block can attend to
+        all previous blocks but not future ones.
+        <Tip warning={true}>
+        This is a specialized generation method. The quality and speed of the output are highly dependent on the interplay
+        between `block_length`, `steps`, and `threshold`. It aims to achieve faster generation through parallel
+        decoding within blocks, which is a departure from the token-by-token generation of standard `.generate()` methods.
+        </Tip>
+        Parameters:
+            inputs (`torch.Tensor`):
+                The token sequence used as a prompt for the generation.
+            temperature (`float`, *optional*, defaults to 0.0):
+                The value used to module the next token probabilities. A value of 0.0 corresponds to greedy decoding.
+            block_length (`int`, *optional*, defaults to 32):
+                The size of each generation block. The model generates text in parallel within these blocks. This is a
+                key parameter for controlling the granularity of the generation process.
+            steps (`int`, *optional*, defaults to 32):
+                The number of iterative refinement (or "denoising") steps to perform for each block. Within each block,
+                the model will try to replace `mask_id` tokens with real tokens for this many iterations.
+            gen_length (`int`, *optional*, defaults to 2048):
+                The maximum number of tokens to generate, excluding the prompt.
+            top_p (`float`, *optional*):
+                If set to a float value between 0 and 1, only the most probable tokens with probabilities that add up to
+                `top_p` or higher are kept for generation (nucleus sampling).
+            top_k (`int`, *optional*):
+                The number of highest probability vocabulary tokens to keep for top-k-filtering.
+            eos_early_stop (`bool`, *optional*, defaults to `False`):
+                If `True`, generation will stop as soon as a valid End-Of-Sequence token is generated and confirmed,
+                even if `gen_length` has not been reached.
+            minimal_topk (`int`, *optional*, defaults to 1):
+                A parameter used to dynamically adjust the number of refinement `steps`. The effective number of steps
+                is capped at `gen_length // minimal_topk`.
+            threshold (`float`, *optional*, defaults to 0.95):
+                The confidence probability threshold for accepting a sampled token. During each refinement step, a
+                sampled token is only kept if its probability is above this threshold. If not enough tokens meet the
+                threshold, the ones with the highest confidence are chosen.
+            eos_id (`int`, *optional*, defaults to 156892):
+                The token ID for the end-of-sequence token. Used for `eos_early_stop`.
+            mask_id (`int`, *optional*, defaults to 156895):
+                The token ID used as a placeholder for tokens that are yet to be generated. This is central to the
+                iterative refinement algorithm.
+        Return:
+            `torch.Tensor`: A string containing the generated token IDs, starting
+            after the prompt and stopping at the first `eos_id` or `gen_length`.
+        """
+        steps = min(steps, gen_length // minimal_topk)
+        input_ids = inputs.to(self.device)
+
+        prompt_length = input_ids.shape[1]
+        num_blocks = (prompt_length + gen_length + block_length - 1) // block_length
+        total_length = num_blocks * block_length
+
+        block_mask = torch.tril(torch.ones(num_blocks, num_blocks, device=self.device))
+        block_diffusion_attention_mask = (
+            (
+                block_mask.repeat_interleave(block_length, dim=0)
+                .repeat_interleave(block_length, dim=1)
+                .unsqueeze(0)
+                .unsqueeze(0)
+            )
+            .log()
+            .to(torch.bfloat16)
+        )
+
+        position_ids = torch.arange(total_length, device=self.device).unsqueeze(0)
+        x = torch.full((1, total_length), mask_id, dtype=torch.long, device=self.device)
+        x[:, :prompt_length] = input_ids.clone()
+
+        prompt_index_full = torch.zeros_like(x, dtype=torch.bool)
+        prompt_index_full[:, :prompt_length] = True
+
+        prefill_blocks = prompt_length // block_length
+
+        denoising_steps_per_block = steps
+        num_transfer_tokens_schedule = self._get_num_transfer_tokens(
+            block_length, denoising_steps_per_block
+        )
+        for num_block in range(prefill_blocks, num_blocks):
+            current_window_end = (num_block + 1) * block_length
+            cur_x = x[:, :current_window_end]
+            cur_attn_mask = block_diffusion_attention_mask[
+                :, :, :current_window_end, :current_window_end
+            ]
+            cur_position_ids = position_ids[:, :current_window_end]
+
+            for step in range(denoising_steps_per_block):
+                active_block_mask = cur_x[:, -block_length:] == mask_id
+                if active_block_mask.sum() == 0:
+                    break
+
+                logits = self.forward(
+                    cur_x,
+                    attention_mask=cur_attn_mask,
+                    position_ids=cur_position_ids,
+                ).logits
+
+                active_logits = logits[:, -block_length:, :]
+                x0, x0_p = self._sample_with_temperature_topk_topp(
+                    active_logits, temperature=temperature, top_k=top_k, top_p=top_p
+                )
+
+                num_to_transfer = num_transfer_tokens_schedule[step].item()
+                transfer_index = torch.zeros_like(x0, dtype=torch.bool)
+
+                confidence = torch.where(active_block_mask, x0_p, -torch.inf)
+                high_conf_mask = confidence[0] > threshold
+                num_high_confidence = high_conf_mask.sum().item()
+
+                if num_high_confidence >= num_to_transfer:
+                    transfer_index[0] = high_conf_mask
+                else:
+                    _, idx = torch.topk(
+                        confidence[0],
+                        k=min(num_to_transfer, active_block_mask.sum().item()),
+                    )
+                    transfer_index[0, idx] = True
+
+                if transfer_index.any():
+                    cur_x[:, -block_length:][transfer_index] = x0[transfer_index]
+                if eos_early_stop and (x0[transfer_index] == eos_id).any():
+                    eos_pos_in_x = (cur_x[0] == eos_id).nonzero(as_tuple=True)
+                    if len(eos_pos_in_x[0]) > 0:
+                        eos_pos = eos_pos_in_x[0][0].item()
+                        if (cur_x[0, prompt_length:eos_pos] != mask_id).all():
+                            final_x = x[:, :total_length][:, : eos_pos + 1]
+                            return final_x
+
+            x[:, :current_window_end] = cur_x
+            if (
+                eos_id is not None
+                and (x[0, prompt_length:current_window_end] == eos_id).any()
+            ):
+                break
+
+        generated_answer = x[:, : prompt_length + gen_length]
+
+        mask_positions = (generated_answer[0][input_ids.shape[1] :] == eos_id).nonzero(
+            as_tuple=True
+        )[0]
+        if len(mask_positions) > 0:
+            first_mask_position = mask_positions[0].item()
+        else:
+            first_mask_position = gen_length
+        return generated_answer[
+            :, input_ids.shape[1] : input_ids.shape[1] + first_mask_position + 1
+        ]
+
+

special_tokens_map.json

@@ -0,0 +1,37 @@
+{
+  "bos_token": {
+    "content": "<|startoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "cls_token": {
+    "content": "[CLS]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "mask_token": {
+    "content": "<|mask|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:878eb900ea4a42da7e54bc18ad38352bec005bbfd81bdf76a5ecf1aad1093a6c
+size 12205801

tokenizer_config.json

@@ -0,0 +1,2115 @@
+{
+  "add_bos_token": false,
+  "add_eos_token": false,
+  "added_tokens_decoder": {
+    "156891": {
+      "content": "<|startoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156892": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156893": {
+      "content": "[CLS]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156894": {
+      "content": "[gMASK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156895": {
+      "content": "<|mask|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156896": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156897": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156898": {
+      "content": "<tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156899": {
+      "content": "</tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156900": {
+      "content": "<|role_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156901": {
+      "content": "<|reserved_token_6|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156902": {
+      "content": "<|reserved_token_7|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156903": {
+      "content": "<|reserved_token_8|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156904": {
+      "content": "<|reserved_token_9|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156905": {
+      "content": "<|reserved_token_10|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156906": {
+      "content": "<|reserved_token_11|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156907": {
+      "content": "<|reserved_token_12|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156908": {
+      "content": "<|reserved_token_13|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156909": {
+      "content": "<|reserved_token_14|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156910": {
+      "content": "<|reserved_token_15|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156911": {
+      "content": "<|reserved_token_16|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156912": {
+      "content": "<|reserved_token_17|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156913": {
+      "content": "<|reserved_token_18|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156914": {
+      "content": "<|reserved_token_19|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156915": {
+      "content": "<|reserved_token_20|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156916": {
+      "content": "<|reserved_token_21|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156917": {
+      "content": "<|reserved_token_22|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156918": {
+      "content": "<|reserved_token_23|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156919": {
+      "content": "<|reserved_token_24|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156920": {
+      "content": "<|reserved_token_25|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156921": {
+      "content": "<|reserved_token_26|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156922": {
+      "content": "<|reserved_token_27|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156923": {
+      "content": "<|reserved_token_28|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156924": {
+      "content": "<|reserved_token_29|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156925": {
+      "content": "<|reserved_token_30|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156926": {
+      "content": "<|reserved_token_31|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156927": {
+      "content": "<|reserved_token_32|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156928": {
+      "content": "<|reserved_token_33|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156929": {
+      "content": "<|reserved_token_34|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156930": {
+      "content": "<|reserved_token_35|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156931": {
+      "content": "<|reserved_token_36|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156932": {
+      "content": "<|reserved_token_37|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156933": {
+      "content": "<|reserved_token_38|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156934": {
+      "content": "<|reserved_token_39|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156935": {
+      "content": "<|reserved_token_40|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156936": {
+      "content": "<|reserved_token_41|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156937": {
+      "content": "<|reserved_token_42|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156938": {
+      "content": "<|reserved_token_43|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156939": {
+      "content": "<|reserved_token_44|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156940": {
+      "content": "<|reserved_token_45|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156941": {
+      "content": "<|reserved_token_46|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156942": {
+      "content": "<|reserved_token_47|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156943": {
+      "content": "<|reserved_token_48|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156944": {
+      "content": "<|reserved_token_49|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156945": {
+      "content": "<|reserved_token_50|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156946": {
+      "content": "<|reserved_token_51|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156947": {
+      "content": "<|reserved_token_52|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156948": {
+      "content": "<|reserved_token_53|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156949": {
+      "content": "<|reserved_token_54|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156950": {
+      "content": "<|reserved_token_55|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156951": {
+      "content": "<|reserved_token_56|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156952": {
+      "content": "<|reserved_token_57|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156953": {
+      "content": "<|reserved_token_58|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156954": {
+      "content": "<|reserved_token_59|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156955": {
+      "content": "<|reserved_token_60|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156956": {
+      "content": "<|reserved_token_61|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156957": {
+      "content": "<|reserved_token_62|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156958": {
+      "content": "<|reserved_token_63|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156959": {
+      "content": "<|reserved_token_64|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156960": {
+      "content": "<|reserved_token_65|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156961": {
+      "content": "<|reserved_token_66|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156962": {
+      "content": "<|reserved_token_67|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156963": {
+      "content": "<|reserved_token_68|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156964": {
+      "content": "<|reserved_token_69|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156965": {
+      "content": "<|reserved_token_70|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156966": {
+      "content": "<|reserved_token_71|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156967": {
+      "content": "<|reserved_token_72|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156968": {
+      "content": "<|reserved_token_73|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156969": {
+      "content": "<|reserved_token_74|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156970": {
+      "content": "<|reserved_token_75|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156971": {
+      "content": "<|reserved_token_76|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156972": {
+      "content": "<|reserved_token_77|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156973": {
+      "content": "<|reserved_token_78|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156974": {
+      "content": "<|reserved_token_79|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156975": {
+      "content": "<|reserved_token_80|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156976": {
+      "content": "<|reserved_token_81|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156977": {
+      "content": "<|reserved_token_82|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156978": {
+      "content": "<|reserved_token_83|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156979": {
+      "content": "<|reserved_token_84|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156980": {
+      "content": "<|reserved_token_85|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156981": {
+      "content": "<|reserved_token_86|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156982": {
+      "content": "<|reserved_token_87|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156983": {
+      "content": "<|reserved_token_88|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156984": {
+      "content": "<|reserved_token_89|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156985": {
+      "content": "<|reserved_token_90|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156986": {
+      "content": "<|reserved_token_91|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156987": {
+      "content": "<|reserved_token_92|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156988": {
+      "content": "<|reserved_token_93|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156989": {
+      "content": "<|reserved_token_94|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156990": {
+      "content": "<|reserved_token_95|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156991": {
+      "content": "<|reserved_token_96|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156992": {
+      "content": "<|reserved_token_97|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156993": {
+      "content": "<|reserved_token_98|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156994": {
+      "content": "<|reserved_token_99|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156995": {
+      "content": "<|reserved_token_100|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156996": {
+      "content": "<|reserved_token_101|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156997": {
+      "content": "<|reserved_token_102|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156998": {
+      "content": "<|reserved_token_103|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156999": {
+      "content": "<|reserved_token_104|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157000": {
+      "content": "<|reserved_token_105|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157001": {
+      "content": "<|reserved_token_106|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157002": {
+      "content": "<|reserved_token_107|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157003": {
+      "content": "<|reserved_token_108|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157004": {
+      "content": "<|reserved_token_109|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157005": {
+      "content": "<|reserved_token_110|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157006": {
+      "content": "<|reserved_token_111|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157007": {
+      "content": "<|reserved_token_112|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157008": {
+      "content": "<|reserved_token_113|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157009": {
+      "content": "<|reserved_token_114|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157010": {
+      "content": "<|reserved_token_115|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157011": {
+      "content": "<|reserved_token_116|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157012": {
+      "content": "<|reserved_token_117|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157013": {
+      "content": "<|reserved_token_118|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157014": {
+      "content": "<|reserved_token_119|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157015": {
+      "content": "<|reserved_token_120|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157016": {
+      "content": "<|reserved_token_121|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157017": {
+      "content": "<|reserved_token_122|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157018": {
+      "content": "<|reserved_token_123|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157019": {
+      "content": "<|reserved_token_124|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157020": {
+      "content": "<|reserved_token_125|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157021": {
+      "content": "<|reserved_token_126|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157022": {
+      "content": "<|reserved_token_127|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157023": {
+      "content": "<|reserved_token_128|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157024": {
+      "content": "<|reserved_token_129|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157025": {
+      "content": "<|reserved_token_130|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157026": {
+      "content": "<|reserved_token_131|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157027": {
+      "content": "<|reserved_token_132|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157028": {
+      "content": "<|reserved_token_133|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157029": {
+      "content": "<|reserved_token_134|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157030": {
+      "content": "<|reserved_token_135|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157031": {
+      "content": "<|reserved_token_136|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157032": {
+      "content": "<|reserved_token_137|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157033": {
+      "content": "<|reserved_token_138|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157034": {
+      "content": "<|reserved_token_139|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157035": {
+      "content": "<|reserved_token_140|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157036": {
+      "content": "<|reserved_token_141|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157037": {
+      "content": "<|reserved_token_142|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157038": {
+      "content": "<|reserved_token_143|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157039": {
+      "content": "<|reserved_token_144|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157040": {
+      "content": "<|reserved_token_145|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157041": {
+      "content": "<|reserved_token_146|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157042": {
+      "content": "<|reserved_token_147|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157043": {
+      "content": "<|reserved_token_148|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157044": {
+      "content": "<|reserved_token_149|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157045": {
+      "content": "<|reserved_token_150|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157046": {
+      "content": "<|reserved_token_151|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157047": {
+      "content": "<|reserved_token_152|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157048": {
+      "content": "<|reserved_token_153|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157049": {
+      "content": "<|reserved_token_154|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157050": {
+      "content": "<|reserved_token_155|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157051": {
+      "content": "<|reserved_token_156|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157052": {
+      "content": "<|reserved_token_157|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157053": {
+      "content": "<|reserved_token_158|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157054": {
+      "content": "<|reserved_token_159|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157055": {
+      "content": "<|reserved_token_160|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157056": {
+      "content": "<|reserved_token_161|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157057": {
+      "content": "<|reserved_token_162|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157058": {
+      "content": "<|reserved_token_163|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157059": {
+      "content": "<|reserved_token_164|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157060": {
+      "content": "<|reserved_token_165|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157061": {
+      "content": "<|reserved_token_166|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157062": {
+      "content": "<|reserved_token_167|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157063": {
+      "content": "<|reserved_token_168|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157064": {
+      "content": "<|reserved_token_169|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157065": {
+      "content": "<|reserved_token_170|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157066": {
+      "content": "<|reserved_token_171|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157067": {
+      "content": "<|reserved_token_172|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157068": {
+      "content": "<|reserved_token_173|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157069": {
+      "content": "<|reserved_token_174|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157070": {
+      "content": "<|reserved_token_175|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157071": {
+      "content": "<|reserved_token_176|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157072": {
+      "content": "<|reserved_token_177|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157073": {
+      "content": "<|reserved_token_178|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157074": {
+      "content": "<|reserved_token_179|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157075": {
+      "content": "<|reserved_token_180|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157076": {
+      "content": "<|reserved_token_181|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157077": {
+      "content": "<|reserved_token_182|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157078": {
+      "content": "<|reserved_token_183|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157079": {
+      "content": "<|reserved_token_184|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157080": {
+      "content": "<|reserved_token_185|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157081": {
+      "content": "<|reserved_token_186|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157082": {
+      "content": "<|reserved_token_187|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157083": {
+      "content": "<|reserved_token_188|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157084": {
+      "content": "<|reserved_token_189|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157085": {
+      "content": "<|reserved_token_190|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157086": {
+      "content": "<|reserved_token_191|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157087": {
+      "content": "<|reserved_token_192|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157088": {
+      "content": "<|reserved_token_193|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157089": {
+      "content": "<|reserved_token_194|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157090": {
+      "content": "<|reserved_token_195|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157091": {
+      "content": "<|reserved_token_196|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157092": {
+      "content": "<|reserved_token_197|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157093": {
+      "content": "<|reserved_token_198|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157094": {
+      "content": "<|reserved_token_199|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157095": {
+      "content": "<|reserved_token_200|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157096": {
+      "content": "<|reserved_token_201|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157097": {
+      "content": "<|reserved_token_202|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157098": {
+      "content": "<|reserved_token_203|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157099": {
+      "content": "<|reserved_token_204|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157100": {
+      "content": "<|reserved_token_205|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157101": {
+      "content": "<|reserved_token_206|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157102": {
+      "content": "<|reserved_token_207|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157103": {
+      "content": "<|reserved_token_208|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157104": {
+      "content": "<|reserved_token_209|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157105": {
+      "content": "<|reserved_token_210|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157106": {
+      "content": "<|reserved_token_211|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157107": {
+      "content": "<|reserved_token_212|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157108": {
+      "content": "<|reserved_token_213|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157109": {
+      "content": "<|reserved_token_214|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157110": {
+      "content": "<|reserved_token_215|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157111": {
+      "content": "<|reserved_token_216|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157112": {
+      "content": "<|reserved_token_217|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157113": {
+      "content": "<|reserved_token_218|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157114": {
+      "content": "<|reserved_token_219|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157115": {
+      "content": "<|reserved_token_220|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157116": {
+      "content": "<|reserved_token_221|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157117": {
+      "content": "<|reserved_token_222|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157118": {
+      "content": "<|reserved_token_223|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157119": {
+      "content": "<|reserved_token_224|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157120": {
+      "content": "<|reserved_token_225|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157121": {
+      "content": "<|reserved_token_226|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157122": {
+      "content": "<|reserved_token_227|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157123": {
+      "content": "<|reserved_token_228|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157124": {
+      "content": "<|reserved_token_229|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157125": {
+      "content": "<|reserved_token_230|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157126": {
+      "content": "<|reserved_token_231|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157127": {
+      "content": "<|reserved_token_232|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157128": {
+      "content": "<|reserved_token_233|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157129": {
+      "content": "<|reserved_token_234|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157130": {
+      "content": "<|reserved_token_235|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157131": {
+      "content": "<|reserved_token_236|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157132": {
+      "content": "<|reserved_token_237|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157133": {
+      "content": "<|reserved_token_238|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157134": {
+      "content": "<|reserved_token_239|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157135": {
+      "content": "<|reserved_token_240|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157136": {
+      "content": "<|reserved_token_241|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157137": {
+      "content": "<|reserved_token_242|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157138": {
+      "content": "<|reserved_token_243|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157139": {
+      "content": "<|reserved_token_244|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157140": {
+      "content": "<|reserved_token_245|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157141": {
+      "content": "<|reserved_token_246|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157142": {
+      "content": "<|reserved_token_247|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157143": {
+      "content": "<|reserved_token_248|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157144": {
+      "content": "<|reserved_token_249|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157145": {
+      "content": "<|reserved_token_250|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157146": {
+      "content": "<|reserved_token_251|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157147": {
+      "content": "<|reserved_token_252|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157148": {
+      "content": "<|reserved_token_253|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157149": {
+      "content": "<|reserved_token_254|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157150": {
+      "content": "<|reserved_token_255|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157151": {
+      "content": "<role>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157152": {
+      "content": "</role>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<|startoftext|>",
+  "clean_up_tokenization_spaces": false,
+  "cls_token": "[CLS]",
+  "eos_token": "<|endoftext|>",
+  "extra_special_tokens": {},
+  "fast_tokenizer": true,
+  "gmask_token": "[gMASK]",
+  "mask_token": "<|mask|>",
+  "merges_file": null,
+  "model_max_length": 32768,
+  "pad_token": "<|endoftext|>",
+  "tokenizer_class": "PreTrainedTokenizerFast",
+  "trust_remote_code": true
+}