Upload folder using huggingface_hub

.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 +1,199 @@
+---
+library_name: transformers
+tags: []
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]

added_tokens.json

@@ -0,0 +1,28 @@
+{
+  "</think>": 151668,
+  "</tool_call>": 151658,
+  "</tool_response>": 151666,
+  "<think>": 151667,
+  "<tool_call>": 151657,
+  "<tool_response>": 151665,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}

chat_template.jinja

@@ -0,0 +1,130 @@
+{% macro render_extra_keys(json_dict, handled_keys) %}
+    {%- if json_dict is mapping %}
+        {%- for json_key in json_dict if json_key not in handled_keys %}
+            {%- if json_dict[json_key] is mapping or (json_dict[json_key] is sequence and json_dict[json_key] is not string) %}
+                {{- '\n<' ~ json_key ~ '>' ~ (json_dict[json_key] | tojson | safe) ~ '</' ~ json_key ~ '>' }}
+            {%- else %}
+                {{-'\n<' ~ json_key ~ '>' ~ (json_dict[json_key] | string) ~ '</' ~ json_key ~ '>' }}
+            {%- endif %}
+        {%- endfor %}
+    {%- endif %}
+{% endmacro %}
+
+{%- if messages[0]["role"] == "system" %}
+    {%- set system_message = messages[0]["content"] %}
+    {%- set loop_messages = messages[1:] %}
+{%- else %}
+    {%- set loop_messages = messages %}
+{%- endif %}
+
+{%- if not tools is defined %}
+    {%- set tools = [] %}
+{%- endif %}
+
+{%- if system_message is defined %}
+    {{- "<|im_start|>system\n" + system_message }}
+{%- else %}
+    {%- if tools is iterable and tools | length > 0 %}
+        {{- "<|im_start|>system\nYou are Qwen, a helpful AI assistant that can interact with a computer to solve tasks." }}
+    {%- endif %}
+{%- endif %}
+{%- if tools is iterable and tools | length > 0 %}
+    {{- "\n\n# Tools\n\nYou have access to the following functions:\n\n" }}
+    {{- "<tools>" }}
+    {%- for tool in tools %}
+        {%- if tool.function is defined %}
+            {%- set tool = tool.function %}
+        {%- endif %}
+        {{- "\n<function>\n<name>" ~ tool.name ~ "</name>" }}
+        {%- if tool.description is defined %}
+            {{- '\n<description>' ~ (tool.description | trim) ~ '</description>' }}
+        {%- endif %}
+        {{- '\n<parameters>' }}
+        {%- if tool.parameters is defined and tool.parameters is mapping and tool.parameters.properties is defined and tool.parameters.properties is mapping %}
+            {%- for param_name, param_fields in tool.parameters.properties|items %}
+                {{- '\n<parameter>' }}
+                {{- '\n<name>' ~ param_name ~ '</name>' }}
+                {%- if param_fields.type is defined %}
+                    {{- '\n<type>' ~ (param_fields.type | string) ~ '</type>' }}
+                {%- endif %}
+                {%- if param_fields.description is defined %}
+                    {{- '\n<description>' ~ (param_fields.description | trim) ~ '</description>' }}
+                {%- endif %}
+                {%- set handled_keys = ['name', 'type', 'description'] %}
+                {{- render_extra_keys(param_fields, handled_keys) }}
+                {{- '\n</parameter>' }}
+            {%- endfor %}
+        {%- endif %}
+        {% set handled_keys = ['type', 'properties'] %}
+        {{- render_extra_keys(tool.parameters, handled_keys) }}
+        {{- '\n</parameters>' }}
+        {%- set handled_keys = ['type', 'name', 'description', 'parameters'] %}
+        {{- render_extra_keys(tool, handled_keys) }}
+        {{- '\n</function>' }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+{%- endif %}
+{%- if system_message is defined %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if tools is iterable and tools | length > 0 %}
+        {{- '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- for message in loop_messages %}
+    {%- if message.role == "assistant" and message.tool_calls is defined and message.tool_calls is iterable and message.tool_calls | length > 0 %}
+        {{- '<|im_start|>' + message.role }}
+        {%- if message.content is defined and message.content is string and message.content | trim | length > 0 %}
+            {{- '\n' + message.content | trim + '\n' }}
+        {%- endif %}
+        {%- for tool_call in message.tool_calls %}
+            {%- if tool_call.function is defined %}
+                {%- set tool_call = tool_call.function %}
+            {%- endif %}
+            {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+            {%- if tool_call.arguments is defined %}
+                {%- for args_name, args_value in tool_call.arguments|items %}
+                    {{- '<parameter=' + args_name + '>\n' }}
+                    {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                    {{- args_value }}
+                    {{- '\n</parameter>\n' }}
+                {%- endfor %}
+            {%- endif %}
+            {{- '</function>\n</tool_call>' }}
+        {%- endfor %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "user" or message.role == "system" or message.role == "assistant" %}
+        {{- '<|im_start|>' + message.role }}
+        {%- if message.role == "assistant" and message.reasoning_content is defined %}
+            {%- if message.reasoning_content -%}
+                {{ '\n<think>' + message.reasoning_content.strip() + '</think>' }}
+            {%- else -%}
+                {{ '\n<think></think>' }}
+            {%- endif -%}
+            {%- if message.content.strip() -%}
+                {{ '\n' + message.content.strip() }}
+            {%- endif -%}
+        {%- else %}
+            {{- '\n' + message.content }}
+        {%- endif %}
+        {{- '<|im_end|>' + '\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user\n' }}
+        {%- endif %}
+        {{- '<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>\n' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+{%- endif %}

config.json

@@ -0,0 +1,43 @@
+{
+  "architectures": [
+    "Qwen3MoeForCausalLM"
+  ],
+  "auto_map": {
+    "AutoConfig": "configuration_qwen3_moe.Qwen3MoeConfig",
+    "AutoModelForCausalLM": "modeling_qwen3_moe.Qwen3MoeForCausalLM",
+    "AutoModel": "modeling_qwen3_moe.Qwen3MoeModel"
+  },
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "decoder_sparse_step": 1,
+  "eos_token_id": 151645,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 2048,
+  "initializer_range": 0.02,
+  "intermediate_size": 6144,
+  "max_position_embeddings": 40960,
+  "max_window_layers": 48,
+  "mlp_only_layers": [],
+  "model_type": "qwen3_moe",
+  "moe_intermediate_size": 768,
+  "norm_topk_prob": true,
+  "num_attention_heads": 32,
+  "num_experts": 128,
+  "num_experts_per_tok": 8,
+  "num_hidden_layers": 48,
+  "num_key_value_heads": 4,
+  "output_router_logits": false,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 1000000.0,
+  "router_aux_loss_coef": 0.001,
+  "sliding_window": null,
+  "tie_word_embeddings": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.51.0",
+  "use_cache": true,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}

configuration_qwen3_moe.py

@@ -0,0 +1,243 @@
+# coding=utf-8
+# Copyright 2024 The Qwen team, Alibaba Group and the HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""Qwen3MoE model configuration"""
+
+from transformers.configuration_utils import PretrainedConfig
+from transformers.modeling_rope_utils import rope_config_validation
+from transformers.utils import logging
+
+
+logger = logging.get_logger(__name__)
+
+
+class Qwen3MoeConfig(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`Qwen3MoeModel`]. It is used to instantiate a
+    Qwen3MoE model according to the specified arguments, defining the model architecture. Instantiating a configuration
+    with the defaults will yield a similar configuration to that of [Qwen/Qwen3-15B-A2B](https://huggingface.co/Qwen/Qwen3-15B-A2B).
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+
+    Args:
+        vocab_size (`int`, *optional*, defaults to 151936):
+            Vocabulary size of the Qwen3MoE model. Defines the number of different tokens that can be represented by the
+            `inputs_ids` passed when calling [`Qwen3MoeModel`]
+        hidden_size (`int`, *optional*, defaults to 2048):
+            Dimension of the hidden representations.
+        intermediate_size (`int`, *optional*, defaults to 6144):
+            Dimension of the MLP representations.
+        num_hidden_layers (`int`, *optional*, defaults to 24):
+            Number of hidden layers in the Transformer encoder.
+        num_attention_heads (`int`, *optional*, defaults to 32):
+            Number of attention heads for each attention layer in the Transformer encoder.
+        num_key_value_heads (`int`, *optional*, defaults to 4):
+            This is the number of key_value heads that should be used to implement Grouped Query Attention. If
+            `num_key_value_heads=num_attention_heads`, the model will use Multi Head Attention (MHA), if
+            `num_key_value_heads=1` the model will use Multi Query Attention (MQA) otherwise GQA is used. When
+            converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed
+            by meanpooling all the original heads within that group. For more details, check out [this
+            paper](https://huggingface.co/papers/2305.13245). If it is not specified, will default to `32`.
+
+        hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
+            The non-linear activation function (function or string) in the decoder.
+        max_position_embeddings (`int`, *optional*, defaults to 32768):
+            The maximum sequence length that this model might ever be used with.
+        initializer_range (`float`, *optional*, defaults to 0.02):
+            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
+        rms_norm_eps (`float`, *optional*, defaults to 1e-06):
+            The epsilon used by the rms normalization layers.
+        use_cache (`bool`, *optional*, defaults to `True`):
+            Whether or not the model should return the last key/values attentions (not used by all models). Only
+            relevant if `config.is_decoder=True`.
+        tie_word_embeddings (`bool`, *optional*, defaults to `False`):
+            Whether the model's input and output word embeddings should be tied.
+        rope_theta (`float`, *optional*, defaults to 10000.0):
+            The base period of the RoPE embeddings.
+        rope_scaling (`Dict`, *optional*):
+            Dictionary containing the scaling configuration for the RoPE embeddings. NOTE: if you apply new rope type
+            and you expect the model to work on longer `max_position_embeddings`, we recommend you to update this value
+            accordingly.
+            Expected contents:
+                `rope_type` (`str`):
+                    The sub-variant of RoPE to use. Can be one of ['default', 'linear', 'dynamic', 'yarn', 'longrope',
+                    'llama3'], with 'default' being the original RoPE implementation.
+                `factor` (`float`, *optional*):
+                    Used with all rope types except 'default'. The scaling factor to apply to the RoPE embeddings. In
+                    most scaling types, a `factor` of x will enable the model to handle sequences of length x *
+                    original maximum pre-trained length.
+                `original_max_position_embeddings` (`int`, *optional*):
+                    Used with 'dynamic', 'longrope' and 'llama3'. The original max position embeddings used during
+                    pretraining.
+                `attention_factor` (`float`, *optional*):
+                    Used with 'yarn' and 'longrope'. The scaling factor to be applied on the attention
+                    computation. If unspecified, it defaults to value recommended by the implementation, using the
+                    `factor` field to infer the suggested value.
+                `beta_fast` (`float`, *optional*):
+                    Only used with 'yarn'. Parameter to set the boundary for extrapolation (only) in the linear
+                    ramp function. If unspecified, it defaults to 32.
+                `beta_slow` (`float`, *optional*):
+                    Only used with 'yarn'. Parameter to set the boundary for interpolation (only) in the linear
+                    ramp function. If unspecified, it defaults to 1.
+                `short_factor` (`list[float]`, *optional*):
+                    Only used with 'longrope'. The scaling factor to be applied to short contexts (<
+                    `original_max_position_embeddings`). Must be a list of numbers with the same length as the hidden
+                    size divided by the number of attention heads divided by 2
+                `long_factor` (`list[float]`, *optional*):
+                    Only used with 'longrope'. The scaling factor to be applied to long contexts (<
+                    `original_max_position_embeddings`). Must be a list of numbers with the same length as the hidden
+                    size divided by the number of attention heads divided by 2
+                `low_freq_factor` (`float`, *optional*):
+                    Only used with 'llama3'. Scaling factor applied to low frequency components of the RoPE
+                `high_freq_factor` (`float`, *optional*):
+                    Only used with 'llama3'. Scaling factor applied to high frequency components of the RoPE
+        attention_bias (`bool`, defaults to `False`, *optional*, defaults to `False`):
+            Whether to use a bias in the query, key, value and output projection layers during self-attention.
+        use_sliding_window (`bool`, *optional*, defaults to `False`):
+            Whether to use sliding window attention.
+        sliding_window (`int`, *optional*, defaults to 4096):
+            Sliding window attention (SWA) window size. If not specified, will default to `4096`.
+        attention_dropout (`float`, *optional*, defaults to 0.0):
+            The dropout ratio for the attention probabilities.
+        decoder_sparse_step (`int`, *optional*, defaults to 1):
+            The frequency of the MoE layer.
+        moe_intermediate_size (`int`, *optional*, defaults to 768):
+            Intermediate size of the routed expert.
+        num_experts_per_tok (`int`, *optional*, defaults to 8):
+            Number of selected experts.
+        num_experts (`int`, *optional*, defaults to 128):
+            Number of routed experts.
+        norm_topk_prob (`bool`, *optional*, defaults to `False`):
+            Whether to normalize the topk probabilities.
+        output_router_logits (`bool`, *optional*, defaults to `False`):
+            Whether or not the router logits should be returned by the model. Enabling this will also
+            allow the model to output the auxiliary loss, including load balancing loss and router z-loss.
+        router_aux_loss_coef (`float`, *optional*, defaults to 0.001):
+            The aux loss factor for the total loss.
+        mlp_only_layers (`list[int]`, *optional*, defaults to `[]`):
+            Indicate which layers use Qwen3MoeMLP rather than Qwen3MoeSparseMoeBlock
+            The list contains layer index, from 0 to num_layers-1 if we have num_layers layers
+            If `mlp_only_layers` is empty, `decoder_sparse_step` is used to determine the sparsity.
+
+    ```python
+    >>> from transformers import Qwen3MoeModel, Qwen3MoeConfig
+
+    >>> # Initializing a Qwen3MoE style configuration
+    >>> configuration = Qwen3MoeConfig()
+
+    >>> # Initializing a model from the Qwen3-15B-A2B" style configuration
+    >>> model = Qwen3MoeModel(configuration)
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+
+    model_type = "qwen3_moe"
+    keys_to_ignore_at_inference = ["past_key_values"]
+
+    # Default tensor parallel plan for base model `Qwen3Moe`
+    base_model_tp_plan = {
+        "layers.*.self_attn.q_proj": "colwise",
+        "layers.*.self_attn.k_proj": "colwise",
+        "layers.*.self_attn.v_proj": "colwise",
+        "layers.*.self_attn.o_proj": "rowwise",
+        "layers.*.mlp.experts.*.gate_proj": "colwise",
+        "layers.*.mlp.experts.*.up_proj": "colwise",
+        "layers.*.mlp.experts.*.down_proj": "rowwise",
+        "layers.*.mlp.gate_proj": "colwise",
+        "layers.*.mlp.up_proj": "colwise",
+        "layers.*.mlp.down_proj": "rowwise",
+    }
+    base_model_pp_plan = {
+        "embed_tokens": (["input_ids"], ["inputs_embeds"]),
+        "layers": (["hidden_states", "attention_mask"], ["hidden_states"]),
+        "norm": (["hidden_states"], ["hidden_states"]),
+    }
+
+    def __init__(
+        self,
+        vocab_size=151936,
+        hidden_size=2048,
+        intermediate_size=6144,
+        num_hidden_layers=24,
+        num_attention_heads=32,
+        num_key_value_heads=4,
+        hidden_act="silu",
+        max_position_embeddings=32768,
+        initializer_range=0.02,
+        rms_norm_eps=1e-6,
+        use_cache=True,
+        tie_word_embeddings=False,
+        rope_theta=10000.0,
+        rope_scaling=None,
+        attention_bias=False,
+        use_sliding_window=False,
+        sliding_window=4096,
+        attention_dropout=0.0,
+        decoder_sparse_step=1,
+        moe_intermediate_size=768,
+        num_experts_per_tok=8,
+        num_experts=128,
+        norm_topk_prob=False,
+        output_router_logits=False,
+        router_aux_loss_coef=0.001,
+        mlp_only_layers=None,
+        load_balance_coeff=1e-3,
+        **kwargs,
+    ):
+        self.vocab_size = vocab_size
+        self.max_position_embeddings = max_position_embeddings
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.use_sliding_window = use_sliding_window
+        self.sliding_window = sliding_window if use_sliding_window else None
+
+        self.num_key_value_heads = num_key_value_heads
+        self.hidden_act = hidden_act
+        self.initializer_range = initializer_range
+        self.rms_norm_eps = rms_norm_eps
+        self.use_cache = use_cache
+        self.rope_theta = rope_theta
+        self.rope_scaling = rope_scaling
+        self.attention_bias = attention_bias
+        self.attention_dropout = attention_dropout
+        # Validate the correctness of rotary position embeddings parameters
+        # BC: if there is a 'type' field, move it to 'rope_type'.
+        if self.rope_scaling is not None and "type" in self.rope_scaling:
+            self.rope_scaling["rope_type"] = self.rope_scaling["type"]
+        rope_config_validation(self)
+
+        # MoE arguments
+        self.decoder_sparse_step = decoder_sparse_step
+        self.moe_intermediate_size = moe_intermediate_size
+        self.num_experts_per_tok = num_experts_per_tok
+        self.num_experts = num_experts
+        self.norm_topk_prob = norm_topk_prob
+        self.output_router_logits = output_router_logits
+        self.router_aux_loss_coef = router_aux_loss_coef
+        self.mlp_only_layers = [] if mlp_only_layers is None else mlp_only_layers
+        self.load_balance_coeff = load_balance_coeff
+
+        super().__init__(
+            tie_word_embeddings=tie_word_embeddings,
+            **kwargs,
+        )
+
+
+__all__ = ["Qwen3MoeConfig"]
+

generation_config.json

@@ -0,0 +1,13 @@
+{
+  "bos_token_id": 151643,
+  "do_sample": true,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "pad_token_id": 151643,
+  "temperature": 0.6,
+  "top_k": 20,
+  "top_p": 0.95,
+  "transformers_version": "4.53.0"
+}

model-00001-of-00017.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cab9f07033dcda74dda82884150b1e948cbe637f37ffdcd3a2f2205a66157309
+size 1244664136

model-00002-of-00017.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0d48bb623e635950c267cbebc7c8afbcd275e5ff4e12bc5a97069508d06defc8
+size 3738727960

model-00003-of-00017.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57ed41214f9ab6b6d2afaa925571f06db521e721495d3d264056502cdff3c493
+size 3738727960

model-00004-of-00017.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:93df7e6c3a4afb30fb0930f44c4f6708c5b3dcd94a591d1b351c2aaaa5df6c0d
+size 3738727960

model-00005-of-00017.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f6d9ec4c54bb8a7ac83e85c815d0811d842b149a988eae7f29a08bacd60a71c4
+size 3738727984

model-00006-of-00017.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:acb7ecc29778fa2dd370ebcffed0dee249866d2ea36f11215a410ba85fa8aa9b
+size 3738727992

model-00007-of-00017.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec033a07a762a24c23500110817f0b589a0db2526df2d90d91f8dfb825be911c
+size 3738727992

model-00008-of-00017.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7624d0eba94bb1255f382c0adf2091d3683777c23c79b091d102b7808329cb43
+size 3738727992

model-00009-of-00017.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c562a2b0c71dcdfa9f6ca6e5bdf4321814ba0055093cc90a9d34904ae993b35
+size 3738727992

model-00010-of-00017.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:953b62a923c82023cf675e644266376c43c2b089f31d08809fa780d3ee4b8719
+size 3738727992

model-00011-of-00017.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:26d0b5cefc55533a0f91e71d90e046d2fd3ad1a4a10bd6009257d08ee0f5285d
+size 3738727992

model-00012-of-00017.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd036c7d9a50485a61afef1c28f91033e67613b27b7b97955bd89f14e11c634e
+size 3738727992

model-00013-of-00017.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90d44b839e3c7cc4d38a02c0ece6958027596d125507d6d311ab1b7dfcd1ee74
+size 3738727992

model-00014-of-00017.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:36bba42c89a95001e28caa75c5b578197139d6f220b19d731e343adc600533f9
+size 3738727992

model-00015-of-00017.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:15c63236cd02eff65295d45bcc9aa32d823186773e03561ddd1b92565af32d84
+size 3738727992

model-00016-of-00017.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6fae2ec67887d1b7f55b6941d20aeeebc0888fe9b1f4b6f75c7f582983377ee7
+size 3738727992

model-00017-of-00017.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ef0d58901646d5eeb86b04958a48d7b5bf802d2d5dd90430fd47f71cb3387d7
+size 3738727992

model.safetensors.index.json

@@ -0,0 +1 @@
+{"metadata": {"total_size": 61064245248}, "weight_map": {"lm_head.weight": "model-00001-of-00017.safetensors", "model.embed_tokens.weight": "model-00001-of-00017.safetensors", "model.norm.weight": "model-00001-of-00017.safetensors", "model.layers.0.input_layernorm.weight": "model-00002-of-00017.safetensors", "model.layers.0.post_attention_layernorm.weight": "model-00002-of-00017.safetensors", "model.layers.0.self_attn.k_norm.weight": "model-00002-of-00017.safetensors", "model.layers.0.self_attn.k_proj.weight": "model-00002-of-00017.safetensors", "model.layers.0.self_attn.o_proj.weight": "model-00002-of-00017.safetensors", "model.layers.0.self_attn.q_norm.weight": "model-00002-of-00017.safetensors", "model.layers.0.self_attn.q_proj.weight": "model-00002-of-00017.safetensors", "model.layers.0.self_attn.v_proj.weight": "model-00002-of-00017.safetensors", "model.layers.0.mlp.router.gate.weight": "model-00002-of-00017.safetensors", "model.layers.0.mlp.experts.w1": "model-00002-of-00017.safetensors", "model.layers.0.mlp.experts.w2": "model-00002-of-00017.safetensors", "model.layers.0.mlp.experts.w3": "model-00002-of-00017.safetensors", "model.layers.1.input_layernorm.weight": "model-00002-of-00017.safetensors", "model.layers.1.post_attention_layernorm.weight": "model-00002-of-00017.safetensors", "model.layers.1.self_attn.k_norm.weight": "model-00002-of-00017.safetensors", "model.layers.1.self_attn.k_proj.weight": "model-00002-of-00017.safetensors", "model.layers.1.self_attn.o_proj.weight": "model-00002-of-00017.safetensors", "model.layers.1.self_attn.q_norm.weight": "model-00002-of-00017.safetensors", "model.layers.1.self_attn.q_proj.weight": "model-00002-of-00017.safetensors", "model.layers.1.self_attn.v_proj.weight": "model-00002-of-00017.safetensors", "model.layers.19.input_layernorm.weight": "model-00008-of-00017.safetensors", "model.layers.19.post_attention_layernorm.weight": "model-00008-of-00017.safetensors", "model.layers.10.input_layernorm.weight": "model-00005-of-00017.safetensors", "model.layers.10.post_attention_layernorm.weight": "model-00005-of-00017.safetensors", "model.layers.10.self_attn.k_norm.weight": "model-00005-of-00017.safetensors", "model.layers.10.self_attn.k_proj.weight": "model-00005-of-00017.safetensors", "model.layers.10.self_attn.o_proj.weight": "model-00005-of-00017.safetensors", "model.layers.10.self_attn.q_norm.weight": "model-00005-of-00017.safetensors", "model.layers.10.self_attn.q_proj.weight": "model-00005-of-00017.safetensors", "model.layers.10.self_attn.v_proj.weight": "model-00005-of-00017.safetensors", "model.layers.11.self_attn.k_norm.weight": "model-00005-of-00017.safetensors", "model.layers.11.self_attn.k_proj.weight": "model-00005-of-00017.safetensors", "model.layers.11.self_attn.o_proj.weight": "model-00005-of-00017.safetensors", "model.layers.11.self_attn.q_norm.weight": "model-00005-of-00017.safetensors", "model.layers.11.self_attn.q_proj.weight": "model-00005-of-00017.safetensors", "model.layers.11.self_attn.v_proj.weight": "model-00005-of-00017.safetensors", "model.layers.11.input_layernorm.weight": "model-00005-of-00017.safetensors", "model.layers.11.post_attention_layernorm.weight": "model-00005-of-00017.safetensors", "model.layers.12.input_layernorm.weight": "model-00006-of-00017.safetensors", "model.layers.12.post_attention_layernorm.weight": "model-00006-of-00017.safetensors", "model.layers.12.self_attn.k_norm.weight": "model-00006-of-00017.safetensors", "model.layers.12.self_attn.k_proj.weight": "model-00006-of-00017.safetensors", "model.layers.12.self_attn.o_proj.weight": "model-00006-of-00017.safetensors", "model.layers.12.self_attn.q_norm.weight": "model-00006-of-00017.safetensors", "model.layers.12.self_attn.q_proj.weight": "model-00006-of-00017.safetensors", "model.layers.12.self_attn.v_proj.weight": "model-00006-of-00017.safetensors", "model.layers.13.input_layernorm.weight": "model-00006-of-00017.safetensors", "model.layers.13.post_attention_layernorm.weight": "model-00006-of-00017.safetensors", "model.layers.13.self_attn.k_norm.weight": "model-00006-of-00017.safetensors", "model.layers.13.self_attn.k_proj.weight": "model-00006-of-00017.safetensors", "model.layers.13.self_attn.o_proj.weight": "model-00006-of-00017.safetensors", "model.layers.13.self_attn.q_norm.weight": "model-00006-of-00017.safetensors", "model.layers.13.self_attn.q_proj.weight": "model-00006-of-00017.safetensors", "model.layers.13.self_attn.v_proj.weight": "model-00006-of-00017.safetensors", "model.layers.14.input_layernorm.weight": "model-00006-of-00017.safetensors", "model.layers.14.post_attention_layernorm.weight": "model-00006-of-00017.safetensors", "model.layers.14.self_attn.k_norm.weight": "model-00006-of-00017.safetensors", "model.layers.14.self_attn.k_proj.weight": "model-00006-of-00017.safetensors", "model.layers.14.self_attn.o_proj.weight": "model-00006-of-00017.safetensors", "model.layers.14.self_attn.q_norm.weight": "model-00006-of-00017.safetensors", "model.layers.14.self_attn.q_proj.weight": "model-00006-of-00017.safetensors", "model.layers.14.self_attn.v_proj.weight": "model-00006-of-00017.safetensors", "model.layers.15.self_attn.k_norm.weight": "model-00007-of-00017.safetensors", "model.layers.15.self_attn.k_proj.weight": "model-00007-of-00017.safetensors", "model.layers.15.self_attn.o_proj.weight": "model-00007-of-00017.safetensors", "model.layers.15.self_attn.q_norm.weight": "model-00007-of-00017.safetensors", "model.layers.15.self_attn.q_proj.weight": "model-00007-of-00017.safetensors", "model.layers.15.self_attn.v_proj.weight": "model-00007-of-00017.safetensors", "model.layers.15.input_layernorm.weight": "model-00007-of-00017.safetensors", "model.layers.15.post_attention_layernorm.weight": "model-00007-of-00017.safetensors", "model.layers.16.input_layernorm.weight": "model-00007-of-00017.safetensors", "model.layers.16.post_attention_layernorm.weight": "model-00007-of-00017.safetensors", "model.layers.16.self_attn.k_norm.weight": "model-00007-of-00017.safetensors", "model.layers.16.self_attn.k_proj.weight": "model-00007-of-00017.safetensors", "model.layers.16.self_attn.o_proj.weight": "model-00007-of-00017.safetensors", "model.layers.16.self_attn.q_norm.weight": "model-00007-of-00017.safetensors", "model.layers.16.self_attn.q_proj.weight": "model-00007-of-00017.safetensors", "model.layers.16.self_attn.v_proj.weight": "model-00007-of-00017.safetensors", "model.layers.17.input_layernorm.weight": "model-00007-of-00017.safetensors", "model.layers.17.post_attention_layernorm.weight": "model-00007-of-00017.safetensors", "model.layers.17.self_attn.k_norm.weight": "model-00007-of-00017.safetensors", "model.layers.17.self_attn.k_proj.weight": "model-00007-of-00017.safetensors", "model.layers.17.self_attn.o_proj.weight": "model-00007-of-00017.safetensors", "model.layers.17.self_attn.q_norm.weight": "model-00007-of-00017.safetensors", "model.layers.17.self_attn.q_proj.weight": "model-00007-of-00017.safetensors", "model.layers.17.self_attn.v_proj.weight": "model-00007-of-00017.safetensors", "model.layers.18.input_layernorm.weight": "model-00008-of-00017.safetensors", "model.layers.18.post_attention_layernorm.weight": "model-00008-of-00017.safetensors", "model.layers.18.self_attn.k_norm.weight": "model-00008-of-00017.safetensors", "model.layers.18.self_attn.k_proj.weight": "model-00008-of-00017.safetensors", "model.layers.18.self_attn.o_proj.weight": "model-00008-of-00017.safetensors", "model.layers.18.self_attn.q_norm.weight": "model-00008-of-00017.safetensors", "model.layers.18.self_attn.q_proj.weight": "model-00008-of-00017.safetensors", "model.layers.18.self_attn.v_proj.weight": "model-00008-of-00017.safetensors", "model.layers.19.self_attn.k_norm.weight": "model-00008-of-00017.safetensors", "model.layers.19.self_attn.k_proj.weight": "model-00008-of-00017.safetensors", "model.layers.19.self_attn.o_proj.weight": "model-00008-of-00017.safetensors", "model.layers.19.self_attn.q_norm.weight": "model-00008-of-00017.safetensors", "model.layers.19.self_attn.q_proj.weight": "model-00008-of-00017.safetensors", "model.layers.19.self_attn.v_proj.weight": "model-00008-of-00017.safetensors", "model.layers.1.mlp.router.gate.weight": "model-00002-of-00017.safetensors", "model.layers.1.mlp.experts.w1": "model-00002-of-00017.safetensors", "model.layers.1.mlp.experts.w2": "model-00002-of-00017.safetensors", "model.layers.1.mlp.experts.w3": "model-00002-of-00017.safetensors", "model.layers.10.mlp.router.gate.weight": "model-00005-of-00017.safetensors", "model.layers.10.mlp.experts.w1": "model-00005-of-00017.safetensors", "model.layers.10.mlp.experts.w2": "model-00005-of-00017.safetensors", "model.layers.10.mlp.experts.w3": "model-00005-of-00017.safetensors", "model.layers.11.mlp.router.gate.weight": "model-00005-of-00017.safetensors", "model.layers.11.mlp.experts.w1": "model-00005-of-00017.safetensors", "model.layers.11.mlp.experts.w2": "model-00005-of-00017.safetensors", "model.layers.11.mlp.experts.w3": "model-00005-of-00017.safetensors", "model.layers.12.mlp.router.gate.weight": "model-00006-of-00017.safetensors", "model.layers.12.mlp.experts.w1": "model-00006-of-00017.safetensors", "model.layers.12.mlp.experts.w2": "model-00006-of-00017.safetensors", "model.layers.12.mlp.experts.w3": "model-00006-of-00017.safetensors", "model.layers.13.mlp.router.gate.weight": "model-00006-of-00017.safetensors", "model.layers.13.mlp.experts.w1": "model-00006-of-00017.safetensors", "model.layers.13.mlp.experts.w2": "model-00006-of-00017.safetensors", "model.layers.13.mlp.experts.w3": "model-00006-of-00017.safetensors", "model.layers.14.mlp.router.gate.weight": "model-00006-of-00017.safetensors", "model.layers.14.mlp.experts.w1": "model-00006-of-00017.safetensors", "model.layers.14.mlp.experts.w2": "model-00006-of-00017.safetensors", "model.layers.14.mlp.experts.w3": "model-00006-of-00017.safetensors", "model.layers.15.mlp.router.gate.weight": "model-00007-of-00017.safetensors", "model.layers.15.mlp.experts.w1": "model-00007-of-00017.safetensors", "model.layers.15.mlp.experts.w2": "model-00007-of-00017.safetensors", "model.layers.15.mlp.experts.w3": "model-00007-of-00017.safetensors", "model.layers.16.mlp.router.gate.weight": "model-00007-of-00017.safetensors", "model.layers.16.mlp.experts.w1": "model-00007-of-00017.safetensors", "model.layers.16.mlp.experts.w2": "model-00007-of-00017.safetensors", "model.layers.16.mlp.experts.w3": "model-00007-of-00017.safetensors", "model.layers.17.mlp.router.gate.weight": "model-00007-of-00017.safetensors", "model.layers.17.mlp.experts.w1": "model-00007-of-00017.safetensors", "model.layers.17.mlp.experts.w2": "model-00007-of-00017.safetensors", "model.layers.17.mlp.experts.w3": "model-00007-of-00017.safetensors", "model.layers.18.mlp.router.gate.weight": "model-00008-of-00017.safetensors", "model.layers.18.mlp.experts.w1": "model-00008-of-00017.safetensors", "model.layers.18.mlp.experts.w2": "model-00008-of-00017.safetensors", "model.layers.18.mlp.experts.w3": "model-00008-of-00017.safetensors", "model.layers.19.mlp.router.gate.weight": "model-00008-of-00017.safetensors", "model.layers.19.mlp.experts.w1": "model-00008-of-00017.safetensors", "model.layers.19.mlp.experts.w2": "model-00008-of-00017.safetensors", "model.layers.19.mlp.experts.w3": "model-00008-of-00017.safetensors", "model.layers.2.input_layernorm.weight": "model-00002-of-00017.safetensors", "model.layers.2.post_attention_layernorm.weight": "model-00002-of-00017.safetensors", "model.layers.2.self_attn.k_norm.weight": "model-00002-of-00017.safetensors", "model.layers.2.self_attn.k_proj.weight": "model-00002-of-00017.safetensors", "model.layers.2.self_attn.o_proj.weight": "model-00002-of-00017.safetensors", "model.layers.2.self_attn.q_norm.weight": "model-00002-of-00017.safetensors", "model.layers.2.self_attn.q_proj.weight": "model-00002-of-00017.safetensors", "model.layers.2.self_attn.v_proj.weight": "model-00002-of-00017.safetensors", "model.layers.20.input_layernorm.weight": "model-00008-of-00017.safetensors", "model.layers.20.post_attention_layernorm.weight": "model-00008-of-00017.safetensors", "model.layers.20.self_attn.k_norm.weight": "model-00008-of-00017.safetensors", "model.layers.20.self_attn.k_proj.weight": "model-00008-of-00017.safetensors", "model.layers.20.self_attn.o_proj.weight": "model-00008-of-00017.safetensors", "model.layers.20.self_attn.q_norm.weight": "model-00008-of-00017.safetensors", "model.layers.20.self_attn.q_proj.weight": "model-00008-of-00017.safetensors", "model.layers.20.self_attn.v_proj.weight": "model-00008-of-00017.safetensors", "model.layers.21.input_layernorm.weight": "model-00009-of-00017.safetensors", "model.layers.21.post_attention_layernorm.weight": "model-00009-of-00017.safetensors", "model.layers.21.self_attn.k_norm.weight": "model-00009-of-00017.safetensors", "model.layers.21.self_attn.k_proj.weight": "model-00009-of-00017.safetensors", "model.layers.21.self_attn.o_proj.weight": "model-00009-of-00017.safetensors", "model.layers.21.self_attn.q_norm.weight": "model-00009-of-00017.safetensors", "model.layers.21.self_attn.q_proj.weight": "model-00009-of-00017.safetensors", "model.layers.21.self_attn.v_proj.weight": "model-00009-of-00017.safetensors", "model.layers.22.input_layernorm.weight": "model-00009-of-00017.safetensors", "model.layers.22.post_attention_layernorm.weight": "model-00009-of-00017.safetensors", "model.layers.22.self_attn.k_norm.weight": "model-00009-of-00017.safetensors", "model.layers.22.self_attn.k_proj.weight": "model-00009-of-00017.safetensors", "model.layers.22.self_attn.o_proj.weight": "model-00009-of-00017.safetensors", "model.layers.22.self_attn.q_norm.weight": "model-00009-of-00017.safetensors", "model.layers.22.self_attn.q_proj.weight": "model-00009-of-00017.safetensors", "model.layers.22.self_attn.v_proj.weight": "model-00009-of-00017.safetensors", "model.layers.23.self_attn.k_norm.weight": "model-00009-of-00017.safetensors", "model.layers.23.self_attn.k_proj.weight": "model-00009-of-00017.safetensors", "model.layers.23.self_attn.o_proj.weight": "model-00009-of-00017.safetensors", "model.layers.23.self_attn.q_norm.weight": "model-00009-of-00017.safetensors", "model.layers.23.self_attn.q_proj.weight": "model-00009-of-00017.safetensors", "model.layers.23.self_attn.v_proj.weight": "model-00009-of-00017.safetensors", "model.layers.23.input_layernorm.weight": "model-00009-of-00017.safetensors", "model.layers.23.post_attention_layernorm.weight": "model-00009-of-00017.safetensors", "model.layers.24.input_layernorm.weight": "model-00010-of-00017.safetensors", "model.layers.24.post_attention_layernorm.weight": "model-00010-of-00017.safetensors", "model.layers.24.self_attn.k_norm.weight": "model-00010-of-00017.safetensors", "model.layers.24.self_attn.k_proj.weight": "model-00010-of-00017.safetensors", "model.layers.24.self_attn.o_proj.weight": "model-00010-of-00017.safetensors", "model.layers.24.self_attn.q_norm.weight": "model-00010-of-00017.safetensors", "model.layers.24.self_attn.q_proj.weight": "model-00010-of-00017.safetensors", "model.layers.24.self_attn.v_proj.weight": "model-00010-of-00017.safetensors", "model.layers.25.input_layernorm.weight": "model-00010-of-00017.safetensors", "model.layers.25.post_attention_layernorm.weight": "model-00010-of-00017.safetensors", "model.layers.25.self_attn.k_norm.weight": "model-00010-of-00017.safetensors", "model.layers.25.self_attn.k_proj.weight": "model-00010-of-00017.safetensors", "model.layers.25.self_attn.o_proj.weight": "model-00010-of-00017.safetensors", "model.layers.25.self_attn.q_norm.weight": "model-00010-of-00017.safetensors", "model.layers.25.self_attn.q_proj.weight": "model-00010-of-00017.safetensors", "model.layers.25.self_attn.v_proj.weight": "model-00010-of-00017.safetensors", "model.layers.26.input_layernorm.weight": "model-00010-of-00017.safetensors", "model.layers.26.post_attention_layernorm.weight": "model-00010-of-00017.safetensors", "model.layers.26.self_attn.k_norm.weight": "model-00010-of-00017.safetensors", "model.layers.26.self_attn.k_proj.weight": "model-00010-of-00017.safetensors", "model.layers.26.self_attn.o_proj.weight": "model-00010-of-00017.safetensors", "model.layers.26.self_attn.q_norm.weight": "model-00010-of-00017.safetensors", "model.layers.26.self_attn.q_proj.weight": "model-00010-of-00017.safetensors", "model.layers.26.self_attn.v_proj.weight": "model-00010-of-00017.safetensors", "model.layers.27.self_attn.k_norm.weight": "model-00011-of-00017.safetensors", "model.layers.27.self_attn.k_proj.weight": "model-00011-of-00017.safetensors", "model.layers.27.self_attn.o_proj.weight": "model-00011-of-00017.safetensors", "model.layers.27.self_attn.q_norm.weight": "model-00011-of-00017.safetensors", "model.layers.27.self_attn.q_proj.weight": "model-00011-of-00017.safetensors", "model.layers.27.self_attn.v_proj.weight": "model-00011-of-00017.safetensors", "model.layers.27.input_layernorm.weight": "model-00011-of-00017.safetensors", "model.layers.27.post_attention_layernorm.weight": "model-00011-of-00017.safetensors", "model.layers.28.input_layernorm.weight": "model-00011-of-00017.safetensors", "model.layers.28.post_attention_layernorm.weight": "model-00011-of-00017.safetensors", "model.layers.28.self_attn.k_norm.weight": "model-00011-of-00017.safetensors", "model.layers.28.self_attn.k_proj.weight": "model-00011-of-00017.safetensors", "model.layers.28.self_attn.o_proj.weight": "model-00011-of-00017.safetensors", "model.layers.28.self_attn.q_norm.weight": "model-00011-of-00017.safetensors", "model.layers.28.self_attn.q_proj.weight": "model-00011-of-00017.safetensors", "model.layers.28.self_attn.v_proj.weight": "model-00011-of-00017.safetensors", "model.layers.29.input_layernorm.weight": "model-00011-of-00017.safetensors", "model.layers.29.post_attention_layernorm.weight": "model-00011-of-00017.safetensors", "model.layers.29.self_attn.k_norm.weight": "model-00011-of-00017.safetensors", "model.layers.29.self_attn.k_proj.weight": "model-00011-of-00017.safetensors", "model.layers.29.self_attn.o_proj.weight": "model-00011-of-00017.safetensors", "model.layers.29.self_attn.q_norm.weight": "model-00011-of-00017.safetensors", "model.layers.29.self_attn.q_proj.weight": "model-00011-of-00017.safetensors", "model.layers.29.self_attn.v_proj.weight": "model-00011-of-00017.safetensors", "model.layers.2.mlp.router.gate.weight": "model-00002-of-00017.safetensors", "model.layers.2.mlp.experts.w1": "model-00002-of-00017.safetensors", "model.layers.2.mlp.experts.w2": "model-00002-of-00017.safetensors", "model.layers.2.mlp.experts.w3": "model-00002-of-00017.safetensors", "model.layers.20.mlp.router.gate.weight": "model-00008-of-00017.safetensors", "model.layers.20.mlp.experts.w1": "model-00008-of-00017.safetensors", "model.layers.20.mlp.experts.w2": "model-00008-of-00017.safetensors", "model.layers.20.mlp.experts.w3": "model-00008-of-00017.safetensors", "model.layers.21.mlp.router.gate.weight": "model-00009-of-00017.safetensors", "model.layers.21.mlp.experts.w1": "model-00009-of-00017.safetensors", "model.layers.21.mlp.experts.w2": "model-00009-of-00017.safetensors", "model.layers.21.mlp.experts.w3": "model-00009-of-00017.safetensors", "model.layers.22.mlp.router.gate.weight": "model-00009-of-00017.safetensors", "model.layers.22.mlp.experts.w1": "model-00009-of-00017.safetensors", "model.layers.22.mlp.experts.w2": "model-00009-of-00017.safetensors", "model.layers.22.mlp.experts.w3": "model-00009-of-00017.safetensors", "model.layers.23.mlp.router.gate.weight": "model-00009-of-00017.safetensors", "model.layers.23.mlp.experts.w1": "model-00009-of-00017.safetensors", "model.layers.23.mlp.experts.w2": "model-00009-of-00017.safetensors", "model.layers.23.mlp.experts.w3": "model-00009-of-00017.safetensors", "model.layers.24.mlp.router.gate.weight": "model-00010-of-00017.safetensors", "model.layers.24.mlp.experts.w1": "model-00010-of-00017.safetensors", "model.layers.24.mlp.experts.w2": "model-00010-of-00017.safetensors", "model.layers.24.mlp.experts.w3": "model-00010-of-00017.safetensors", "model.layers.25.mlp.router.gate.weight": "model-00010-of-00017.safetensors", "model.layers.25.mlp.experts.w1": "model-00010-of-00017.safetensors", "model.layers.25.mlp.experts.w2": "model-00010-of-00017.safetensors", "model.layers.25.mlp.experts.w3": "model-00010-of-00017.safetensors", "model.layers.26.mlp.router.gate.weight": "model-00010-of-00017.safetensors", "model.layers.26.mlp.experts.w1": "model-00010-of-00017.safetensors", "model.layers.26.mlp.experts.w2": "model-00010-of-00017.safetensors", "model.layers.26.mlp.experts.w3": "model-00010-of-00017.safetensors", "model.layers.27.mlp.router.gate.weight": "model-00011-of-00017.safetensors", "model.layers.27.mlp.experts.w1": "model-00011-of-00017.safetensors", "model.layers.27.mlp.experts.w2": "model-00011-of-00017.safetensors", "model.layers.27.mlp.experts.w3": "model-00011-of-00017.safetensors", "model.layers.28.mlp.router.gate.weight": "model-00011-of-00017.safetensors", "model.layers.28.mlp.experts.w1": "model-00011-of-00017.safetensors", "model.layers.28.mlp.experts.w2": "model-00011-of-00017.safetensors", "model.layers.28.mlp.experts.w3": "model-00011-of-00017.safetensors", "model.layers.29.mlp.router.gate.weight": "model-00011-of-00017.safetensors", "model.layers.29.mlp.experts.w1": "model-00011-of-00017.safetensors", "model.layers.29.mlp.experts.w2": "model-00011-of-00017.safetensors", "model.layers.29.mlp.experts.w3": "model-00011-of-00017.safetensors", "model.layers.3.self_attn.k_norm.weight": "model-00003-of-00017.safetensors", "model.layers.3.self_attn.k_proj.weight": "model-00003-of-00017.safetensors", "model.layers.3.self_attn.o_proj.weight": "model-00003-of-00017.safetensors", "model.layers.3.self_attn.q_norm.weight": "model-00003-of-00017.safetensors", "model.layers.3.self_attn.q_proj.weight": "model-00003-of-00017.safetensors", "model.layers.3.self_attn.v_proj.weight": "model-00003-of-00017.safetensors", "model.layers.35.input_layernorm.weight": "model-00013-of-00017.safetensors", "model.layers.35.post_attention_layernorm.weight": "model-00013-of-00017.safetensors", "model.layers.36.input_layernorm.weight": "model-00014-of-00017.safetensors", "model.layers.36.post_attention_layernorm.weight": "model-00014-of-00017.safetensors", "model.layers.36.self_attn.k_norm.weight": "model-00014-of-00017.safetensors", "model.layers.36.self_attn.k_proj.weight": "model-00014-of-00017.safetensors", "model.layers.36.self_attn.o_proj.weight": "model-00014-of-00017.safetensors", "model.layers.36.self_attn.q_norm.weight": "model-00014-of-00017.safetensors", "model.layers.36.self_attn.q_proj.weight": "model-00014-of-00017.safetensors", "model.layers.36.self_attn.v_proj.weight": "model-00014-of-00017.safetensors", "model.layers.37.input_layernorm.weight": "model-00014-of-00017.safetensors", "model.layers.37.post_attention_layernorm.weight": "model-00014-of-00017.safetensors", "model.layers.37.self_attn.k_norm.weight": "model-00014-of-00017.safetensors", "model.layers.37.self_attn.k_proj.weight": "model-00014-of-00017.safetensors", "model.layers.37.self_attn.o_proj.weight": "model-00014-of-00017.safetensors", "model.layers.37.self_attn.q_norm.weight": "model-00014-of-00017.safetensors", "model.layers.37.self_attn.q_proj.weight": "model-00014-of-00017.safetensors", "model.layers.37.self_attn.v_proj.weight": "model-00014-of-00017.safetensors", "model.layers.38.input_layernorm.weight": "model-00014-of-00017.safetensors", "model.layers.38.post_attention_layernorm.weight": "model-00014-of-00017.safetensors", "model.layers.38.self_attn.k_norm.weight": "model-00014-of-00017.safetensors", "model.layers.38.self_attn.k_proj.weight": "model-00014-of-00017.safetensors", "model.layers.38.self_attn.o_proj.weight": "model-00014-of-00017.safetensors", "model.layers.38.self_attn.q_norm.weight": "model-00014-of-00017.safetensors", "model.layers.38.self_attn.q_proj.weight": "model-00014-of-00017.safetensors", "model.layers.38.self_attn.v_proj.weight": "model-00014-of-00017.safetensors", "model.layers.39.self_attn.k_norm.weight": "model-00015-of-00017.safetensors", "model.layers.39.self_attn.k_proj.weight": "model-00015-of-00017.safetensors", "model.layers.39.self_attn.o_proj.weight": "model-00015-of-00017.safetensors", "model.layers.39.self_attn.q_norm.weight": "model-00015-of-00017.safetensors", "model.layers.39.self_attn.q_proj.weight": "model-00015-of-00017.safetensors", "model.layers.39.self_attn.v_proj.weight": "model-00015-of-00017.safetensors", "model.layers.31.input_layernorm.weight": "model-00012-of-00017.safetensors", "model.layers.31.post_attention_layernorm.weight": "model-00012-of-00017.safetensors", "model.layers.32.input_layernorm.weight": "model-00012-of-00017.safetensors", "model.layers.32.post_attention_layernorm.weight": "model-00012-of-00017.safetensors", "model.layers.32.self_attn.k_norm.weight": "model-00012-of-00017.safetensors", "model.layers.32.self_attn.k_proj.weight": "model-00012-of-00017.safetensors", "model.layers.32.self_attn.o_proj.weight": "model-00012-of-00017.safetensors", "model.layers.32.self_attn.q_norm.weight": "model-00012-of-00017.safetensors", "model.layers.32.self_attn.q_proj.weight": "model-00012-of-00017.safetensors", "model.layers.32.self_attn.v_proj.weight": "model-00012-of-00017.safetensors", "model.layers.33.input_layernorm.weight": "model-00013-of-00017.safetensors", "model.layers.33.post_attention_layernorm.weight": "model-00013-of-00017.safetensors", "model.layers.33.self_attn.k_norm.weight": "model-00013-of-00017.safetensors", "model.layers.33.self_attn.k_proj.weight": "model-00013-of-00017.safetensors", "model.layers.33.self_attn.o_proj.weight": "model-00013-of-00017.safetensors", "model.layers.33.self_attn.q_norm.weight": "model-00013-of-00017.safetensors", "model.layers.33.self_attn.q_proj.weight": "model-00013-of-00017.safetensors", "model.layers.33.self_attn.v_proj.weight": "model-00013-of-00017.safetensors", "model.layers.34.input_layernorm.weight": "model-00013-of-00017.safetensors", "model.layers.34.post_attention_layernorm.weight": "model-00013-of-00017.safetensors", "model.layers.34.self_attn.k_norm.weight": "model-00013-of-00017.safetensors", "model.layers.34.self_attn.k_proj.weight": "model-00013-of-00017.safetensors", "model.layers.34.self_attn.o_proj.weight": "model-00013-of-00017.safetensors", "model.layers.34.self_attn.q_norm.weight": "model-00013-of-00017.safetensors", "model.layers.34.self_attn.q_proj.weight": "model-00013-of-00017.safetensors", "model.layers.34.self_attn.v_proj.weight": "model-00013-of-00017.safetensors", "model.layers.35.self_attn.k_norm.weight": "model-00013-of-00017.safetensors", "model.layers.35.self_attn.k_proj.weight": "model-00013-of-00017.safetensors", "model.layers.35.self_attn.o_proj.weight": "model-00013-of-00017.safetensors", "model.layers.35.self_attn.q_norm.weight": "model-00013-of-00017.safetensors", "model.layers.35.self_attn.q_proj.weight": "model-00013-of-00017.safetensors", "model.layers.35.self_attn.v_proj.weight": "model-00013-of-00017.safetensors", "model.layers.3.input_layernorm.weight": "model-00003-of-00017.safetensors", "model.layers.3.post_attention_layernorm.weight": "model-00003-of-00017.safetensors", "model.layers.39.input_layernorm.weight": "model-00015-of-00017.safetensors", "model.layers.39.post_attention_layernorm.weight": "model-00015-of-00017.safetensors", "model.layers.30.input_layernorm.weight": "model-00012-of-00017.safetensors", "model.layers.30.post_attention_layernorm.weight": "model-00012-of-00017.safetensors", "model.layers.30.self_attn.k_norm.weight": "model-00012-of-00017.safetensors", "model.layers.30.self_attn.k_proj.weight": "model-00012-of-00017.safetensors", "model.layers.30.self_attn.o_proj.weight": "model-00012-of-00017.safetensors", "model.layers.30.self_attn.q_norm.weight": "model-00012-of-00017.safetensors", "model.layers.30.self_attn.q_proj.weight": "model-00012-of-00017.safetensors", "model.layers.30.self_attn.v_proj.weight": "model-00012-of-00017.safetensors", "model.layers.31.self_attn.k_norm.weight": "model-00012-of-00017.safetensors", "model.layers.31.self_attn.k_proj.weight": "model-00012-of-00017.safetensors", "model.layers.31.self_attn.o_proj.weight": "model-00012-of-00017.safetensors", "model.layers.31.self_attn.q_norm.weight": "model-00012-of-00017.safetensors", "model.layers.31.self_attn.q_proj.weight": "model-00012-of-00017.safetensors", "model.layers.31.self_attn.v_proj.weight": "model-00012-of-00017.safetensors", "model.layers.3.mlp.router.gate.weight": "model-00003-of-00017.safetensors", "model.layers.3.mlp.experts.w1": "model-00003-of-00017.safetensors", "model.layers.3.mlp.experts.w2": "model-00003-of-00017.safetensors", "model.layers.3.mlp.experts.w3": "model-00003-of-00017.safetensors", "model.layers.30.mlp.router.gate.weight": "model-00012-of-00017.safetensors", "model.layers.30.mlp.experts.w1": "model-00012-of-00017.safetensors", "model.layers.30.mlp.experts.w2": "model-00012-of-00017.safetensors", "model.layers.30.mlp.experts.w3": "model-00012-of-00017.safetensors", "model.layers.31.mlp.router.gate.weight": "model-00012-of-00017.safetensors", "model.layers.31.mlp.experts.w1": "model-00012-of-00017.safetensors", "model.layers.31.mlp.experts.w2": "model-00012-of-00017.safetensors", "model.layers.31.mlp.experts.w3": "model-00012-of-00017.safetensors", "model.layers.32.mlp.router.gate.weight": "model-00012-of-00017.safetensors", "model.layers.32.mlp.experts.w1": "model-00012-of-00017.safetensors", "model.layers.32.mlp.experts.w2": "model-00012-of-00017.safetensors", "model.layers.32.mlp.experts.w3": "model-00012-of-00017.safetensors", "model.layers.33.mlp.router.gate.weight": "model-00013-of-00017.safetensors", "model.layers.33.mlp.experts.w1": "model-00013-of-00017.safetensors", "model.layers.33.mlp.experts.w2": "model-00013-of-00017.safetensors", "model.layers.33.mlp.experts.w3": "model-00013-of-00017.safetensors", "model.layers.34.mlp.router.gate.weight": "model-00013-of-00017.safetensors", "model.layers.34.mlp.experts.w1": "model-00013-of-00017.safetensors", "model.layers.34.mlp.experts.w2": "model-00013-of-00017.safetensors", "model.layers.34.mlp.experts.w3": "model-00013-of-00017.safetensors", "model.layers.35.mlp.router.gate.weight": "model-00013-of-00017.safetensors", "model.layers.35.mlp.experts.w1": "model-00013-of-00017.safetensors", "model.layers.35.mlp.experts.w2": "model-00013-of-00017.safetensors", "model.layers.35.mlp.experts.w3": "model-00013-of-00017.safetensors", "model.layers.36.mlp.router.gate.weight": "model-00014-of-00017.safetensors", "model.layers.36.mlp.experts.w1": "model-00014-of-00017.safetensors", "model.layers.36.mlp.experts.w2": "model-00014-of-00017.safetensors", "model.layers.36.mlp.experts.w3": "model-00014-of-00017.safetensors", "model.layers.37.mlp.router.gate.weight": "model-00014-of-00017.safetensors", "model.layers.37.mlp.experts.w1": "model-00014-of-00017.safetensors", "model.layers.37.mlp.experts.w2": "model-00014-of-00017.safetensors", "model.layers.37.mlp.experts.w3": "model-00014-of-00017.safetensors", "model.layers.38.mlp.router.gate.weight": "model-00014-of-00017.safetensors", "model.layers.38.mlp.experts.w1": "model-00014-of-00017.safetensors", "model.layers.38.mlp.experts.w2": "model-00014-of-00017.safetensors", "model.layers.38.mlp.experts.w3": "model-00014-of-00017.safetensors", "model.layers.39.mlp.router.gate.weight": "model-00015-of-00017.safetensors", "model.layers.39.mlp.experts.w1": "model-00015-of-00017.safetensors", "model.layers.39.mlp.experts.w2": "model-00015-of-00017.safetensors", "model.layers.39.mlp.experts.w3": "model-00015-of-00017.safetensors", "model.layers.43.input_layernorm.weight": "model-00016-of-00017.safetensors", "model.layers.43.post_attention_layernorm.weight": "model-00016-of-00017.safetensors", "model.layers.44.input_layernorm.weight": "model-00016-of-00017.safetensors", "model.layers.44.post_attention_layernorm.weight": "model-00016-of-00017.safetensors", "model.layers.44.self_attn.k_norm.weight": "model-00016-of-00017.safetensors", "model.layers.44.self_attn.k_proj.weight": "model-00016-of-00017.safetensors", "model.layers.44.self_attn.o_proj.weight": "model-00016-of-00017.safetensors", "model.layers.44.self_attn.q_norm.weight": "model-00016-of-00017.safetensors", "model.layers.44.self_attn.q_proj.weight": "model-00016-of-00017.safetensors", "model.layers.44.self_attn.v_proj.weight": "model-00016-of-00017.safetensors", "model.layers.45.input_layernorm.weight": "model-00017-of-00017.safetensors", "model.layers.45.post_attention_layernorm.weight": "model-00017-of-00017.safetensors", "model.layers.45.self_attn.k_norm.weight": "model-00017-of-00017.safetensors", "model.layers.45.self_attn.k_proj.weight": "model-00017-of-00017.safetensors", "model.layers.45.self_attn.o_proj.weight": "model-00017-of-00017.safetensors", "model.layers.45.self_attn.q_norm.weight": "model-00017-of-00017.safetensors", "model.layers.45.self_attn.q_proj.weight": "model-00017-of-00017.safetensors", "model.layers.45.self_attn.v_proj.weight": "model-00017-of-00017.safetensors", "model.layers.46.input_layernorm.weight": "model-00017-of-00017.safetensors", "model.layers.46.post_attention_layernorm.weight": "model-00017-of-00017.safetensors", "model.layers.46.self_attn.k_norm.weight": "model-00017-of-00017.safetensors", "model.layers.46.self_attn.k_proj.weight": "model-00017-of-00017.safetensors", "model.layers.46.self_attn.o_proj.weight": "model-00017-of-00017.safetensors", "model.layers.46.self_attn.q_norm.weight": "model-00017-of-00017.safetensors", "model.layers.46.self_attn.q_proj.weight": "model-00017-of-00017.safetensors", "model.layers.46.self_attn.v_proj.weight": "model-00017-of-00017.safetensors", "model.layers.47.self_attn.k_norm.weight": "model-00017-of-00017.safetensors", "model.layers.47.self_attn.k_proj.weight": "model-00017-of-00017.safetensors", "model.layers.47.self_attn.o_proj.weight": "model-00017-of-00017.safetensors", "model.layers.47.self_attn.q_norm.weight": "model-00017-of-00017.safetensors", "model.layers.47.self_attn.q_proj.weight": "model-00017-of-00017.safetensors", "model.layers.47.self_attn.v_proj.weight": "model-00017-of-00017.safetensors", "model.layers.47.input_layernorm.weight": "model-00017-of-00017.safetensors", "model.layers.47.post_attention_layernorm.weight": "model-00017-of-00017.safetensors", "model.layers.4.input_layernorm.weight": "model-00003-of-00017.safetensors", "model.layers.4.post_attention_layernorm.weight": "model-00003-of-00017.safetensors", "model.layers.4.self_attn.k_norm.weight": "model-00003-of-00017.safetensors", "model.layers.4.self_attn.k_proj.weight": "model-00003-of-00017.safetensors", "model.layers.4.self_attn.o_proj.weight": "model-00003-of-00017.safetensors", "model.layers.4.self_attn.q_norm.weight": "model-00003-of-00017.safetensors", "model.layers.4.self_attn.q_proj.weight": "model-00003-of-00017.safetensors", "model.layers.4.self_attn.v_proj.weight": "model-00003-of-00017.safetensors", "model.layers.40.input_layernorm.weight": "model-00015-of-00017.safetensors", "model.layers.40.post_attention_layernorm.weight": "model-00015-of-00017.safetensors", "model.layers.40.self_attn.k_norm.weight": "model-00015-of-00017.safetensors", "model.layers.40.self_attn.k_proj.weight": "model-00015-of-00017.safetensors", "model.layers.40.self_attn.o_proj.weight": "model-00015-of-00017.safetensors", "model.layers.40.self_attn.q_norm.weight": "model-00015-of-00017.safetensors", "model.layers.40.self_attn.q_proj.weight": "model-00015-of-00017.safetensors", "model.layers.40.self_attn.v_proj.weight": "model-00015-of-00017.safetensors", "model.layers.41.input_layernorm.weight": "model-00015-of-00017.safetensors", "model.layers.41.post_attention_layernorm.weight": "model-00015-of-00017.safetensors", "model.layers.41.self_attn.k_norm.weight": "model-00015-of-00017.safetensors", "model.layers.41.self_attn.k_proj.weight": "model-00015-of-00017.safetensors", "model.layers.41.self_attn.o_proj.weight": "model-00015-of-00017.safetensors", "model.layers.41.self_attn.q_norm.weight": "model-00015-of-00017.safetensors", "model.layers.41.self_attn.q_proj.weight": "model-00015-of-00017.safetensors", "model.layers.41.self_attn.v_proj.weight": "model-00015-of-00017.safetensors", "model.layers.42.input_layernorm.weight": "model-00016-of-00017.safetensors", "model.layers.42.post_attention_layernorm.weight": "model-00016-of-00017.safetensors", "model.layers.42.self_attn.k_norm.weight": "model-00016-of-00017.safetensors", "model.layers.42.self_attn.k_proj.weight": "model-00016-of-00017.safetensors", "model.layers.42.self_attn.o_proj.weight": "model-00016-of-00017.safetensors", "model.layers.42.self_attn.q_norm.weight": "model-00016-of-00017.safetensors", "model.layers.42.self_attn.q_proj.weight": "model-00016-of-00017.safetensors", "model.layers.42.self_attn.v_proj.weight": "model-00016-of-00017.safetensors", "model.layers.43.self_attn.k_norm.weight": "model-00016-of-00017.safetensors", "model.layers.43.self_attn.k_proj.weight": "model-00016-of-00017.safetensors", "model.layers.43.self_attn.o_proj.weight": "model-00016-of-00017.safetensors", "model.layers.43.self_attn.q_norm.weight": "model-00016-of-00017.safetensors", "model.layers.43.self_attn.q_proj.weight": "model-00016-of-00017.safetensors", "model.layers.43.self_attn.v_proj.weight": "model-00016-of-00017.safetensors", "model.layers.4.mlp.router.gate.weight": "model-00003-of-00017.safetensors", "model.layers.4.mlp.experts.w1": "model-00003-of-00017.safetensors", "model.layers.4.mlp.experts.w2": "model-00003-of-00017.safetensors", "model.layers.4.mlp.experts.w3": "model-00003-of-00017.safetensors", "model.layers.40.mlp.router.gate.weight": "model-00015-of-00017.safetensors", "model.layers.40.mlp.experts.w1": "model-00015-of-00017.safetensors", "model.layers.40.mlp.experts.w2": "model-00015-of-00017.safetensors", "model.layers.40.mlp.experts.w3": "model-00015-of-00017.safetensors", "model.layers.41.mlp.router.gate.weight": "model-00015-of-00017.safetensors", "model.layers.41.mlp.experts.w1": "model-00015-of-00017.safetensors", "model.layers.41.mlp.experts.w2": "model-00015-of-00017.safetensors", "model.layers.41.mlp.experts.w3": "model-00015-of-00017.safetensors", "model.layers.42.mlp.router.gate.weight": "model-00016-of-00017.safetensors", "model.layers.42.mlp.experts.w1": "model-00016-of-00017.safetensors", "model.layers.42.mlp.experts.w2": "model-00016-of-00017.safetensors", "model.layers.42.mlp.experts.w3": "model-00016-of-00017.safetensors", "model.layers.43.mlp.router.gate.weight": "model-00016-of-00017.safetensors", "model.layers.43.mlp.experts.w1": "model-00016-of-00017.safetensors", "model.layers.43.mlp.experts.w2": "model-00016-of-00017.safetensors", "model.layers.43.mlp.experts.w3": "model-00016-of-00017.safetensors", "model.layers.44.mlp.router.gate.weight": "model-00016-of-00017.safetensors", "model.layers.44.mlp.experts.w1": "model-00016-of-00017.safetensors", "model.layers.44.mlp.experts.w2": "model-00016-of-00017.safetensors", "model.layers.44.mlp.experts.w3": "model-00016-of-00017.safetensors", "model.layers.45.mlp.router.gate.weight": "model-00017-of-00017.safetensors", "model.layers.45.mlp.experts.w1": "model-00017-of-00017.safetensors", "model.layers.45.mlp.experts.w2": "model-00017-of-00017.safetensors", "model.layers.45.mlp.experts.w3": "model-00017-of-00017.safetensors", "model.layers.46.mlp.router.gate.weight": "model-00017-of-00017.safetensors", "model.layers.46.mlp.experts.w1": "model-00017-of-00017.safetensors", "model.layers.46.mlp.experts.w2": "model-00017-of-00017.safetensors", "model.layers.46.mlp.experts.w3": "model-00017-of-00017.safetensors", "model.layers.47.mlp.router.gate.weight": "model-00017-of-00017.safetensors", "model.layers.47.mlp.experts.w1": "model-00017-of-00017.safetensors", "model.layers.47.mlp.experts.w2": "model-00017-of-00017.safetensors", "model.layers.47.mlp.experts.w3": "model-00017-of-00017.safetensors", "model.layers.5.input_layernorm.weight": "model-00003-of-00017.safetensors", "model.layers.5.post_attention_layernorm.weight": "model-00003-of-00017.safetensors", "model.layers.5.self_attn.k_norm.weight": "model-00003-of-00017.safetensors", "model.layers.5.self_attn.k_proj.weight": "model-00003-of-00017.safetensors", "model.layers.5.self_attn.o_proj.weight": "model-00003-of-00017.safetensors", "model.layers.5.self_attn.q_norm.weight": "model-00003-of-00017.safetensors", "model.layers.5.self_attn.q_proj.weight": "model-00003-of-00017.safetensors", "model.layers.5.self_attn.v_proj.weight": "model-00003-of-00017.safetensors", "model.layers.5.mlp.router.gate.weight": "model-00003-of-00017.safetensors", "model.layers.5.mlp.experts.w1": "model-00003-of-00017.safetensors", "model.layers.5.mlp.experts.w2": "model-00003-of-00017.safetensors", "model.layers.5.mlp.experts.w3": "model-00003-of-00017.safetensors", "model.layers.6.input_layernorm.weight": "model-00004-of-00017.safetensors", "model.layers.6.post_attention_layernorm.weight": "model-00004-of-00017.safetensors", "model.layers.6.self_attn.k_norm.weight": "model-00004-of-00017.safetensors", "model.layers.6.self_attn.k_proj.weight": "model-00004-of-00017.safetensors", "model.layers.6.self_attn.o_proj.weight": "model-00004-of-00017.safetensors", "model.layers.6.self_attn.q_norm.weight": "model-00004-of-00017.safetensors", "model.layers.6.self_attn.q_proj.weight": "model-00004-of-00017.safetensors", "model.layers.6.self_attn.v_proj.weight": "model-00004-of-00017.safetensors", "model.layers.6.mlp.router.gate.weight": "model-00004-of-00017.safetensors", "model.layers.6.mlp.experts.w1": "model-00004-of-00017.safetensors", "model.layers.6.mlp.experts.w2": "model-00004-of-00017.safetensors", "model.layers.6.mlp.experts.w3": "model-00004-of-00017.safetensors", "model.layers.7.input_layernorm.weight": "model-00004-of-00017.safetensors", "model.layers.7.post_attention_layernorm.weight": "model-00004-of-00017.safetensors", "model.layers.7.self_attn.k_norm.weight": "model-00004-of-00017.safetensors", "model.layers.7.self_attn.k_proj.weight": "model-00004-of-00017.safetensors", "model.layers.7.self_attn.o_proj.weight": "model-00004-of-00017.safetensors", "model.layers.7.self_attn.q_norm.weight": "model-00004-of-00017.safetensors", "model.layers.7.self_attn.q_proj.weight": "model-00004-of-00017.safetensors", "model.layers.7.self_attn.v_proj.weight": "model-00004-of-00017.safetensors", "model.layers.7.mlp.router.gate.weight": "model-00004-of-00017.safetensors", "model.layers.7.mlp.experts.w1": "model-00004-of-00017.safetensors", "model.layers.7.mlp.experts.w2": "model-00004-of-00017.safetensors", "model.layers.7.mlp.experts.w3": "model-00004-of-00017.safetensors", "model.layers.8.input_layernorm.weight": "model-00004-of-00017.safetensors", "model.layers.8.post_attention_layernorm.weight": "model-00004-of-00017.safetensors", "model.layers.8.self_attn.k_norm.weight": "model-00004-of-00017.safetensors", "model.layers.8.self_attn.k_proj.weight": "model-00004-of-00017.safetensors", "model.layers.8.self_attn.o_proj.weight": "model-00004-of-00017.safetensors", "model.layers.8.self_attn.q_norm.weight": "model-00004-of-00017.safetensors", "model.layers.8.self_attn.q_proj.weight": "model-00004-of-00017.safetensors", "model.layers.8.self_attn.v_proj.weight": "model-00004-of-00017.safetensors", "model.layers.8.mlp.router.gate.weight": "model-00004-of-00017.safetensors", "model.layers.8.mlp.experts.w1": "model-00004-of-00017.safetensors", "model.layers.8.mlp.experts.w2": "model-00004-of-00017.safetensors", "model.layers.8.mlp.experts.w3": "model-00004-of-00017.safetensors", "model.layers.9.input_layernorm.weight": "model-00005-of-00017.safetensors", "model.layers.9.post_attention_layernorm.weight": "model-00005-of-00017.safetensors", "model.layers.9.self_attn.k_norm.weight": "model-00005-of-00017.safetensors", "model.layers.9.self_attn.k_proj.weight": "model-00005-of-00017.safetensors", "model.layers.9.self_attn.o_proj.weight": "model-00005-of-00017.safetensors", "model.layers.9.self_attn.q_norm.weight": "model-00005-of-00017.safetensors", "model.layers.9.self_attn.q_proj.weight": "model-00005-of-00017.safetensors", "model.layers.9.self_attn.v_proj.weight": "model-00005-of-00017.safetensors", "model.layers.9.mlp.router.gate.weight": "model-00005-of-00017.safetensors", "model.layers.9.mlp.experts.w1": "model-00005-of-00017.safetensors", "model.layers.9.mlp.experts.w2": "model-00005-of-00017.safetensors", "model.layers.9.mlp.experts.w3": "model-00005-of-00017.safetensors"}}

modeling_qwen3_moe.py

@@ -0,0 +1,730 @@
+# coding=utf-8
+# Copyright 2025 The Qwen team, Alibaba Group and the HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from typing import Callable, Optional, Union
+
+import torch
+import torch.nn.functional as F
+from torch import nn
+
+from transformers.activations import ACT2FN
+from transformers.cache_utils import Cache, DynamicCache
+from transformers.generation import GenerationMixin
+from transformers.integrations import use_kernel_forward_from_hub
+from transformers.masking_utils import create_causal_mask, create_sliding_window_causal_mask
+from transformers.modeling_flash_attention_utils import FlashAttentionKwargs
+from transformers.modeling_layers import (
+    GenericForQuestionAnswering,
+    GenericForSequenceClassification,
+    GenericForTokenClassification,
+    GradientCheckpointingLayer,
+)
+from transformers.modeling_outputs import MoeCausalLMOutputWithPast, MoeModelOutputWithPast
+from transformers.modeling_rope_utils import ROPE_INIT_FUNCTIONS, dynamic_rope_update
+from transformers.modeling_utils import ALL_ATTENTION_FUNCTIONS, PreTrainedModel
+from transformers.processing_utils import Unpack
+from transformers.utils import TransformersKwargs, auto_docstring, can_return_tuple
+from transformers.utils.generic import OutputRecorder, check_model_inputs
+from .configuration_qwen3_moe import Qwen3MoeConfig
+
+from torchtitan.models.moe import MoE, MoEArgs
+
+
+def rotate_half(x):
+    """Rotates half the hidden dims of the input."""
+    x1 = x[..., : x.shape[-1] // 2]
+    x2 = x[..., x.shape[-1] // 2 :]
+    return torch.cat((-x2, x1), dim=-1)
+
+
+def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_dim=1):
+    """Applies Rotary Position Embedding to the query and key tensors.
+
+    Args:
+        q (`torch.Tensor`): The query tensor.
+        k (`torch.Tensor`): The key tensor.
+        cos (`torch.Tensor`): The cosine part of the rotary embedding.
+        sin (`torch.Tensor`): The sine part of the rotary embedding.
+        position_ids (`torch.Tensor`, *optional*):
+            Deprecated and unused.
+        unsqueeze_dim (`int`, *optional*, defaults to 1):
+            The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and
+            sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note
+            that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and
+            k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes
+            cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have
+            the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.
+    Returns:
+        `tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding.
+    """
+    cos = cos.unsqueeze(unsqueeze_dim)
+    sin = sin.unsqueeze(unsqueeze_dim)
+    q_embed = (q * cos) + (rotate_half(q) * sin)
+    k_embed = (k * cos) + (rotate_half(k) * sin)
+    return q_embed, k_embed
+
+
+def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
+    """
+    This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch,
+    num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim)
+    """
+    batch, num_key_value_heads, slen, head_dim = hidden_states.shape
+    if n_rep == 1:
+        return hidden_states
+    hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)
+    return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)
+
+
+def eager_attention_forward(
+    module: nn.Module,
+    query: torch.Tensor,
+    key: torch.Tensor,
+    value: torch.Tensor,
+    attention_mask: Optional[torch.Tensor],
+    scaling: float,
+    dropout: float = 0.0,
+    **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:
+        causal_mask = attention_mask[:, :, :, : key_states.shape[-2]]
+        attn_weights = attn_weights + causal_mask
+
+    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
+
+
+class Qwen3MoeAttention(nn.Module):
+    """Multi-headed attention from 'Attention Is All You Need' paper"""
+
+    def __init__(self, config: Qwen3MoeConfig, layer_idx: int):
+        super().__init__()
+        self.config = config
+        self.layer_idx = layer_idx
+        self.head_dim = getattr(config, "head_dim", config.hidden_size // config.num_attention_heads)
+        self.num_key_value_groups = config.num_attention_heads // config.num_key_value_heads
+        self.scaling = self.head_dim**-0.5
+        self.attention_dropout = config.attention_dropout
+        self.is_causal = True
+
+        self.q_proj = nn.Linear(
+            config.hidden_size, config.num_attention_heads * self.head_dim, bias=config.attention_bias
+        )
+        self.k_proj = nn.Linear(
+            config.hidden_size, config.num_key_value_heads * self.head_dim, bias=config.attention_bias
+        )
+        self.v_proj = nn.Linear(
+            config.hidden_size, config.num_key_value_heads * self.head_dim, bias=config.attention_bias
+        )
+        self.o_proj = nn.Linear(
+            config.num_attention_heads * self.head_dim, config.hidden_size, bias=config.attention_bias
+        )
+        self.q_norm = Qwen3MoeRMSNorm(self.head_dim, eps=config.rms_norm_eps)  # unlike olmo, only on the head dim!
+        self.k_norm = Qwen3MoeRMSNorm(self.head_dim, eps=config.rms_norm_eps)  # thus post q_norm does not need reshape
+        self.sliding_window = getattr(config, "sliding_window", None)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        position_embeddings: tuple[torch.Tensor, torch.Tensor],
+        attention_mask: Optional[torch.Tensor],
+        past_key_value: Optional[Cache] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        **kwargs: Unpack[FlashAttentionKwargs],
+    ) -> tuple[torch.Tensor, Optional[torch.Tensor], Optional[tuple[torch.Tensor]]]:
+        input_shape = hidden_states.shape[:-1]
+        hidden_shape = (*input_shape, -1, self.head_dim)
+
+        query_states = self.q_norm(self.q_proj(hidden_states).view(hidden_shape)).transpose(1, 2)
+        key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)
+        value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)
+
+        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:
+            # sin and cos are specific to RoPE models; cache_position needed for the static cache
+            cache_kwargs = {"sin": sin, "cos": cos, "cache_position": cache_position}
+            key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
+
+        if self.config._attn_implementation == "sdpa":
+            key_states = key_states.repeat_interleave(self.num_key_value_groups, dim=1)
+            value_states = value_states.repeat_interleave(self.num_key_value_groups, dim=1)
+            out = F.scaled_dot_product_attention(query_states, key_states, value_states, is_causal=True)
+            out = out.transpose(1, 2).contiguous() #.view(out.shape[0], out.shape[1], -1)
+            attn_output = out.view(out.shape[0], out.shape[1], -1)
+            attn_weights = None
+        else:
+            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.o_proj(attn_output)
+        return attn_output, attn_weights
+
+
+class Qwen3MoeMLP(nn.Module):
+    def __init__(self, config, intermediate_size=None):
+        super().__init__()
+        self.config = config
+        self.hidden_size = config.hidden_size
+        self.intermediate_size = intermediate_size if intermediate_size is not None else config.intermediate_size
+        self.gate_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False)
+        self.up_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False)
+        self.down_proj = nn.Linear(self.intermediate_size, self.hidden_size, bias=False)
+        self.act_fn = ACT2FN[config.hidden_act]
+
+    def forward(self, x):
+        down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))
+        return down_proj
+
+
+class Qwen3MoeSparseMoeBlock(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.num_experts = config.num_experts
+        self.top_k = config.num_experts_per_tok
+        self.norm_topk_prob = config.norm_topk_prob
+
+        # gating
+        self.gate = nn.Linear(config.hidden_size, config.num_experts, bias=False)
+        self.experts = nn.ModuleList(
+            [Qwen3MoeMLP(config, intermediate_size=config.moe_intermediate_size) for _ in range(self.num_experts)]
+        )
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        """ """
+        batch_size, sequence_length, hidden_dim = hidden_states.shape
+        hidden_states = hidden_states.view(-1, hidden_dim)
+        # router_logits: (batch * sequence_length, n_experts)
+        router_logits = self.gate(hidden_states)
+
+        routing_weights = F.softmax(router_logits, dim=1, dtype=torch.float)
+        routing_weights, selected_experts = torch.topk(routing_weights, self.top_k, dim=-1)
+        if self.norm_topk_prob:  # only diff with mixtral sparse moe block!
+            routing_weights /= routing_weights.sum(dim=-1, keepdim=True)
+        # we cast back to the input dtype
+        routing_weights = routing_weights.to(hidden_states.dtype)
+
+        final_hidden_states = torch.zeros(
+            (batch_size * sequence_length, hidden_dim), dtype=hidden_states.dtype, device=hidden_states.device
+        )
+
+        # One hot encode the selected experts to create an expert mask
+        # this will be used to easily index which expert is going to be sollicitated
+        expert_mask = torch.nn.functional.one_hot(selected_experts, num_classes=self.num_experts).permute(2, 1, 0)
+
+        # Loop over all available experts in the model and perform the computation on each expert
+        expert_hitted = torch.greater(expert_mask.sum(dim=(-1, -2)), 0).nonzero()
+        for expert_idx in expert_hitted:
+            expert_layer = self.experts[expert_idx]
+            idx, top_x = torch.where(expert_mask[expert_idx].squeeze(0))
+
+            # Index the correct hidden states and compute the expert hidden state for
+            # the current expert. We need to make sure to multiply the output hidden
+            # states by `routing_weights` on the corresponding tokens (top-1 and top-2)
+            current_state = hidden_states[None, top_x].reshape(-1, hidden_dim)
+            current_hidden_states = expert_layer(current_state) * routing_weights[top_x, idx, None]
+
+            # However `index_add_` only support torch tensors for indexing so we'll use
+            # the `top_x` tensor here.
+            final_hidden_states.index_add_(0, top_x, current_hidden_states.to(hidden_states.dtype))
+        final_hidden_states = final_hidden_states.reshape(batch_size, sequence_length, hidden_dim)
+        return final_hidden_states, router_logits
+
+
+@use_kernel_forward_from_hub("RMSNorm")
+class Qwen3MoeRMSNorm(nn.Module):
+    def __init__(self, hidden_size, eps=1e-6):
+        """
+        Qwen3MoeRMSNorm 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)
+
+    def extra_repr(self):
+        return f"{tuple(self.weight.shape)}, eps={self.variance_epsilon}"
+
+
+class Qwen3MoeDecoderLayer(GradientCheckpointingLayer):
+    def __init__(self, config: Qwen3MoeConfig, layer_idx: int):
+        super().__init__()
+        self.hidden_size = config.hidden_size
+
+        self.self_attn = Qwen3MoeAttention(config, layer_idx)
+
+        moe_args = MoEArgs(
+            num_experts=config.num_experts,
+            num_shared_experts=0,
+            score_func="softmax",
+            route_norm=config.norm_topk_prob,
+            route_scale=1.0,
+            score_before_experts=False,
+            top_k=config.num_experts_per_tok,
+            use_grouped_mm=True,
+            load_balance_coeff=config.load_balance_coeff,
+        )
+
+        if (layer_idx not in config.mlp_only_layers) and (
+            config.num_experts > 0 and (layer_idx + 1) % config.decoder_sparse_step == 0
+        ):
+            self.mlp = MoE(moe_args, dim=config.hidden_size, hidden_dim=config.moe_intermediate_size)
+        else:
+            self.mlp = Qwen3MoeMLP(config, intermediate_size=config.intermediate_size)
+
+        self.input_layernorm = Qwen3MoeRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.post_attention_layernorm = Qwen3MoeRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        position_embeddings: tuple[torch.Tensor, torch.Tensor],
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[tuple[torch.Tensor]] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        **kwargs: Unpack[FlashAttentionKwargs],
+    ) -> 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, sequence_length)` where padding elements are indicated by 0.
+            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_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`).
+            past_key_value (`Tuple(torch.FloatTensor)`, *optional*): cached past key and value projection states
+            cache_position (`torch.LongTensor` of shape `(sequence_length)`, *optional*):
+                Indices depicting the position of the input sequence tokens in the sequence.
+            position_embeddings (`tuple[torch.FloatTensor, torch.FloatTensor]`, *optional*):
+                Tuple containing the cosine and sine positional embeddings of shape `(batch_size, seq_len, head_dim)`,
+                with `head_dim` being the embedding dimension of each attention head.
+            kwargs (`dict`, *optional*):
+                Arbitrary kwargs to be ignored, used for FSDP and other methods that injects code
+                into the model
+        """
+        residual = hidden_states
+
+        hidden_states = self.input_layernorm(hidden_states)
+
+        # Self Attention
+        hidden_states, _ = self.self_attn(
+            hidden_states=hidden_states,
+            position_embeddings=position_embeddings,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_value=past_key_value,
+            cache_position=cache_position,
+            **kwargs,
+        )
+        hidden_states = residual + hidden_states
+
+        # Fully Connected
+        residual = hidden_states
+        hidden_states = self.post_attention_layernorm(hidden_states)
+        hidden_states = self.mlp(hidden_states)
+        # For the MoE layers, we need to unpack
+        if isinstance(hidden_states, tuple):
+            hidden_states, _ = hidden_states
+        hidden_states = residual + hidden_states
+
+        return hidden_states
+
+
+class Qwen3MoeRotaryEmbedding(nn.Module):
+    def __init__(self, config: Qwen3MoeConfig, device=None):
+        super().__init__()
+        # BC: "rope_type" was originally "type"
+        if hasattr(config, "rope_scaling") and isinstance(config.rope_scaling, dict):
+            self.rope_type = config.rope_scaling.get("rope_type", config.rope_scaling.get("type"))
+        else:
+            self.rope_type = "default"
+        self.max_seq_len_cached = config.max_position_embeddings
+        self.original_max_seq_len = config.max_position_embeddings
+
+        self.config = config
+        self.rope_init_fn = ROPE_INIT_FUNCTIONS[self.rope_type]
+
+        inv_freq, self.attention_scaling = self.rope_init_fn(self.config, device)
+        self.register_buffer("inv_freq", inv_freq, persistent=False)
+        self.original_inv_freq = self.inv_freq
+
+    @torch.no_grad()
+    @dynamic_rope_update  # power user: used with advanced RoPE types (e.g. dynamic rope)
+    def forward(self, x, position_ids):
+        inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1).to(x.device)
+        position_ids_expanded = position_ids[:, None, :].float()
+
+        device_type = x.device.type if isinstance(x.device.type, str) and x.device.type != "mps" else "cpu"
+        with torch.autocast(device_type=device_type, enabled=False):  # Force float32
+            freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).transpose(1, 2)
+            emb = torch.cat((freqs, freqs), dim=-1)
+            cos = emb.cos() * self.attention_scaling
+            sin = emb.sin() * self.attention_scaling
+
+        return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)
+
+
+@auto_docstring
+class Qwen3MoePreTrainedModel(PreTrainedModel):
+    config: Qwen3MoeConfig
+    base_model_prefix = "model"
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["Qwen3MoeDecoderLayer"]
+    _skip_keys_device_placement = ["past_key_values"]
+    _supports_flash_attn = True
+    _supports_sdpa = True
+    _supports_flex_attn = True
+    _can_compile_fullgraph = False  # MoE models don't work with torch.compile (`torch.where(condition)` not supported)
+    _supports_attention_backend = True
+    _can_record_outputs = {
+        "router_logits": OutputRecorder(Qwen3MoeSparseMoeBlock, index=1),
+        "hidden_states": Qwen3MoeDecoderLayer,
+        "attentions": Qwen3MoeAttention,
+    }
+
+
+@auto_docstring
+class Qwen3MoeModel(Qwen3MoePreTrainedModel):
+    def __init__(self, config: Qwen3MoeConfig):
+        super().__init__(config)
+        self.padding_idx = config.pad_token_id
+        self.vocab_size = config.vocab_size
+
+        self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size, self.padding_idx)
+        self.layers = nn.ModuleList(
+            [Qwen3MoeDecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
+        )
+        self.norm = Qwen3MoeRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.rotary_emb = Qwen3MoeRotaryEmbedding(config=config)
+        self.gradient_checkpointing = False
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    @check_model_inputs
+    @auto_docstring
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Cache] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        use_cache: Optional[bool] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        **kwargs: Unpack[TransformersKwargs],
+    ) -> MoeModelOutputWithPast:
+        if (input_ids is None) ^ (inputs_embeds is not None):
+            raise ValueError("You must specify exactly one of input_ids or inputs_embeds")
+
+        if use_cache and past_key_values is None:
+            past_key_values = DynamicCache()
+
+        if inputs_embeds is None:
+            inputs_embeds = self.embed_tokens(input_ids)
+
+        if cache_position is None:
+            past_seen_tokens = past_key_values.get_seq_length() if past_key_values is not None else 0
+            cache_position = torch.arange(
+                past_seen_tokens, past_seen_tokens + inputs_embeds.shape[1], device=inputs_embeds.device
+            )
+        if position_ids is None:
+            position_ids = cache_position.unsqueeze(0)
+
+        mask_function = create_causal_mask if self.config.sliding_window is None else create_sliding_window_causal_mask
+        causal_mask = mask_function(
+            config=self.config,
+            input_embeds=inputs_embeds,
+            attention_mask=attention_mask,
+            cache_position=cache_position,
+            past_key_values=past_key_values,
+            position_ids=position_ids,
+        )
+
+        hidden_states = inputs_embeds
+
+        # create position embeddings to be shared across the decoder layers
+        position_embeddings = self.rotary_emb(hidden_states, position_ids)
+
+        for decoder_layer in self.layers[: self.config.num_hidden_layers]:
+            hidden_states = decoder_layer(
+                hidden_states,
+                position_embeddings=position_embeddings,
+                attention_mask=causal_mask,
+                position_ids=position_ids,
+                past_key_value=past_key_values,
+                use_cache=use_cache,
+                cache_position=cache_position,
+                **kwargs,
+            )
+
+        hidden_states = self.norm(hidden_states)
+
+        return MoeModelOutputWithPast(  # only diff with Mistral is the output type, we need MoE
+            last_hidden_state=hidden_states,
+            past_key_values=past_key_values,
+        )
+
+
+def load_balancing_loss_func(
+    gate_logits: Union[torch.Tensor, tuple[torch.Tensor], None],
+    num_experts: Optional[int] = None,
+    top_k=2,
+    attention_mask: Optional[torch.Tensor] = None,
+) -> Union[torch.Tensor, int]:
+    r"""
+    Computes auxiliary load balancing loss as in Switch Transformer - implemented in Pytorch.
+
+    See Switch Transformer (https://huggingface.co/papers/2101.03961) for more details. This function implements the loss
+    function presented in equations (4) - (6) of the paper. It aims at penalizing cases where the routing between
+    experts is too unbalanced.
+
+    Args:
+        gate_logits:
+            Logits from the `gate`, should be a tuple of model.config.num_hidden_layers tensors of
+            shape [batch_size X sequence_length, num_experts].
+        num_experts:
+            Number of experts
+        top_k:
+            The number of experts to route per-token, can be also interpreted as the `top-k` routing
+            parameter.
+        attention_mask (`torch.Tensor`, *optional*):
+            The attention_mask used in forward function
+            shape [batch_size X sequence_length] if not None.
+
+    Returns:
+        The auxiliary loss.
+    """
+    if gate_logits is None or not isinstance(gate_logits, tuple):
+        return 0
+
+    if isinstance(gate_logits, tuple):
+        compute_device = gate_logits[0].device
+        concatenated_gate_logits = torch.cat([layer_gate.to(compute_device) for layer_gate in gate_logits], dim=0)
+
+    routing_weights = torch.nn.functional.softmax(concatenated_gate_logits, dim=-1)
+
+    _, selected_experts = torch.topk(routing_weights, top_k, dim=-1)
+
+    expert_mask = torch.nn.functional.one_hot(selected_experts, num_experts)
+
+    if attention_mask is None:
+        # Compute the percentage of tokens routed to each experts
+        tokens_per_expert = torch.mean(expert_mask.float(), dim=0)
+
+        # Compute the average probability of routing to these experts
+        router_prob_per_expert = torch.mean(routing_weights, dim=0)
+    else:
+        batch_size, sequence_length = attention_mask.shape
+        num_hidden_layers = concatenated_gate_logits.shape[0] // (batch_size * sequence_length)
+
+        # Compute the mask that masks all padding tokens as 0 with the same shape of expert_mask
+        expert_attention_mask = (
+            attention_mask[None, :, :, None, None]
+            .expand((num_hidden_layers, batch_size, sequence_length, top_k, num_experts))
+            .reshape(-1, top_k, num_experts)
+            .to(compute_device)
+        )
+
+        # Compute the percentage of tokens routed to each experts
+        tokens_per_expert = torch.sum(expert_mask.float() * expert_attention_mask, dim=0) / torch.sum(
+            expert_attention_mask, dim=0
+        )
+
+        # Compute the mask that masks all padding tokens as 0 with the same shape of tokens_per_expert
+        router_per_expert_attention_mask = (
+            attention_mask[None, :, :, None]
+            .expand((num_hidden_layers, batch_size, sequence_length, num_experts))
+            .reshape(-1, num_experts)
+            .to(compute_device)
+        )
+
+        # Compute the average probability of routing to these experts
+        router_prob_per_expert = torch.sum(routing_weights * router_per_expert_attention_mask, dim=0) / torch.sum(
+            router_per_expert_attention_mask, dim=0
+        )
+
+    overall_loss = torch.sum(tokens_per_expert * router_prob_per_expert.unsqueeze(0))
+    return overall_loss * num_experts
+
+
+@auto_docstring
+class Qwen3MoeForCausalLM(Qwen3MoePreTrainedModel, GenerationMixin):
+    _tied_weights_keys = ["lm_head.weight"]
+    _tp_plan = {"lm_head": "colwise_rep"}
+    _pp_plan = {"lm_head": (["hidden_states"], ["logits"])}
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.model = Qwen3MoeModel(config)
+        self.vocab_size = config.vocab_size
+        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
+        self.router_aux_loss_coef = config.router_aux_loss_coef
+        self.num_experts = config.num_experts
+        self.num_experts_per_tok = config.num_experts_per_tok
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def set_decoder(self, decoder):
+        self.model = decoder
+
+    def get_decoder(self):
+        return self.model
+
+    @can_return_tuple
+    @auto_docstring
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Cache] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_router_logits: Optional[bool] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        logits_to_keep: Union[int, torch.Tensor] = 0,
+        **kwargs: Unpack[TransformersKwargs],
+    ) -> MoeCausalLMOutputWithPast:
+        r"""
+        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]`.
+
+        Example:
+
+        ```python
+        >>> from transformers import AutoTokenizer, Qwen3MoeForCausalLM
+
+        >>> model = Qwen3MoeForCausalLM.from_pretrained("Qwen/Qwen3-MoE-15B-A2B")
+        >>> tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3-MoE-15B-A2B")
+
+        >>> 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_router_logits = (
+            output_router_logits if output_router_logits is not None else self.config.output_router_logits
+        )
+
+        # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
+        outputs: MoeModelOutputWithPast = 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_router_logits=output_router_logits,
+            cache_position=cache_position,
+            **kwargs,
+        )
+
+        hidden_states = outputs.last_hidden_state
+        # Only compute necessary logits, and do not upcast them to float if we are not computing the loss
+        slice_indices = slice(-logits_to_keep, None) if isinstance(logits_to_keep, int) else logits_to_keep
+        logits = self.lm_head(hidden_states[:, slice_indices, :])
+
+        loss = None
+        if labels is not None:
+            loss = self.loss_function(logits, labels, self.vocab_size, **kwargs)
+
+        aux_loss = None
+        if output_router_logits:
+            aux_loss = load_balancing_loss_func(
+                outputs.router_logits,
+                self.num_experts,
+                self.num_experts_per_tok,
+                attention_mask,
+            )
+            if labels is not None:
+                loss += self.router_aux_loss_coef * aux_loss.to(loss.device)  # make sure to reside in the same device
+
+        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,
+        )
+
+
+class Qwen3MoeForSequenceClassification(GenericForSequenceClassification, Qwen3MoePreTrainedModel):
+    pass
+
+
+class Qwen3MoeForTokenClassification(GenericForTokenClassification, Qwen3MoePreTrainedModel):
+    pass
+
+
+class Qwen3MoeForQuestionAnswering(GenericForQuestionAnswering, Qwen3MoePreTrainedModel):
+    base_model_prefix = "transformer"  # For BC, where `transformer` was used instead of `model`
+
+
+__all__ = [
+    "Qwen3MoeForCausalLM",
+    "Qwen3MoeForQuestionAnswering",
+    "Qwen3MoeModel",
+    "Qwen3MoePreTrainedModel",
+    "Qwen3MoeForSequenceClassification",
+    "Qwen3MoeForTokenClassification",
+]
+

special_tokens_map.json

@@ -0,0 +1,31 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "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:aeb13307a71acd8fe81861d94ad54ab689df773318809eed3cbe794b4492dae4
+size 11422654

tokenizer_config.json

@@ -0,0 +1,239 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151665": {
+      "content": "<tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151666": {
+      "content": "</tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151667": {
+      "content": "<think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151668": {
+      "content": "</think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}