Add NVFP4 quantized checkpoint

README.md

@@ -0,0 +1,27 @@
+---
+datasets:
+- nvidia/OpenCodeInstruct
+base_model:
+- IQuestLab/IQuest-Coder-V1-40B-Instruct
+tags:
+  - nvfp4
+  - fp4
+  - quantized
+---
+# IQuest-Coder-V1-40B-Instruct-nvfp4
+
+**Format:** NVFP4 — weights & activations quantized to FP4 with dual scaling.  
+**Base model:** `IQuestLab/IQuest-Coder-V1-40B-Instruct`  
+**How it was made:** One-shot calibration with LLM Compressor (NVFP4 recipe), long-seq calibration (256 samples of 4096 length) with nvidia/OpenCodeInstruct.  
+
+> Notes: Keep `lm_head` in high precision; calibrate on long, domain-relevant sequences.
+
+Check the original model card for information about this model.
+
+# Running the model with VLLM in Docker
+```sh
+sudo docker run --runtime nvidia --gpus all -p 8000:8000 --ipc=host vllm/vllm-openai:nightly --model Firworks/IQuest-Coder-V1-40B-Instruct-nvfp4 --dtype auto --max-model-len 32768
+```
+This was tested on an RTX Pro 6000 Blackwell cloud instance.
+
+If there are other models you're interested in seeing quantized to NVFP4 for use on the DGX Spark, or other modern Blackwell (or newer) cards let me know. I'm trying to make more NVFP4 models available to allow more people to try them out.

added_tokens.json

@@ -0,0 +1,29 @@
+{
+  "</think>": 75873,
+  "</tool_call>": 75877,
+  "</tool_response>": 75879,
+  "</tools>": 75875,
+  "<CLS>": 75858,
+  "<EOD>": 75860,
+  "<MASK>": 75861,
+  "<PAD>": 75862,
+  "<SEP>": 75859,
+  "<think>": 75872,
+  "<tool_call>": 75876,
+  "<tool_response>": 75878,
+  "<tools>": 75874,
+  "<|CLS|>": 75880,
+  "<|EOD|>": 75882,
+  "<|MASK|>": 75883,
+  "<|PAD|>": 75884,
+  "<|SEP|>": 75881,
+  "<|endoftext|>": 75869,
+  "<|file_sep|>": 75871,
+  "<|fim_middle|>": 75866,
+  "<|fim_pad|>": 75868,
+  "<|fim_prefix|>": 75865,
+  "<|fim_suffix|>": 75867,
+  "<|im_end|>": 75864,
+  "<|im_start|>": 75863,
+  "<|repo_name|>": 75870
+}

chat_template.jinja

@@ -0,0 +1,69 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- else %}
+        {{- 'You are LoopCoder, a helpful assistant developed by IQuest.' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- else %}
+        {{- '<|im_start|>system\nYou are LoopCoder, a helpful assistant developed by IQuest.<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+{%- endif %}

config.json

@@ -0,0 +1,91 @@
+{
+  "architectures": [
+    "IQuestCoderForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "auto_map": {
+    "AutoConfig": "configuration_iquestcoder.IQuestCoderConfig",
+    "AutoModel": "modeling_iquestcoder.IQuestCoderModel",
+    "AutoModelForCausalLM": "modeling_iquestcoder.IQuestCoderForCausalLM",
+    "AutoModelForQuestionAnswering": "modeling_iquestcoder.IQuestCoderForQuestionAnswering",
+    "AutoModelForSequenceClassification": "modeling_iquestcoder.IQuestCoderForSequenceClassification",
+    "AutoModelForTokenClassification": "modeling_iquestcoder.IQuestCoderForTokenClassification"
+  },
+  "bos_token_id": 1,
+  "clip_qkv": null,
+  "eos_token_id": [
+    2,
+    75864,
+    75869
+  ],
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 5120,
+  "initializer_range": 0.02,
+  "intermediate_size": 27648,
+  "max_position_embeddings": 131072,
+  "max_window_layers": 0,
+  "mlp_bias": false,
+  "model_type": "iquestcoder",
+  "num_attention_heads": 40,
+  "num_hidden_layers": 80,
+  "num_key_value_heads": 8,
+  "pretraining_tp": 1,
+  "quantization_config": {
+    "config_groups": {
+      "group_0": {
+        "format": "nvfp4-pack-quantized",
+        "input_activations": {
+          "actorder": null,
+          "block_structure": null,
+          "dynamic": "local",
+          "group_size": 16,
+          "num_bits": 4,
+          "observer": "minmax",
+          "observer_kwargs": {},
+          "strategy": "tensor_group",
+          "symmetric": true,
+          "type": "float"
+        },
+        "output_activations": null,
+        "targets": [
+          "Linear"
+        ],
+        "weights": {
+          "actorder": null,
+          "block_structure": null,
+          "dynamic": false,
+          "group_size": 16,
+          "num_bits": 4,
+          "observer": "minmax",
+          "observer_kwargs": {},
+          "strategy": "tensor_group",
+          "symmetric": true,
+          "type": "float"
+        }
+      }
+    },
+    "format": "nvfp4-pack-quantized",
+    "global_compression_ratio": null,
+    "ignore": [
+      "lm_head"
+    ],
+    "kv_cache_scheme": null,
+    "quant_method": "compressed-tensors",
+    "quantization_status": "compressed",
+    "sparsity_config": {},
+    "transform_config": {},
+    "version": "0.12.2"
+  },
+  "rms_norm_eps": 1e-05,
+  "rope_scaling": null,
+  "rope_theta": 500000.0,
+  "sliding_window": null,
+  "tie_word_embeddings": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.53.3",
+  "use_cache": true,
+  "use_sliding_window": false,
+  "vocab_size": 76800
+}

configuration_iquestcoder.py

@@ -0,0 +1,182 @@
+"""IQuestCoder model configuration."""
+
+from transformers.configuration_utils import PretrainedConfig
+from transformers.utils import logging
+
+
+logger = logging.get_logger(__name__)
+
+
+class IQuestCoderConfig(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`IQuestCoderModel`]. It is used to instantiate
+    an IQuestCoder model according to the specified arguments, defining the model architecture.
+
+    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 76800):
+            Vocabulary size of the IQuestCoder model. Defines the number of different tokens that can be represented
+            by the `inputs_ids` passed when calling [`IQuestCoderModel`].
+        hidden_size (`int`, *optional*, defaults to 5120):
+            Dimension of the hidden representations.
+        intermediate_size (`int`, *optional*, defaults to 27648):
+            Dimension of the MLP representations.
+        num_hidden_layers (`int`, *optional*, defaults to 80):
+            Number of hidden layers in the Transformer decoder.
+        num_attention_heads (`int`, *optional*, defaults to 40):
+            Number of attention heads for each attention layer in the Transformer decoder.
+        num_key_value_heads (`int`, *optional*, defaults to 8):
+            This is the number of key_value heads that should be used to implement Grouped Query Attention (GQA).
+            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).
+        head_dim (`int`, *optional*, defaults to 128):
+            The dimension of each attention head. If not specified, defaults to `hidden_size // num_attention_heads`.
+        hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
+            The non-linear activation function (function or string) in the decoder.
+        max_position_embeddings (`int`, *optional*, defaults to 16384):
+            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-05):
+            The epsilon used by the rms normalization layers.
+        use_cache (`bool`, *optional*, defaults to `True`):
+            Whether or not the model should return the last key/values attentions (not used by all models).
+        pad_token_id (`int`, *optional*):
+            Padding token id.
+        bos_token_id (`int`, *optional*, defaults to 1):
+            Beginning of stream token id.
+        eos_token_id (`int`, *optional*, defaults to 2):
+            End of stream token id.
+        tie_word_embeddings (`bool`, *optional*, defaults to `False`):
+            Whether to tie weight embeddings.
+        rope_theta (`float`, *optional*, defaults to 500000.0):
+            The base period of the RoPE embeddings.
+        rope_scaling (`Dict`, *optional*):
+            Dictionary containing the scaling configuration for the RoPE embeddings. Supports various RoPE scaling
+            types including "linear", "dynamic", "yarn", "longrope", etc.
+        attention_bias (`bool`, *optional*, defaults to `False`):
+            Whether to use a bias in the query, key, value and output projection layers during self-attention.
+        attention_dropout (`float`, *optional*, defaults to 0.0):
+            The dropout ratio for the attention probabilities.
+        mlp_bias (`bool`, *optional*, defaults to `False`):
+            Whether to use a bias in up_proj, down_proj and gate_proj layers in the MLP layers.
+        clip_qkv (`float`, *optional*):
+            If set, clip the query, key, and value tensors to this value. Borrowed from OLMo for training stability.
+        use_sliding_window (`bool`, *optional*, defaults to `False`):
+            Whether to use sliding window attention. Borrowed from Qwen2.
+        sliding_window (`int`, *optional*):
+            The sliding window size. Only effective when `use_sliding_window=True`.
+        max_window_layers (`int`, *optional*, defaults to 0):
+            The number of layers that don't use sliding window attention. Borrowed from Qwen2.
+
+    Example:
+        ```python
+        >>> from configuration_iquestcoder import IQuestCoderConfig
+        >>> from modeling_iquestcoder import IQuestCoderModel
+
+        >>> # Initializing a IQuestCoder configuration
+        >>> configuration = IQuestCoderConfig()
+
+        >>> # Initializing a model from the configuration
+        >>> model = IQuestCoderModel(configuration)
+
+        >>> # Accessing the model configuration
+        >>> configuration = model.config
+        ```
+    """
+
+    model_type = "iquestcoder"
+    keys_to_ignore_at_inference = ["past_key_values"]
+
+    def __init__(
+        self,
+        vocab_size=76800,
+        hidden_size=5120,
+        intermediate_size=27648,
+        num_hidden_layers=80,
+        num_attention_heads=40,
+        num_key_value_heads=8,
+        head_dim=128,
+        hidden_act="silu",
+        max_position_embeddings=16384,
+        initializer_range=0.02,
+        rms_norm_eps=1e-5,
+        use_cache=True,
+        pad_token_id=None,
+        bos_token_id=1,
+        eos_token_id=2,
+        tie_word_embeddings=False,
+        rope_theta=500000.0,
+        rope_scaling=None,
+        attention_bias=False,
+        attention_dropout=0.0,
+        mlp_bias=False,
+        # IQuestCoder specific (borrowed from OLMo)
+        clip_qkv=None,
+        # IQuestCoder specific (borrowed from Qwen2)
+        use_sliding_window=False,
+        sliding_window=None,
+        max_window_layers=0,
+        **kwargs,
+    ):
+        self.vocab_size = vocab_size
+        self.max_position_embeddings = max_position_embeddings
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.num_key_value_heads = num_key_value_heads
+        self.head_dim = head_dim
+        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
+        self.mlp_bias = mlp_bias
+        # IQuestCoder specific
+        self.clip_qkv = clip_qkv
+        self.use_sliding_window = use_sliding_window
+        self.sliding_window = sliding_window
+        self.max_window_layers = max_window_layers
+
+        # Validate rope_scaling configuration
+        self._rope_scaling_validation()
+
+        super().__init__(
+            pad_token_id=pad_token_id,
+            bos_token_id=bos_token_id,
+            eos_token_id=eos_token_id,
+            tie_word_embeddings=tie_word_embeddings,
+            **kwargs,
+        )
+
+    def _rope_scaling_validation(self):
+        """Validate the `rope_scaling` configuration."""
+        if self.rope_scaling is None:
+            return
+
+        if not isinstance(self.rope_scaling, dict) or len(self.rope_scaling) < 1:
+            raise ValueError(
+                "`rope_scaling` must be a dictionary with a minimum of one field, `type` or `rope_type`."
+            )
+        
+        rope_scaling_type = self.rope_scaling.get("type", None) or self.rope_scaling.get("rope_type", None)
+        if rope_scaling_type is None:
+            raise ValueError(
+                "`rope_scaling` must have a `type` or `rope_type` field."
+            )
+        
+        valid_rope_types = ["linear", "dynamic", "yarn", "longrope", "llama3"]
+        if rope_scaling_type not in valid_rope_types:
+            raise ValueError(
+                f"`rope_scaling`'s type field must be one of {valid_rope_types}, got {rope_scaling_type}"
+            )
+
+
+__all__ = ["IQuestCoderConfig"]
+

generation_config.json

@@ -0,0 +1,11 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 1,
+  "do_sample": true,
+  "eos_token_id": [
+    2,
+    75864,
+    75869
+  ],
+  "transformers_version": "4.53.3"
+}

model-00001-of-00005.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:43385a18f3303086da57a8c8d62e4e1a283a8bc08c4befcc5c3f1789e64e90fb
+size 4936205040

model-00002-of-00005.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f2c23098f70fdafc994c496c22d2d5ae0241da736670c863e8a677b52d650f0
+size 4937257880

model-00003-of-00005.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c9a4b6d8b33e9194a3dbb31135aaa9f32ab1893355ee2e26a05d75b505f1e8b
+size 4937257880

model-00004-of-00005.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af2128a0871511daab50e8c924500ceacfe2328343cef67960f9a3370f604eba
+size 4937257880

model-00005-of-00005.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:02c4f8f09aba457aa4fab2bfd026bf8b6cc402f5a8d46fafb690e83238ae5c27
+size 3768264472

modeling_iquestcoder.py

@@ -0,0 +1,1068 @@
+"""
+Modified MIT License 
+
+Software Copyright© 2025 IQuest Research
+
+Our only modification is that, if the Software (or any derivative works
+thereof) is used for any of your commercial products or services, you shall
+prominently display "IQuest Coder" on the user interface of such product or
+service.
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+"""
+
+from typing import Callable, List, Optional, Tuple, Union
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+from transformers.activations import ACT2FN
+from transformers.cache_utils import Cache, DynamicCache, SlidingWindowCache, StaticCache
+from transformers.generation import GenerationMixin
+from transformers.modeling_attn_mask_utils import AttentionMaskConverter
+from transformers.modeling_flash_attention_utils import FlashAttentionKwargs
+from transformers.modeling_layers import GradientCheckpointingLayer
+from transformers.modeling_outputs import (
+    BaseModelOutputWithPast,
+    CausalLMOutputWithPast,
+    QuestionAnsweringModelOutput,
+    SequenceClassifierOutputWithPast,
+    TokenClassifierOutput,
+)
+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 (
+    LossKwargs,
+    auto_docstring,
+    can_return_tuple,
+    is_torch_flex_attn_available,
+    logging,
+)
+
+from .configuration_iquestcoder import IQuestCoderConfig
+
+
+if is_torch_flex_attn_available():
+    from torch.nn.attention.flex_attention import BlockMask
+    from transformers.integrations.flex_attention import make_flex_block_causal_mask
+
+
+logger = logging.get_logger(__name__)
+
+
+# =============================================================================
+# Helper Functions
+# =============================================================================
+
+def rotate_half(x: torch.Tensor) -> torch.Tensor:
+    """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: torch.Tensor,
+    k: torch.Tensor,
+    cos: torch.Tensor,
+    sin: torch.Tensor,
+    position_ids: Optional[torch.Tensor] = None,
+    unsqueeze_dim: int = 1,
+) -> Tuple[torch.Tensor, torch.Tensor]:
+    """Applies Rotary Position Embedding to the query and key tensors.
+
+    Args:
+        q: The query tensor.
+        k: The key tensor.
+        cos: The cosine part of the rotary embedding.
+        sin: The sine part of the rotary embedding.
+        position_ids: Deprecated and unused.
+        unsqueeze_dim: The dimension along which to unsqueeze cos and sin.
+
+    Returns:
+        Tuple of query and key tensors rotated using the Rotary Position Embedding.
+    """
+    # Borrowed from OLMo: preserve original dtypes for numerical stability
+    q_dtype, k_dtype = q.dtype, k.dtype
+    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.to(q_dtype), k_embed.to(k_dtype)
+
+
+def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
+    """
+    Expands key/value heads for Grouped Query Attention.
+    
+    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,
+) -> Tuple[torch.Tensor, torch.Tensor]:
+    """Standard eager attention implementation."""
+    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 = F.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)
+    attn_weights = F.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
+
+
+# =============================================================================
+# Model Components
+# =============================================================================
+
+class IQuestCoderRMSNorm(nn.Module):
+    """Root Mean Square Layer Normalization.
+    
+    RMSNorm is computationally simpler than LayerNorm while achieving similar
+    performance. It normalizes the input by its RMS value.
+    """
+
+    def __init__(self, hidden_size: int, eps: float = 1e-6):
+        super().__init__()
+        self.weight = nn.Parameter(torch.ones(hidden_size))
+        self.variance_epsilon = eps
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        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) -> str:
+        return f"{tuple(self.weight.shape)}, eps={self.variance_epsilon}"
+
+
+class IQuestCoderRotaryEmbedding(nn.Module):
+    """Rotary Position Embedding (RoPE).
+    
+    Implements rotary positional embeddings as described in the RoFormer paper.
+    Supports various RoPE scaling methods for extended context lengths.
+    """
+
+    def __init__(self, config: IQuestCoderConfig, device=None):
+        super().__init__()
+        # BC: "rope_type" was originally "type"
+        if hasattr(config, "rope_scaling") and config.rope_scaling is not None:
+            self.rope_type = config.rope_scaling.get("rope_type", config.rope_scaling.get("type"))
+        else:
+            self.rope_type = "default"
+        self.max_seq_len_cached = config.max_position_embeddings
+        self.original_max_seq_len = config.max_position_embeddings
+
+        self.config = config
+        self.rope_init_fn = ROPE_INIT_FUNCTIONS[self.rope_type]
+
+        inv_freq, self.attention_scaling = self.rope_init_fn(self.config, device)
+        self.register_buffer("inv_freq", inv_freq, persistent=False)
+        self.original_inv_freq = self.inv_freq
+
+    @torch.no_grad()
+    @dynamic_rope_update
+    def forward(self, x: torch.Tensor, position_ids: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:
+        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):
+            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)
+
+
+class IQuestCoderMLP(nn.Module):
+    """Feed-forward network with SwiGLU activation.
+    
+    Uses the gated linear unit variant with SiLU activation for improved
+    performance compared to standard FFN.
+    """
+
+    def __init__(self, config: IQuestCoderConfig):
+        super().__init__()
+        self.config = config
+        self.hidden_size = config.hidden_size
+        self.intermediate_size = config.intermediate_size
+        self.gate_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=config.mlp_bias)
+        self.up_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=config.mlp_bias)
+        self.down_proj = nn.Linear(self.intermediate_size, self.hidden_size, bias=config.mlp_bias)
+        self.act_fn = ACT2FN[config.hidden_act]
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        # SwiGLU: down_proj(act_fn(gate_proj(x)) * up_proj(x))
+        return self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))
+
+
+class IQuestCoderAttention(nn.Module):
+    """Multi-headed attention with support for Grouped Query Attention (GQA).
+    
+    Features:
+    - Grouped Query Attention for memory efficiency
+    - Optional QKV clipping for training stability (from OLMo)
+    - Optional sliding window attention (from Qwen2)
+    - Rotary Position Embeddings
+    """
+
+    def __init__(self, config: IQuestCoderConfig, 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
+
+        # Projection layers
+        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
+        )
+
+    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)
+
+        # Compute Q, K, V projections
+        query_states = self.q_proj(hidden_states)
+        key_states = self.k_proj(hidden_states)
+        value_states = self.v_proj(hidden_states)
+
+        # [OLMo Feature] Optional QKV clipping for training stability
+        if self.config.clip_qkv is not None:
+            query_states = query_states.clamp(min=-self.config.clip_qkv, max=self.config.clip_qkv)
+            key_states = key_states.clamp(min=-self.config.clip_qkv, max=self.config.clip_qkv)
+            value_states = value_states.clamp(min=-self.config.clip_qkv, max=self.config.clip_qkv)
+
+        # Reshape to (batch, heads, seq_len, head_dim)
+        query_states = query_states.view(hidden_shape).transpose(1, 2)
+        key_states = key_states.view(hidden_shape).transpose(1, 2)
+        value_states = value_states.view(hidden_shape).transpose(1, 2)
+
+        # Apply rotary position embeddings
+        cos, sin = position_embeddings
+        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)
+
+        # Update KV cache if provided
+        if past_key_value is not None:
+            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)
+
+        # [Qwen2 Feature] Sliding window attention
+        sliding_window = None
+        if (
+            self.config.use_sliding_window
+            and getattr(self.config, "sliding_window", None) is not None
+            and self.layer_idx >= self.config.max_window_layers
+        ):
+            sliding_window = self.config.sliding_window
+
+        # Select attention implementation
+        attention_interface: Callable = eager_attention_forward
+        if self.config._attn_implementation != "eager":
+            if self.config._attn_implementation == "sdpa" and kwargs.get("output_attentions", False):
+                logger.warning_once(
+                    "`torch.nn.functional.scaled_dot_product_attention` does not support `output_attentions=True`. "
+                    'Falling back to eager attention. This warning can be removed using the argument '
+                    '`attn_implementation="eager"` when loading the model.'
+                )
+            else:
+                attention_interface = ALL_ATTENTION_FUNCTIONS[self.config._attn_implementation]
+
+        # Compute attention
+        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=sliding_window,
+            **kwargs,
+        )
+
+        # Reshape and project output
+        attn_output = attn_output.reshape(*input_shape, -1).contiguous()
+        attn_output = self.o_proj(attn_output)
+
+        return attn_output, attn_weights
+
+
+class IQuestCoderDecoderLayer(GradientCheckpointingLayer):
+    """Transformer decoder layer with pre-normalization.
+    
+    Architecture: Pre-RMSNorm -> Attention -> Residual -> Pre-RMSNorm -> MLP -> Residual
+    """
+
+    def __init__(self, config: IQuestCoderConfig, layer_idx: int):
+        super().__init__()
+        self.hidden_size = config.hidden_size
+        self.self_attn = IQuestCoderAttention(config=config, layer_idx=layer_idx)
+        self.mlp = IQuestCoderMLP(config)
+        self.input_layernorm = IQuestCoderRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.post_attention_layernorm = IQuestCoderRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+
+        # Warn if sliding window is enabled but not properly supported
+        if config.use_sliding_window and config._attn_implementation != "flash_attention_2":
+            logger.warning_once(
+                f"Sliding Window Attention is enabled but not implemented for `{config._attn_implementation}`; "
+                "unexpected results may be encountered."
+            )
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Cache] = None,
+        output_attentions: Optional[bool] = False,
+        use_cache: Optional[bool] = False,
+        cache_position: Optional[torch.LongTensor] = None,
+        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
+        **kwargs: Unpack[FlashAttentionKwargs],
+    ) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]:
+        # Pre-norm + Self Attention
+        residual = hidden_states
+        hidden_states = self.input_layernorm(hidden_states)
+
+        hidden_states, self_attn_weights = self.self_attn(
+            hidden_states=hidden_states,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_value=past_key_value,
+            output_attentions=output_attentions,
+            use_cache=use_cache,
+            cache_position=cache_position,
+            position_embeddings=position_embeddings,
+            **kwargs,
+        )
+        hidden_states = residual + hidden_states
+
+        # Pre-norm + MLP
+        residual = hidden_states
+        hidden_states = self.post_attention_layernorm(hidden_states)
+        hidden_states = self.mlp(hidden_states)
+        hidden_states = residual + hidden_states
+
+        outputs = (hidden_states,)
+        if output_attentions:
+            outputs += (self_attn_weights,)
+
+        return outputs
+
+
+# =============================================================================
+# Base Model
+# =============================================================================
+
+@auto_docstring
+class IQuestCoderPreTrainedModel(PreTrainedModel):
+    """Base class for IQuestCoder models."""
+
+    config_class = IQuestCoderConfig
+    base_model_prefix = "model"
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["IQuestCoderDecoderLayer"]
+    _skip_keys_device_placement = ["past_key_values"]
+    _supports_flash_attn_2 = True
+    _supports_sdpa = True
+    _supports_flex_attn = True
+    _supports_cache_class = True
+    _supports_quantized_cache = True
+    _supports_static_cache = True
+    _supports_attention_backend = True
+
+    def _init_weights(self, module: nn.Module):
+        std = self.config.initializer_range
+        if isinstance(module, nn.Linear):
+            module.weight.data.normal_(mean=0.0, std=std)
+            if module.bias is not None:
+                module.bias.data.zero_()
+        elif isinstance(module, nn.Embedding):
+            module.weight.data.normal_(mean=0.0, std=std)
+            if module.padding_idx is not None:
+                module.weight.data[module.padding_idx].zero_()
+        elif isinstance(module, IQuestCoderRMSNorm):
+            module.weight.data.fill_(1.0)
+
+
+@auto_docstring
+class IQuestCoderModel(IQuestCoderPreTrainedModel):
+    """
+    IQuestCoder Model outputting raw hidden-states without any specific head on top.
+    
+    This model is compatible with LLaMA weights while incorporating features from OLMo and Qwen2.
+    """
+
+    def __init__(self, config: IQuestCoderConfig):
+        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(
+            [IQuestCoderDecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
+        )
+        self.norm = IQuestCoderRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.rotary_emb = IQuestCoderRotaryEmbedding(config=config)
+        self.gradient_checkpointing = False
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self) -> nn.Embedding:
+        return self.embed_tokens
+
+    def set_input_embeddings(self, value: nn.Embedding):
+        self.embed_tokens = value
+
+    @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,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        **flash_attn_kwargs: Unpack[FlashAttentionKwargs],
+    ) -> BaseModelOutputWithPast:
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        use_cache = use_cache if use_cache is not None else self.config.use_cache
+
+        if (input_ids is None) ^ (inputs_embeds is not None):
+            raise ValueError("You must specify exactly one of input_ids or inputs_embeds")
+
+        if self.gradient_checkpointing and self.training and use_cache:
+            logger.warning_once(
+                "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`."
+            )
+            use_cache = False
+
+        if not isinstance(past_key_values, (type(None), Cache)):
+            raise ValueError("The `past_key_values` should be either a `Cache` object or `None`.")
+
+        if inputs_embeds is None:
+            inputs_embeds = self.embed_tokens(input_ids)
+
+        if use_cache and past_key_values is None:
+            past_key_values = DynamicCache()
+
+        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)
+
+        causal_mask = self._update_causal_mask(
+            attention_mask, inputs_embeds, cache_position, past_key_values, output_attentions
+        )
+
+        hidden_states = inputs_embeds
+
+        # Create position embeddings to be shared across the decoder layers
+        position_embeddings = self.rotary_emb(hidden_states, position_ids)
+
+        # Decoder layers
+        all_hidden_states = () if output_hidden_states else None
+        all_self_attns = () if output_attentions else None
+
+        for decoder_layer in self.layers[: self.config.num_hidden_layers]:
+            if output_hidden_states:
+                all_hidden_states += (hidden_states,)
+
+            layer_outputs = decoder_layer(
+                hidden_states,
+                attention_mask=causal_mask,
+                position_ids=position_ids,
+                past_key_value=past_key_values,
+                output_attentions=output_attentions,
+                use_cache=use_cache,
+                cache_position=cache_position,
+                position_embeddings=position_embeddings,
+                **flash_attn_kwargs,
+            )
+
+            hidden_states = layer_outputs[0]
+
+            if output_attentions:
+                all_self_attns += (layer_outputs[1],)
+
+        hidden_states = self.norm(hidden_states)
+
+        # Add hidden states from the last decoder layer
+        if output_hidden_states:
+            all_hidden_states += (hidden_states,)
+
+        return BaseModelOutputWithPast(
+            last_hidden_state=hidden_states,
+            past_key_values=past_key_values if use_cache else None,
+            hidden_states=all_hidden_states,
+            attentions=all_self_attns,
+        )
+
+    def _update_causal_mask(
+        self,
+        attention_mask: Union[torch.Tensor, "BlockMask"],
+        input_tensor: torch.Tensor,
+        cache_position: torch.Tensor,
+        past_key_values: Cache,
+        output_attentions: bool = False,
+    ):
+        if self.config._attn_implementation == "flash_attention_2":
+            if attention_mask is not None and past_key_values is not None:
+                is_padding_right = attention_mask[:, -1].sum().item() != input_tensor.size()[0]
+                if is_padding_right:
+                    raise ValueError(
+                        "You are attempting to perform batched generation with padding_side='right'. "
+                        "This may lead to unexpected behaviour for Flash Attention version of IQuestCoder. "
+                        "Make sure to call `tokenizer.padding_side = 'left'` before tokenizing the input."
+                    )
+            if attention_mask is not None and 0.0 in attention_mask:
+                return attention_mask
+            return None
+
+        if self.config._attn_implementation == "flex_attention":
+            if isinstance(attention_mask, torch.Tensor):
+                attention_mask = make_flex_block_causal_mask(attention_mask)
+            return attention_mask
+
+        past_seen_tokens = past_key_values.get_seq_length() if past_key_values is not None else 0
+        using_static_cache = isinstance(past_key_values, StaticCache)
+        using_sliding_window_cache = isinstance(past_key_values, SlidingWindowCache)
+
+        if (
+            self.config._attn_implementation == "sdpa"
+            and not (using_static_cache or using_sliding_window_cache)
+            and not output_attentions
+        ):
+            if AttentionMaskConverter._ignore_causal_mask_sdpa(
+                attention_mask,
+                inputs_embeds=input_tensor,
+                past_key_values_length=past_seen_tokens,
+                sliding_window=self.config.sliding_window if self.config.use_sliding_window else None,
+                is_training=self.training,
+            ):
+                return None
+
+        dtype = input_tensor.dtype
+        min_dtype = torch.finfo(dtype).min
+        sequence_length = input_tensor.shape[1]
+
+        if using_sliding_window_cache or using_static_cache:
+            target_length = past_key_values.get_max_cache_shape()
+        else:
+            target_length = (
+                attention_mask.shape[-1]
+                if isinstance(attention_mask, torch.Tensor)
+                else past_seen_tokens + sequence_length + 1
+            )
+
+        causal_mask = self._prepare_4d_causal_attention_mask_with_cache_position(
+            attention_mask,
+            sequence_length=sequence_length,
+            target_length=target_length,
+            dtype=dtype,
+            cache_position=cache_position,
+            batch_size=input_tensor.shape[0],
+            config=self.config,
+            past_key_values=past_key_values,
+        )
+
+        if (
+            self.config._attn_implementation == "sdpa"
+            and attention_mask is not None
+            and attention_mask.device.type in ["cuda", "xpu", "npu"]
+            and not output_attentions
+        ):
+            causal_mask = AttentionMaskConverter._unmask_unattended(causal_mask, min_dtype)
+
+        return causal_mask
+
+    @staticmethod
+    def _prepare_4d_causal_attention_mask_with_cache_position(
+        attention_mask: torch.Tensor,
+        sequence_length: int,
+        target_length: int,
+        dtype: torch.dtype,
+        cache_position: torch.Tensor,
+        batch_size: int,
+        config: IQuestCoderConfig,
+        past_key_values: Cache,
+    ):
+        """Creates a causal 4D mask from a 2D mask, or returns the 4D mask if already provided."""
+        if attention_mask is not None and attention_mask.dim() == 4:
+            causal_mask = attention_mask
+        else:
+            min_dtype = torch.finfo(dtype).min
+            causal_mask = torch.full(
+                (sequence_length, target_length), fill_value=min_dtype, dtype=dtype, device=cache_position.device
+            )
+            diagonal_attend_mask = torch.arange(target_length, device=cache_position.device) > cache_position.reshape(
+                -1, 1
+            )
+            
+            # [Qwen2 Feature] Handle sliding window mask
+            if getattr(config, "use_sliding_window", False) and config.sliding_window is not None:
+                if not isinstance(past_key_values, SlidingWindowCache) or sequence_length > target_length:
+                    sliding_attend_mask = torch.arange(target_length, device=cache_position.device) <= (
+                        cache_position.reshape(-1, 1) - config.sliding_window
+                    )
+                    diagonal_attend_mask.bitwise_or_(sliding_attend_mask)
+            
+            causal_mask *= diagonal_attend_mask
+            causal_mask = causal_mask[None, None, :, :].expand(batch_size, 1, -1, -1)
+            
+            if attention_mask is not None:
+                causal_mask = causal_mask.clone()
+                if attention_mask.shape[-1] > target_length:
+                    attention_mask = attention_mask[:, :target_length]
+                mask_length = attention_mask.shape[-1]
+                padding_mask = causal_mask[:, :, :, :mask_length] + attention_mask[:, None, None, :].to(
+                    causal_mask.device
+                )
+                padding_mask = padding_mask == 0
+                causal_mask[:, :, :, :mask_length] = causal_mask[:, :, :, :mask_length].masked_fill(
+                    padding_mask, min_dtype
+                )
+
+        return causal_mask
+
+
+# =============================================================================
+# Model Heads
+# =============================================================================
+
+class KwargsForCausalLM(FlashAttentionKwargs, LossKwargs):
+    ...
+
+
+@auto_docstring
+class IQuestCoderForCausalLM(IQuestCoderPreTrainedModel, GenerationMixin):
+    """IQuestCoder Model with a language modeling head on top for causal LM."""
+
+    _tied_weights_keys = ["lm_head.weight"]
+    _tp_plan = {"lm_head": "colwise_rep"}
+    _pp_plan = {"lm_head": (["hidden_states"], ["logits"])}
+
+    def __init__(self, config: IQuestCoderConfig):
+        super().__init__(config)
+        self.model = IQuestCoderModel(config)
+        self.vocab_size = config.vocab_size
+        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self) -> nn.Embedding:
+        return self.model.embed_tokens
+
+    def set_input_embeddings(self, value: nn.Embedding):
+        self.model.embed_tokens = value
+
+    def get_output_embeddings(self) -> nn.Linear:
+        return self.lm_head
+
+    def set_output_embeddings(self, new_embeddings: nn.Linear):
+        self.lm_head = new_embeddings
+
+    def set_decoder(self, decoder: IQuestCoderModel):
+        self.model = decoder
+
+    def get_decoder(self) -> IQuestCoderModel:
+        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_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        logits_to_keep: Union[int, torch.Tensor] = 0,
+        **kwargs: Unpack[KwargsForCausalLM],
+    ) -> CausalLMOutputWithPast:
+        r"""
+        Args:
+            labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+                Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
+                config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
+                (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`.
+
+        Example:
+            ```python
+            >>> from transformers import AutoTokenizer
+            >>> from modeling_iquestcoder import IQuestCoderForCausalLM
+
+            >>> model = IQuestCoderForCausalLM.from_pretrained("path/to/IQuestCoder")
+            >>> tokenizer = AutoTokenizer.from_pretrained("path/to/IQuestCoder")
+
+            >>> prompt = "Hey, are you conscious? Can you talk to me?"
+            >>> inputs = tokenizer(prompt, return_tensors="pt")
+
+            >>> # Generate
+            >>> generate_ids = model.generate(inputs.input_ids, max_length=30)
+            >>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
+            "Hey, are you conscious? Can you talk to me?\\nI'm not conscious, but I can talk to you."
+            ```
+        """
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+
+        # Decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
+        outputs: BaseModelOutputWithPast = self.model(
+            input_ids=input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            cache_position=cache_position,
+            **kwargs,
+        )
+
+        hidden_states = outputs.last_hidden_state
+        # Only compute necessary logits, and do not upcast them to float if we are not computing the loss
+        slice_indices = slice(-logits_to_keep, None) if isinstance(logits_to_keep, int) else logits_to_keep
+        logits = self.lm_head(hidden_states[:, slice_indices, :])
+
+        loss = None
+        if labels is not None:
+            loss = self.loss_function(logits=logits, labels=labels, vocab_size=self.config.vocab_size, **kwargs)
+
+        return CausalLMOutputWithPast(
+            loss=loss,
+            logits=logits,
+            past_key_values=outputs.past_key_values,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+
+@auto_docstring(
+    custom_intro="""
+    The IQuestCoder Model transformer with a sequence classification head on top (linear layer).
+
+    [`IQuestCoderForSequenceClassification`] uses the last token in order to do the classification, as other causal
+    models (e.g. GPT-2) do.
+
+    Since it does classification on the last token, it requires to know the position of the last token. If a
+    `pad_token_id` is defined in the configuration, it finds the last token that is not a padding token in each row.
+    If no `pad_token_id` is defined, it simply takes the last value in each row of the batch.
+    """
+)
+class IQuestCoderForSequenceClassification(IQuestCoderPreTrainedModel):
+    """IQuestCoder Model with a sequence classification head."""
+
+    def __init__(self, config: IQuestCoderConfig):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+        self.model = IQuestCoderModel(config)
+        self.score = nn.Linear(config.hidden_size, self.num_labels, bias=False)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self) -> nn.Embedding:
+        return self.model.embed_tokens
+
+    def set_input_embeddings(self, value: nn.Embedding):
+        self.model.embed_tokens = value
+
+    @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_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+    ) -> SequenceClassifierOutputWithPast:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
+            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
+            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss),
+            If `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
+        """
+        transformer_outputs: BaseModelOutputWithPast = self.model(
+            input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+        )
+        hidden_states = transformer_outputs.last_hidden_state
+        logits = self.score(hidden_states)
+
+        if input_ids is not None:
+            batch_size = input_ids.shape[0]
+        else:
+            batch_size = inputs_embeds.shape[0]
+
+        if self.config.pad_token_id is None and batch_size != 1:
+            raise ValueError("Cannot handle batch sizes > 1 if no padding token is defined.")
+        if self.config.pad_token_id is None:
+            last_non_pad_token = -1
+        elif input_ids is not None:
+            non_pad_mask = (input_ids != self.config.pad_token_id).to(logits.device, torch.int32)
+            token_indices = torch.arange(input_ids.shape[-1], device=logits.device, dtype=torch.int32)
+            last_non_pad_token = (token_indices * non_pad_mask).argmax(-1)
+        else:
+            last_non_pad_token = -1
+            logger.warning_once(
+                f"{self.__class__.__name__} will not detect padding tokens in `inputs_embeds`. Results may be "
+                "unexpected if using padding tokens in conjunction with `inputs_embeds.`"
+            )
+
+        pooled_logits = logits[torch.arange(batch_size, device=logits.device), last_non_pad_token]
+
+        loss = None
+        if labels is not None:
+            loss = self.loss_function(logits=logits, labels=labels, pooled_logits=pooled_logits, config=self.config)
+
+        return SequenceClassifierOutputWithPast(
+            loss=loss,
+            logits=pooled_logits,
+            past_key_values=transformer_outputs.past_key_values,
+            hidden_states=transformer_outputs.hidden_states,
+            attentions=transformer_outputs.attentions,
+        )
+
+
+@auto_docstring
+class IQuestCoderForTokenClassification(IQuestCoderPreTrainedModel):
+    """IQuestCoder Model with a token classification head."""
+
+    def __init__(self, config: IQuestCoderConfig):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+        self.model = IQuestCoderModel(config)
+        if getattr(config, "classifier_dropout", None) is not None:
+            classifier_dropout = config.classifier_dropout
+        elif getattr(config, "hidden_dropout", None) is not None:
+            classifier_dropout = config.hidden_dropout
+        else:
+            classifier_dropout = 0.1
+        self.dropout = nn.Dropout(classifier_dropout)
+        self.score = nn.Linear(config.hidden_size, config.num_labels)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self) -> nn.Embedding:
+        return self.model.embed_tokens
+
+    def set_input_embeddings(self, value: nn.Embedding):
+        self.model.embed_tokens = value
+
+    @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_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+    ) -> TokenClassifierOutput:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
+            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
+            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss),
+            If `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
+        """
+        outputs: BaseModelOutputWithPast = self.model(
+            input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+        )
+        sequence_output = outputs.last_hidden_state
+        sequence_output = self.dropout(sequence_output)
+        logits = self.score(sequence_output)
+
+        loss = None
+        if labels is not None:
+            loss = self.loss_function(logits, labels, self.config)
+
+        return TokenClassifierOutput(
+            loss=loss,
+            logits=logits,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+
+@auto_docstring
+class IQuestCoderForQuestionAnswering(IQuestCoderPreTrainedModel):
+    """IQuestCoder Model with a span classification head for extractive question-answering."""
+
+    base_model_prefix = "transformer"
+
+    def __init__(self, config: IQuestCoderConfig):
+        super().__init__(config)
+        self.transformer = IQuestCoderModel(config)
+        self.qa_outputs = nn.Linear(config.hidden_size, 2)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self) -> nn.Embedding:
+        return self.transformer.embed_tokens
+
+    def set_input_embeddings(self, value: nn.Embedding):
+        self.transformer.embed_tokens = value
+
+    @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,
+        start_positions: Optional[torch.LongTensor] = None,
+        end_positions: Optional[torch.LongTensor] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        **kwargs,
+    ) -> QuestionAnsweringModelOutput:
+        outputs: BaseModelOutputWithPast = self.transformer(
+            input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+        )
+
+        sequence_output = outputs.last_hidden_state
+
+        logits = self.qa_outputs(sequence_output)
+        start_logits, end_logits = logits.split(1, dim=-1)
+        start_logits = start_logits.squeeze(-1).contiguous()
+        end_logits = end_logits.squeeze(-1).contiguous()
+
+        loss = None
+        if start_positions is not None and end_positions is not None:
+            loss = self.loss_function(start_logits, end_logits, start_positions, end_positions, **kwargs)
+
+        return QuestionAnsweringModelOutput(
+            loss=loss,
+            start_logits=start_logits,
+            end_logits=end_logits,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+
+__all__ = [
+    "IQuestCoderPreTrainedModel",
+    "IQuestCoderModel",
+    "IQuestCoderForCausalLM",
+    "IQuestCoderForSequenceClassification",
+    "IQuestCoderForTokenClassification",
+    "IQuestCoderForQuestionAnswering",
+]
+

recipe.yaml

@@ -0,0 +1,7 @@
+default_stage:
+  default_modifiers:
+    QuantizationModifier:
+      targets: [Linear]
+      ignore: [lm_head, 're:visual.*', 're:.*vision_tower.*', 're:.*video_tower.*', 're:.*audio_tower.*',
+        're:.*multi_modal_projector.*']
+      scheme: NVFP4

special_tokens_map.json

@@ -0,0 +1,48 @@
+{
+  "additional_special_tokens": [
+    "<|CLS|>",
+    "<|SEP|>",
+    "<|EOD|>",
+    "<|MASK|>",
+    "<|PAD|>",
+    "<|fim_prefix|>",
+    "<|fim_middle|>",
+    "<|fim_suffix|>",
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|fim_pad|>",
+    "<|endoftext|>",
+    "<|repo_name|>",
+    "<|file_sep|>",
+    "<think>",
+    "</think>"
+  ],
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": true
+  }
+}

tokenization_iquestcoder.py

@@ -0,0 +1,552 @@
+"""Tokenization classes for IQuestCoder."""
+
+import os
+from shutil import copyfile
+from typing import Any, Dict, List, Optional, Tuple, Union
+
+import sentencepiece as spm
+
+from transformers.tokenization_utils import AddedToken, PreTrainedTokenizer
+from transformers.utils import logging
+
+
+logger = logging.get_logger(__name__)
+
+VOCAB_FILES_NAMES = {"vocab_file": "tokenizer.model"}
+
+PRETRAINED_VOCAB_FILES_MAP = {
+    "vocab_file": {},
+    "tokenizer_file": {},
+}
+PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES = {}
+
+
+
+class IQuestCoderTokenizer(PreTrainedTokenizer):
+
+    vocab_files_names = VOCAB_FILES_NAMES
+    pretrained_vocab_files_map = PRETRAINED_VOCAB_FILES_MAP
+    max_model_input_sizes = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
+    model_input_names = ["input_ids", "attention_mask"]
+
+    def __init__(
+        self,
+        vocab_file,
+        unk_token="<unk>",
+        bos_token="<s>",
+        eos_token="</s>",
+        pad_token=None,
+        sp_model_kwargs: Optional[Dict[str, Any]] = None,
+        add_bos_token=True,
+        add_eos_token=False,
+        clean_up_tokenization_spaces=False,
+        add_prefix_space=False,
+        legacy=None,
+        use_default_system_prompt=False,
+        chat_template=None,
+        **kwargs,
+    ):
+        self.sp_model_kwargs = {} if sp_model_kwargs is None else sp_model_kwargs
+        bos_token = AddedToken(bos_token, lstrip=False, rstrip=False) if isinstance(bos_token, str) else bos_token
+        eos_token = AddedToken(eos_token, lstrip=False, rstrip=False) if isinstance(eos_token, str) else eos_token
+        unk_token = AddedToken(unk_token, lstrip=False, rstrip=False) if isinstance(unk_token, str) else unk_token
+        pad_token = AddedToken(pad_token, lstrip=False, rstrip=False) if isinstance(pad_token, str) else pad_token
+        
+        # Legacy behavior handling
+        if legacy is None:
+            logger.warning_once(
+                f"You are using the default legacy behaviour of the {self.__class__.__name__}. This is"
+                " expected, and simply means that the `legacy` (previous) behavior will be used so nothing changes for you."
+                " If you want to use the new behaviour, set `legacy=False`. This should only be set if you understand what it"
+                " means, and thoroughly read the reason why this was added as explained in"
+                " https://github.com/huggingface/transformers/pull/24565"
+            )
+            legacy = True
+        
+        self.legacy = legacy
+        self.vocab_file = vocab_file
+        self.add_bos_token = add_bos_token
+        self.add_eos_token = add_eos_token
+        self.add_prefix_space = add_prefix_space
+        self.use_default_system_prompt = use_default_system_prompt
+        self.sp_model = spm.SentencePieceProcessor(**self.sp_model_kwargs)
+        self.sp_model.Load(vocab_file)
+        
+
+        
+        super().__init__(
+            bos_token=bos_token,
+            eos_token=eos_token,
+            unk_token=unk_token,
+            pad_token=pad_token,
+            add_bos_token=add_bos_token,
+            add_eos_token=add_eos_token,
+            sp_model_kwargs=self.sp_model_kwargs,
+            clean_up_tokenization_spaces=clean_up_tokenization_spaces,
+            add_prefix_space=add_prefix_space,
+            legacy=legacy,
+            use_default_system_prompt=use_default_system_prompt,
+            chat_template=chat_template,
+            **kwargs,
+        )
+
+    def __getstate__(self):
+        state = self.__dict__.copy()
+        state["sp_model"] = None
+        return state
+
+    def __setstate__(self, d):
+        self.__dict__ = d
+        self.sp_model = spm.SentencePieceProcessor(**self.sp_model_kwargs)
+        self.sp_model.Load(self.vocab_file)
+
+    @property
+    def vocab_size(self) -> int:
+        """Returns the vocabulary size."""
+        return self.sp_model.get_piece_size()
+
+    def get_vocab(self) -> Dict[str, int]:
+        """Returns the vocabulary as a dictionary of token to index."""
+        vocab = {self.convert_ids_to_tokens(i): i for i in range(self.vocab_size)}
+        vocab.update(self.added_tokens_encoder)
+        return vocab
+
+    def _tokenize(self, text: str) -> List[str]:
+        """
+        Tokenize a string.
+        
+        Args:
+            text (`str`): The text to tokenize.
+            
+        Returns:
+            `List[str]`: The list of tokens.
+        """
+        if self.add_prefix_space:
+            text = " " + text
+            
+        if self.legacy:
+            return self.sp_model.encode(text, out_type=str)
+        
+        # Non-legacy behavior: handle special tokens properly
+        return self.sp_model.encode(text, out_type=str)
+
+    def _convert_token_to_id(self, token: str) -> int:
+        """Converts a token (str) to an id using the vocab."""
+        return self.sp_model.piece_to_id(token)
+
+    def _convert_id_to_token(self, index: int) -> str:
+        """Converts an index (integer) to a token (str) using the vocab."""
+        token = self.sp_model.IdToPiece(index)
+        return token
+
+    def convert_tokens_to_string(self, tokens: List[str]) -> str:
+        """
+        Converts a sequence of tokens (strings) to a single string.
+        
+        This method handles special tokens separately to ensure they are not
+        decoded using the SentencePiece model.
+        
+        Args:
+            tokens (`List[str]`): The list of tokens to convert.
+            
+        Returns:
+            `str`: The decoded string.
+        """
+        current_sub_tokens = []
+        out_string = ""
+        prev_is_special = False
+        for i, token in enumerate(tokens):
+            # make sure that special tokens are not decoded using sentencepiece model
+            if token in self.all_special_tokens:
+                if not prev_is_special and i != 0:
+                    out_string += " "
+                out_string += self.sp_model.decode(current_sub_tokens) + token
+                prev_is_special = True
+                current_sub_tokens = []
+            else:
+                current_sub_tokens.append(token)
+                prev_is_special = False
+        out_string += self.sp_model.decode(current_sub_tokens)
+        return out_string
+
+    def save_vocabulary(self, save_directory: str, filename_prefix: Optional[str] = None) -> Tuple[str]:
+        """
+        Save the vocabulary and special tokens file to a directory.
+        
+        Args:
+            save_directory (`str`):
+                The directory in which to save the vocabulary.
+            filename_prefix (`str`, *optional*):
+                An optional prefix to add to the named of the saved files.
+                
+        Returns:
+            `Tuple(str)`: Paths to the files saved.
+        """
+        if not os.path.isdir(save_directory):
+            logger.error(f"Vocabulary path ({save_directory}) should be a directory")
+            return
+        out_vocab_file = os.path.join(
+            save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
+        )
+
+        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file) and os.path.isfile(self.vocab_file):
+            copyfile(self.vocab_file, out_vocab_file)
+        elif not os.path.isfile(self.vocab_file):
+            with open(out_vocab_file, "wb") as fi:
+                content_spiece_model = self.sp_model.serialized_model_proto()
+                fi.write(content_spiece_model)
+
+        return (out_vocab_file,)
+
+    def build_inputs_with_special_tokens(
+        self, 
+        token_ids_0: List[int], 
+        token_ids_1: Optional[List[int]] = None
+    ) -> List[int]:
+        """
+        Build model inputs from a sequence or a pair of sequences for sequence classification tasks by concatenating
+        and adding special tokens.
+        
+        An IQuestCoder sequence has the following format:
+        
+        - single sequence: `<s> X </s>` (if add_eos_token is True) or `<s> X` (default)
+        - pair of sequences: `<s> A </s> <s> B </s>` (if add_eos_token is True) or `<s> A <s> B` (default)
+        
+        Args:
+            token_ids_0 (`List[int]`):
+                List of IDs to which the special tokens will be added.
+            token_ids_1 (`List[int]`, *optional*):
+                Optional second list of IDs for sequence pairs.
+                
+        Returns:
+            `List[int]`: List of input IDs with the appropriate special tokens.
+        """
+        bos_token_id = [self.bos_token_id] if self.add_bos_token else []
+        eos_token_id = [self.eos_token_id] if self.add_eos_token else []
+
+        output = bos_token_id + token_ids_0 + eos_token_id
+
+        if token_ids_1 is not None:
+            output = output + bos_token_id + token_ids_1 + eos_token_id
+
+        return output
+
+    def get_special_tokens_mask(
+        self, 
+        token_ids_0: List[int], 
+        token_ids_1: Optional[List[int]] = None, 
+        already_has_special_tokens: bool = False
+    ) -> List[int]:
+        """
+        Retrieve sequence ids from a token list that has no special tokens added. This method is called when adding
+        special tokens using the tokenizer `prepare_for_model` method.
+        
+        Args:
+            token_ids_0 (`List[int]`):
+                List of IDs.
+            token_ids_1 (`List[int]`, *optional*):
+                Optional second list of IDs for sequence pairs.
+            already_has_special_tokens (`bool`, *optional*, defaults to `False`):
+                Whether or not the token list is already formatted with special tokens for the model.
+                
+        Returns:
+            `List[int]`: A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token.
+        """
+        if already_has_special_tokens:
+            return super().get_special_tokens_mask(
+                token_ids_0=token_ids_0, token_ids_1=token_ids_1, already_has_special_tokens=True
+            )
+
+        bos_token_id = [1] if self.add_bos_token else []
+        eos_token_id = [1] if self.add_eos_token else []
+
+        if token_ids_1 is None:
+            return bos_token_id + ([0] * len(token_ids_0)) + eos_token_id
+        return (
+            bos_token_id
+            + ([0] * len(token_ids_0))
+            + eos_token_id
+            + bos_token_id
+            + ([0] * len(token_ids_1))
+            + eos_token_id
+        )
+
+    def create_token_type_ids_from_sequences(
+        self, 
+        token_ids_0: List[int], 
+        token_ids_1: Optional[List[int]] = None
+    ) -> List[int]:
+        """
+        Create a mask from the two sequences passed to be used in a sequence-pair classification task.
+        
+        An IQuestCoder sequence pair mask has the following format:
+        
+        ```
+        0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1
+        | first sequence    | second sequence |
+        ```
+        
+        If `token_ids_1` is `None`, this method only returns the first portion of the mask (0s).
+        
+        Args:
+            token_ids_0 (`List[int]`):
+                List of IDs.
+            token_ids_1 (`List[int]`, *optional*):
+                Optional second list of IDs for sequence pairs.
+                
+        Returns:
+            `List[int]`: List of token type IDs according to the given sequence(s).
+        """
+        bos_token_id = [self.bos_token_id] if self.add_bos_token else []
+        eos_token_id = [self.eos_token_id] if self.add_eos_token else []
+
+        output = [0] * len(bos_token_id + token_ids_0 + eos_token_id)
+
+        if token_ids_1 is not None:
+            output += [1] * len(bos_token_id + token_ids_1 + eos_token_id)
+
+        return output
+
+    @property
+    def default_chat_template(self) -> str:
+        """
+        Returns the default chat template for IQuestCoder.
+        
+        This template formats conversations with system, user, and assistant roles.
+        """
+        return DEFAULT_CHAT_TEMPLATE
+
+    def apply_chat_template(
+        self,
+        conversation: Union[List[Dict[str, str]], "Conversation"],
+        chat_template: Optional[str] = None,
+        add_generation_prompt: bool = False,
+        tokenize: bool = True,
+        padding: bool = False,
+        truncation: bool = False,
+        max_length: Optional[int] = None,
+        return_tensors: Optional[str] = None,
+        return_dict: bool = False,
+        **tokenizer_kwargs,
+    ):
+        """
+        Apply a chat template to format a conversation.
+        
+        Args:
+            conversation (`List[Dict[str, str]]` or `Conversation`):
+                A list of dicts with "role" and "content" keys, representing the conversation history.
+            chat_template (`str`, *optional*):
+                A Jinja template to use for formatting. If not provided, the tokenizer's default will be used.
+            add_generation_prompt (`bool`, *optional*, defaults to `False`):
+                Whether to add a generation prompt at the end for the assistant to continue.
+            tokenize (`bool`, *optional*, defaults to `True`):
+                Whether to tokenize the output. If `False`, returns a string.
+            padding (`bool`, *optional*, defaults to `False`):
+                Whether to pad sequences.
+            truncation (`bool`, *optional*, defaults to `False`):
+                Whether to truncate sequences.
+            max_length (`int`, *optional*):
+                Maximum length of the output.
+            return_tensors (`str`, *optional*):
+                The type of tensors to return ("pt", "tf", "np", or None).
+            return_dict (`bool`, *optional*, defaults to `False`):
+                Whether to return a dictionary with additional information.
+            **tokenizer_kwargs:
+                Additional keyword arguments passed to the tokenizer.
+                
+        Returns:
+            `Union[str, List[int], BatchEncoding]`: The formatted (and optionally tokenized) conversation.
+            
+        Example:
+            ```python
+            >>> tokenizer = IQuestCoderTokenizer.from_pretrained("path/to/model")
+            >>> conversation = [
+            ...     {"role": "system", "content": "You are a helpful assistant."},
+            ...     {"role": "user", "content": "Hello!"},
+            ...     {"role": "assistant", "content": "Hi there! How can I help you today?"},
+            ...     {"role": "user", "content": "What's the weather like?"},
+            ... ]
+            >>> tokenizer.apply_chat_template(conversation, add_generation_prompt=True, tokenize=False)
+            '<|system|>\\nYou are a helpful assistant.\\n</|system|><|user|>\\nHello!\\n</|user|>...'
+            ```
+        """
+        # Use parent class implementation with our template
+        return super().apply_chat_template(
+            conversation,
+            chat_template=chat_template,
+            add_generation_prompt=add_generation_prompt,
+            tokenize=tokenize,
+            padding=padding,
+            truncation=truncation,
+            max_length=max_length,
+            return_tensors=return_tensors,
+            return_dict=return_dict,
+            **tokenizer_kwargs,
+        )
+
+
+# Try to import and create Fast tokenizer version
+try:
+    from transformers import PreTrainedTokenizerFast
+    from tokenizers import Tokenizer, decoders, models, normalizers, pre_tokenizers, processors
+    
+    class IQuestCoderTokenizerFast(PreTrainedTokenizerFast):
+        """
+        Construct a "fast" IQuestCoder tokenizer (backed by HuggingFace's *tokenizers* library).
+        
+        This is a fast implementation of [`IQuestCoderTokenizer`] using the 🤗 Tokenizers library.
+        
+        Args:
+            vocab_file (`str`, *optional*):
+                Path to the vocabulary file (SentencePiece model).
+            tokenizer_file (`str`, *optional*):
+                Path to a tokenizer JSON file.
+            unk_token (`str`, *optional*, defaults to `"<unk>"`):
+                The unknown token.
+            bos_token (`str`, *optional*, defaults to `"<s>"`):
+                The beginning of sequence token.
+            eos_token (`str`, *optional*, defaults to `"</s>"`):
+                The end of sequence token.
+            pad_token (`str`, *optional*):
+                The token used for padding.
+            add_bos_token (`bool`, *optional*, defaults to `True`):
+                Whether to add a BOS token at the start of sequences.
+            add_eos_token (`bool`, *optional*, defaults to `False`):
+                Whether to add an EOS token at the end of sequences.
+            add_prefix_space (`bool`, *optional*, defaults to `False`):
+                Whether to add an initial space to the input.
+            use_default_system_prompt (`bool`, *optional*, defaults to `False`):
+                Whether to use the default system prompt.
+            chat_template (`str`, *optional*):
+                A Jinja template for formatting conversations.
+                
+        Example:
+            ```python
+            >>> from tokenization_iquestcoder import IQuestCoderTokenizerFast
+            
+            >>> tokenizer = IQuestCoderTokenizerFast.from_pretrained("path/to/model")
+            >>> tokenizer.encode("Hello, world!")
+            [1, 15043, 29892, 3186, 29991]
+            ```
+        """
+        
+        vocab_files_names = VOCAB_FILES_NAMES
+        pretrained_vocab_files_map = PRETRAINED_VOCAB_FILES_MAP
+        max_model_input_sizes = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
+        model_input_names = ["input_ids", "attention_mask"]
+        slow_tokenizer_class = IQuestCoderTokenizer
+        
+        def __init__(
+            self,
+            vocab_file=None,
+            tokenizer_file=None,
+            unk_token="<unk>",
+            bos_token="<s>",
+            eos_token="</s>",
+            pad_token=None,
+            add_bos_token=True,
+            add_eos_token=False,
+            add_prefix_space=False,
+            use_default_system_prompt=False,
+            chat_template=None,
+            **kwargs,
+        ):
+            self.add_bos_token = add_bos_token
+            self.add_eos_token = add_eos_token
+            self.add_prefix_space = add_prefix_space
+            self.use_default_system_prompt = use_default_system_prompt
+            
+            if chat_template is None:
+                chat_template = DEFAULT_CHAT_TEMPLATE
+            
+            super().__init__(
+                vocab_file=vocab_file,
+                tokenizer_file=tokenizer_file,
+                unk_token=unk_token,
+                bos_token=bos_token,
+                eos_token=eos_token,
+                pad_token=pad_token,
+                add_bos_token=add_bos_token,
+                add_eos_token=add_eos_token,
+                add_prefix_space=add_prefix_space,
+                use_default_system_prompt=use_default_system_prompt,
+                chat_template=chat_template,
+                **kwargs,
+            )
+        
+        @property
+        def can_save_slow_tokenizer(self) -> bool:
+            return os.path.isfile(self.vocab_file) if self.vocab_file else False
+        
+        @property
+        def default_chat_template(self) -> str:
+            """Returns the default chat template."""
+            return DEFAULT_CHAT_TEMPLATE
+        
+        def build_inputs_with_special_tokens(
+            self, 
+            token_ids_0: List[int], 
+            token_ids_1: Optional[List[int]] = None
+        ) -> List[int]:
+            """Build model inputs with special tokens."""
+            bos_token_id = [self.bos_token_id] if self.add_bos_token else []
+            eos_token_id = [self.eos_token_id] if self.add_eos_token else []
+
+            output = bos_token_id + token_ids_0 + eos_token_id
+
+            if token_ids_1 is not None:
+                output = output + bos_token_id + token_ids_1 + eos_token_id
+
+            return output
+        
+        def get_special_tokens_mask(
+            self, 
+            token_ids_0: List[int], 
+            token_ids_1: Optional[List[int]] = None, 
+            already_has_special_tokens: bool = False
+        ) -> List[int]:
+            """Retrieve special tokens mask."""
+            if already_has_special_tokens:
+                return super().get_special_tokens_mask(
+                    token_ids_0=token_ids_0, token_ids_1=token_ids_1, already_has_special_tokens=True
+                )
+
+            bos_token_id = [1] if self.add_bos_token else []
+            eos_token_id = [1] if self.add_eos_token else []
+
+            if token_ids_1 is None:
+                return bos_token_id + ([0] * len(token_ids_0)) + eos_token_id
+            return (
+                bos_token_id
+                + ([0] * len(token_ids_0))
+                + eos_token_id
+                + bos_token_id
+                + ([0] * len(token_ids_1))
+                + eos_token_id
+            )
+        
+        def create_token_type_ids_from_sequences(
+            self, 
+            token_ids_0: List[int], 
+            token_ids_1: Optional[List[int]] = None
+        ) -> List[int]:
+            """Create token type IDs from sequences."""
+            bos_token_id = [self.bos_token_id] if self.add_bos_token else []
+            eos_token_id = [self.eos_token_id] if self.add_eos_token else []
+
+            output = [0] * len(bos_token_id + token_ids_0 + eos_token_id)
+
+            if token_ids_1 is not None:
+                output += [1] * len(bos_token_id + token_ids_1 + eos_token_id)
+
+            return output
+
+except ImportError:
+    # tokenizers library not available, Fast tokenizer not supported
+    IQuestCoderTokenizerFast = None
+    logger.info(
+        "The `tokenizers` library is not installed. "
+        "IQuestCoderTokenizerFast will not be available. "
+        "Install it with `pip install tokenizers`."
+    )
+

tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d3be68e090a927f31e0e378d7599b15c206dd47e4a73933775a746cc9c1cd91
+size 1345108

tokenizer_config.json

@@ -0,0 +1,285 @@
+{
+  "add_bos_token": false,
+  "add_eos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": true,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": true,
+      "special": true
+    },
+    "75858": {
+      "content": "<CLS>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "75859": {
+      "content": "<SEP>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "75860": {
+      "content": "<EOD>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "75861": {
+      "content": "<MASK>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "75862": {
+      "content": "<PAD>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "75863": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "75864": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "75865": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "75866": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "75867": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "75868": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "75869": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "75870": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "75871": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "75872": {
+      "content": "<think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "75873": {
+      "content": "</think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "75874": {
+      "content": "<tools>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "75875": {
+      "content": "</tools>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "75876": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "75877": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "75878": {
+      "content": "<tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "75879": {
+      "content": "</tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "75880": {
+      "content": "<|CLS|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "75881": {
+      "content": "<|SEP|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "75882": {
+      "content": "<|EOD|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "75883": {
+      "content": "<|MASK|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "75884": {
+      "content": "<|PAD|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [
+    "<|CLS|>",
+    "<|SEP|>",
+    "<|EOD|>",
+    "<|MASK|>",
+    "<|PAD|>",
+    "<|fim_prefix|>",
+    "<|fim_middle|>",
+    "<|fim_suffix|>",
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|fim_pad|>",
+    "<|endoftext|>",
+    "<|repo_name|>",
+    "<|file_sep|>",
+    "<think>",
+    "</think>"
+  ],
+  "auto_map": {
+    "AutoTokenizer": [
+      "tokenization_iquestcoder.IQuestCoderTokenizer",
+      null
+    ]
+  },
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "extra_special_tokens": {},
+  "legacy": true,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "sp_model_kwargs": {},
+  "split_special_tokens": false,
+  "tokenizer_class": "IQuestCoderTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false,
+  "use_fast": false
+}