Support Sentence Transformers via SparseEncoder

.gitattributes

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

1_SpladePooling/config.json

@@ -0,0 +1,4 @@
+{
+    "pooling_strategy": "max",
+    "activation_function": "relu"
+}

README.md

@@ -1,46 +1,53 @@
 ---
 license: cc-by-nc-sa-4.0
+tags:
+  - sentence-transformers
+  - splade
+  - sparse-encoder
+  - code
+pipeline_tag: feature-extraction
 ---
 
 SPLADE-Code-06B is a sparse retrieval model designed for code retrieval tasks. It is the top-performing models on MTEB for models below 1B (at time of writing, Feb 2026).
 
+## Usage
 
-```python
-from transformers import AutoModelForCausalLM, AutoModel
-import os
-import torch
-
-splade = AutoModelForCausalLM.from_pretrained("naver/splade-code-06B", trust_remote_code=True)
-device = (torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu"))
-splade.to(device)
-splade.eval()
-queries = ["SELECT *\nFROM Student\nWHERE Age = (\nSELECT MAX(Age)\nFROM Student\nWHERE Group = 'specific_group'\n)\nAND Group = 'specific_group';"]
-bow_dict = splade.encode(queries, prompt_type="query", top_k_q=10, return_dict=True, print_dict=True)
-```
+### Using Sentence Transformers
 
+Install Sentence Transformers:
+```bash
+pip install sentence_transformers
 ```
-+--------------------------------------------------------------------+
-|                        TOP ACTIVATED WORDS                         |
-+--------------------------------------------------------------------+
-
-
-* INPUT: SELECT *
-FROM Student
-WHERE Age = (
-SELECT MAX(Age)
-FROM Student
-WHERE Group = 'specific_group'
-)
-AND Group = 'specific_group';
-
-Ġgroup                    | ████████████████████ 2.34
-ĠAge                      | ████████████████████ 2.34
-Ġage                      | ███████████████████ 2.33
-Ġspecific                 | ███████████████████ 2.30
-_group                    | ███████████████████ 2.30
-ĠStudent                  | ███████████████████ 2.30
-Ġmax                      | ██████████████████ 2.22
-ĠMax                      | ██████████████████ 2.22
-Ġstudent                  | ██████████████████ 2.20
-ĠGroup                    | ██████████████████ 2.20
+
+```python
+from sentence_transformers import SparseEncoder
+
+model = SparseEncoder("naver/splade-code-06B", trust_remote_code=True)
+
+queries = [
+    "SELECT *\nFROM Student\nWHERE Age = (\nSELECT MAX(Age)\nFROM Student\nWHERE Group = 'specific_group'\n)\nAND Group = 'specific_group';"
+]
+
+query_embeddings = model.encode(queries)
+print(query_embeddings.shape)
+# torch.Size([1, 151936])
+
+sparsity = model.sparsity(query_embeddings)
+print(sparsity)
+# {'active_dims': 1231.0, 'sparsity_ratio': 0.991897904380792}
+
+decoded = model.decode(query_embeddings, top_k=10)
+print(decoded)
+# [[
+#     ("Ġgroup", 2.34375),
+#     ("Ġage", 2.34375),
+#     ("ĠAge", 2.34375),
+#     ("ĠStudent", 2.296875),
+#     ("Ġspecific", 2.296875),
+#     ("_group", 2.296875),
+#     ("ĠMax", 2.21875),
+#     ("Ġmax", 2.21875),
+#     ("Ġstudent", 2.203125),
+#     ("ĠGroup", 2.1875),
+# ]]
 ```

config.json

@@ -1,22 +1,34 @@
 {
-  "archi_type": "decoder",
   "architectures": [
-    "Splade"
+    "Qwen3ForCausalLM"
   ],
-  "attn_implementation": "flash_attention_2",
-  "attn_type": "causal",
-  "bidirectional": true,
-  "lora": false,
-  "lora_r": 0,
-  "model_name_or_path": "Qwen/Qwen3-0.6B",
-  "model_type": "splade",
-  "n_layers": null,
-  "padding_side": "left",
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "eos_token_id": 151645,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "is_causal": false,
+  "max_position_embeddings": 40960,
+  "max_window_layers": 28,
+  "model_type": "qwen3",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 8,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 1000000,
+  "sliding_window": null,
+  "tie_word_embeddings": true,
   "torch_dtype": "bfloat16",
-  "train_head": false,
-  "transformers_version": "4.53.3",
+  "transformers_version": "4.51.0",
+  "use_cache": true,
+  "use_sliding_window": false,
+  "vocab_size": 151936,
   "auto_map": {
-    "AutoConfig": "splade.SpladeConfig",
-    "AutoModelForCausalLM": "splade.Splade"
+    "AutoModelForMaskedLM": "modeling_splade.Qwen3ForCausalLM"
   }
-}
+}

config_sentence_transformers.json

@@ -0,0 +1,6 @@
+{
+  "model_type": "SparseEncoder",
+  "prompts": {},
+  "default_prompt_name": null,
+  "similarity_fn_name": "dot"
+}

modeling_qwen3_bidir.py

@@ -1,960 +0,0 @@
-###
-# Adapted from https://github.com/huggingface/transformers/blob/v4.51.2/src/transformers/models/qwen3/modeling_qwen3.py
-###
-
-from functools import partial
-from typing import Callable, Optional, Tuple, Union
-
-import torch
-from torch import nn
-
-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_outputs import (
-    BaseModelOutputWithPast,
-    CausalLMOutputWithPast,
-)
-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,
-    add_start_docstrings,
-    add_start_docstrings_to_model_forward,
-    can_return_tuple,
-    logging,
-    replace_return_docstrings,
-)
-from transformers.utils.deprecation import deprecate_kwarg
-from transformers.models.qwen3.configuration_qwen3 import Qwen3Config
-
-
-logger = logging.get_logger(__name__)
-
-_CHECKPOINT_FOR_DOC = "Qwen/Qwen3-8B"
-_CONFIG_FOR_DOC = "Qwen3Config"
-
-
-class Qwen3RMSNorm(nn.Module):
-    def __init__(self, hidden_size, eps=1e-6):
-        """
-        Qwen3RMSNorm is equivalent to T5LayerNorm
-        """
-        super().__init__()
-        self.weight = nn.Parameter(torch.ones(hidden_size))
-        self.variance_epsilon = eps
-
-    def forward(self, hidden_states):
-        input_dtype = hidden_states.dtype
-        hidden_states = hidden_states.to(torch.float32)
-        variance = hidden_states.pow(2).mean(-1, keepdim=True)
-        hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)
-        return self.weight * hidden_states.to(input_dtype)
-
-    def extra_repr(self):
-        return f"{tuple(self.weight.shape)}, eps={self.variance_epsilon}"
-
-
-class Qwen3MLP(nn.Module):
-    def __init__(self, config):
-        super().__init__()
-        self.config = config
-        self.hidden_size = config.hidden_size
-        self.intermediate_size = config.intermediate_size
-        self.gate_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False)
-        self.up_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False)
-        self.down_proj = nn.Linear(self.intermediate_size, self.hidden_size, bias=False)
-        self.act_fn = ACT2FN[config.hidden_act]
-
-    def forward(self, x):
-        down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))
-        return down_proj
-
-
-def rotate_half(x):
-    """Rotates half the hidden dims of the input."""
-    x1 = x[..., : x.shape[-1] // 2]
-    x2 = x[..., x.shape[-1] // 2 :]
-    return torch.cat((-x2, x1), dim=-1)
-
-
-def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_dim=1):
-    """Applies Rotary Position Embedding to the query and key tensors.
-    Args:
-        q (`torch.Tensor`): The query tensor.
-        k (`torch.Tensor`): The key tensor.
-        cos (`torch.Tensor`): The cosine part of the rotary embedding.
-        sin (`torch.Tensor`): The sine part of the rotary embedding.
-        position_ids (`torch.Tensor`, *optional*):
-            Deprecated and unused.
-        unsqueeze_dim (`int`, *optional*, defaults to 1):
-            The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and
-            sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note
-            that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and
-            k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes
-            cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have
-            the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.
-    Returns:
-        `tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding.
-    """
-    cos = cos.unsqueeze(unsqueeze_dim)
-    sin = sin.unsqueeze(unsqueeze_dim)
-    q_embed = (q * cos) + (rotate_half(q) * sin)
-    k_embed = (k * cos) + (rotate_half(k) * sin)
-    return q_embed, k_embed
-
-
-def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
-    """
-    This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch,
-    num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim)
-    """
-    batch, num_key_value_heads, slen, head_dim = hidden_states.shape
-    if n_rep == 1:
-        return hidden_states
-    hidden_states = hidden_states[:, :, None, :, :].expand(
-        batch, num_key_value_heads, n_rep, slen, head_dim
-    )
-    return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)
-
-
-def eager_attention_forward(
-    module: nn.Module,
-    query: torch.Tensor,
-    key: torch.Tensor,
-    value: torch.Tensor,
-    attention_mask: Optional[torch.Tensor],
-    scaling: float,
-    dropout: float = 0.0,
-    **kwargs,
-):
-    key_states = repeat_kv(key, module.num_key_value_groups)
-    value_states = repeat_kv(value, module.num_key_value_groups)
-
-    attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling
-    if attention_mask is not None:
-        causal_mask = attention_mask[:, :, :, : key_states.shape[-2]]
-        attn_weights = attn_weights + causal_mask
-
-    attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(
-        query.dtype
-    )
-    attn_weights = nn.functional.dropout(
-        attn_weights, p=dropout, training=module.training
-    )
-    attn_output = torch.matmul(attn_weights, value_states)
-    attn_output = attn_output.transpose(1, 2).contiguous()
-
-    return attn_output, attn_weights
-
-
-class Qwen3BidirAttention(nn.Module):
-    """Multi-headed attention from 'Attention Is All You Need' paper"""
-
-    def __init__(self, config: Qwen3Config, 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 = False
-
-        self.q_proj = nn.Linear(
-            config.hidden_size,
-            config.num_attention_heads * self.head_dim,
-            bias=config.attention_bias,
-        )
-        self.k_proj = nn.Linear(
-            config.hidden_size,
-            config.num_key_value_heads * self.head_dim,
-            bias=config.attention_bias,
-        )
-        self.v_proj = nn.Linear(
-            config.hidden_size,
-            config.num_key_value_heads * self.head_dim,
-            bias=config.attention_bias,
-        )
-        self.o_proj = nn.Linear(
-            config.num_attention_heads * self.head_dim,
-            config.hidden_size,
-            bias=config.attention_bias,
-        )
-        self.q_norm = Qwen3RMSNorm(
-            self.head_dim, eps=config.rms_norm_eps
-        )  # unlike olmo, only on the head dim!
-        self.k_norm = Qwen3RMSNorm(
-            self.head_dim, eps=config.rms_norm_eps
-        )  # thus post q_norm does not need reshape
-        self.sliding_window = config.sliding_window
-        if not (
-            self.config.use_sliding_window
-            and getattr(self.config, "sliding_window", None) is not None
-            and self.layer_idx >= self.config.max_window_layers
-        ):
-            self.sliding_window = None
-
-    def forward(
-        self,
-        hidden_states: torch.Tensor,
-        position_embeddings: Tuple[torch.Tensor, torch.Tensor],
-        attention_mask: Optional[torch.Tensor],
-        past_key_value: Optional[Cache] = None,
-        cache_position: Optional[torch.LongTensor] = None,
-        **kwargs: Unpack[FlashAttentionKwargs],
-    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
-        input_shape = hidden_states.shape[:-1]
-        hidden_shape = (*input_shape, -1, self.head_dim)
-
-        query_states = self.q_norm(
-            self.q_proj(hidden_states).view(hidden_shape)
-        ).transpose(1, 2)
-        key_states = self.k_norm(
-            self.k_proj(hidden_states).view(hidden_shape)
-        ).transpose(1, 2)
-        value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)
-
-        cos, sin = position_embeddings
-        query_states, key_states = apply_rotary_pos_emb(
-            query_states, key_states, cos, sin
-        )
-
-        if past_key_value is not None:
-            # sin and cos are specific to RoPE models; cache_position needed for the static cache
-            cache_kwargs = {"sin": sin, "cos": cos, "cache_position": cache_position}
-            key_states, value_states = past_key_value.update(
-                key_states, value_states, self.layer_idx, cache_kwargs
-            )
-
-        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
-                ]
-
-        attn_output, attn_weights = attention_interface(
-            self,
-            query_states,
-            key_states,
-            value_states,
-            attention_mask,
-            dropout=0.0 if not self.training else self.attention_dropout,
-            scaling=self.scaling,
-            sliding_window=self.sliding_window,  # diff with Llama
-            **kwargs,
-        )
-
-        attn_output = attn_output.reshape(*input_shape, -1).contiguous()
-        attn_output = self.o_proj(attn_output)
-        return attn_output, attn_weights
-
-
-class Qwen3BidirDecoderLayer(nn.Module):
-    def __init__(self, config: Qwen3Config, layer_idx: int):
-        super().__init__()
-        self.hidden_size = config.hidden_size
-        self.self_attn = Qwen3BidirAttention(config=config, layer_idx=layer_idx)
-        self.mlp = Qwen3MLP(config)
-        self.input_layernorm = Qwen3RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
-        self.post_attention_layernorm = Qwen3RMSNorm(
-            config.hidden_size, eps=config.rms_norm_eps
-        )
-        if (
-            config.sliding_window and config._attn_implementation != "flash_attention_2"
-        ):  # diff with Llama is this warning
-            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,  # necessary, but kept here for BC
-        **kwargs: Unpack[FlashAttentionKwargs],
-    ) -> Tuple[
-        torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]
-    ]:
-        residual = hidden_states
-
-        hidden_states = self.input_layernorm(hidden_states)
-
-        # Self Attention
-        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
-
-        # Fully Connected
-        residual = hidden_states
-        hidden_states = self.post_attention_layernorm(hidden_states)
-        hidden_states = self.mlp(hidden_states)
-        hidden_states = residual + hidden_states
-
-        outputs = (hidden_states,)
-        if output_attentions:
-            outputs += (self_attn_weights,)
-
-        return outputs
-
-
-class Qwen3RotaryEmbedding(nn.Module):
-    def __init__(self, config: Qwen3Config, device=None):
-        super().__init__()
-        # BC: "rope_type" was originally "type"
-        if hasattr(config, "rope_scaling") and config.rope_scaling is not None:
-            self.rope_type = config.rope_scaling.get(
-                "rope_type", config.rope_scaling.get("type")
-            )
-        else:
-            self.rope_type = "default"
-        self.max_seq_len_cached = config.max_position_embeddings
-        self.original_max_seq_len = config.max_position_embeddings
-
-        self.config = config
-        self.rope_init_fn = ROPE_INIT_FUNCTIONS[self.rope_type]
-
-        inv_freq, self.attention_scaling = self.rope_init_fn(self.config, device)
-        self.register_buffer("inv_freq", inv_freq, persistent=False)
-        self.original_inv_freq = self.inv_freq
-
-    @torch.no_grad()
-    @dynamic_rope_update  # power user: used with advanced RoPE types (e.g. dynamic rope)
-    def forward(self, x, position_ids):
-        inv_freq_expanded = (
-            self.inv_freq[None, :, None]
-            .float()
-            .expand(position_ids.shape[0], -1, 1)
-            .to(x.device)
-        )
-        position_ids_expanded = position_ids[:, None, :].float()
-
-        device_type = (
-            x.device.type
-            if isinstance(x.device.type, str) and x.device.type != "mps"
-            else "cpu"
-        )
-        with torch.autocast(device_type=device_type, enabled=False):  # Force float32
-            freqs = (
-                inv_freq_expanded.float() @ position_ids_expanded.float()
-            ).transpose(1, 2)
-            emb = torch.cat((freqs, freqs), dim=-1)
-            cos = emb.cos() * self.attention_scaling
-            sin = emb.sin() * self.attention_scaling
-
-        return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)
-
-
-QWEN3_START_DOCSTRING = r"""
-    This model inherits from [`PreTrainedModel`]. Check the superclass documentation for the generic methods the
-    library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
-    etc.)
-    This model is also a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) subclass.
-    Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage
-    and behavior.
-    Parameters:
-        config ([`Qwen3Config`]):
-            Model configuration class with all the parameters of the model. Initializing with a config file does not
-            load the weights associated with the model, only the configuration. Check out the
-            [`~PreTrainedModel.from_pretrained`] method to load the model weights.
-"""
-
-
-@add_start_docstrings(
-    "The bare Qwen3 Model outputting raw hidden-states without any specific head on top.",
-    QWEN3_START_DOCSTRING,
-)
-class Qwen3PreTrainedModel(PreTrainedModel):
-    config_class = Qwen3Config
-    base_model_prefix = "model"
-    supports_gradient_checkpointing = True
-    _no_split_modules = ["Qwen3DecoderLayer"]
-    _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):
-        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_()
-
-
-QWEN3_INPUTS_DOCSTRING = r"""
-    Args:
-        input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
-            Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide
-            it.
-            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
-            [`PreTrainedTokenizer.__call__`] for details.
-            [What are input IDs?](../glossary#input-ids)
-        attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
-            Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
-            - 1 for tokens that are **not masked**,
-            - 0 for tokens that are **masked**.
-            [What are attention masks?](../glossary#attention-mask)
-            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
-            [`PreTrainedTokenizer.__call__`] for details.
-            If `past_key_values` is used, optionally only the last `input_ids` have to be input (see
-            `past_key_values`).
-            If you want to change padding behavior, you should read [`modeling_opt._prepare_decoder_attention_mask`]
-            and modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more
-            information on the default strategy.
-            - 1 indicates the head is **not masked**,
-            - 0 indicates the head is **masked**.
-        position_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
-            Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0,
-            config.n_positions - 1]`.
-            [What are position IDs?](../glossary#position-ids)
-        past_key_values (`Cache`, *optional*):
-            Pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention
-            blocks) that can be used to speed up sequential decoding. This typically consists in the `past_key_values`
-            returned by the model at a previous stage of decoding, when `use_cache=True` or `config.use_cache=True`.
-            It is a [`~cache_utils.Cache`] instance. For more details, see our [kv cache guide](https://huggingface.co/docs/transformers/en/kv_cache).
-            If `past_key_values` are used, the user can optionally input only the last `input_ids` (those that don't
-            have their past key value states given to this model) of shape `(batch_size, 1)` instead of all `input_ids`
-            of shape `(batch_size, sequence_length)`.
-        inputs_embeds (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
-            Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation. This
-            is useful if you want more control over how to convert `input_ids` indices into associated vectors than the
-            model's internal embedding lookup matrix.
-        use_cache (`bool`, *optional*):
-            If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see
-            `past_key_values`).
-        output_attentions (`bool`, *optional*):
-            Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned
-            tensors for more detail.
-        output_hidden_states (`bool`, *optional*):
-            Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
-            more detail.
-        return_dict (`bool`, *optional*):
-            Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
-        cache_position (`torch.LongTensor` of shape `(sequence_length)`, *optional*):
-            Indices depicting the position of the input sequence tokens in the sequence. Contrarily to `position_ids`,
-            this tensor is not affected by padding. It is used to update the cache in the correct position and to infer
-            the complete sequence length.
-"""
-
-
-@add_start_docstrings(
-    "The bare Qwen3 Model outputting raw hidden-states without any specific head on top.",
-    QWEN3_START_DOCSTRING,
-)
-class Qwen3BidirModel(Qwen3PreTrainedModel):
-    """
-    Transformer decoder consisting of *config.num_hidden_layers* layers. Each layer is a [`Qwen3DecoderLayer`]
-    Args:
-        config: Qwen3Config
-    """
-
-    def __init__(self, config: Qwen3Config):
-        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(
-            [
-                Qwen3BidirDecoderLayer(config, layer_idx)
-                for layer_idx in range(config.num_hidden_layers)
-            ]
-        )
-        self.norm = Qwen3RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
-        self.rotary_emb = Qwen3RotaryEmbedding(config=config)
-        self.gradient_checkpointing = False
-
-        # Initialize weights and apply final processing
-        self.post_init()
-
-    def get_input_embeddings(self):
-        return self.embed_tokens
-
-    def set_input_embeddings(self, value):
-        self.embed_tokens = value
-
-    @can_return_tuple
-    @add_start_docstrings_to_model_forward(QWEN3_INPUTS_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
-
-        # TODO (joao): remove this exception in v4.56 -- it exists for users that try to pass a legacy cache
-        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,)
-
-            if self.gradient_checkpointing and self.training:
-                layer_outputs = self._gradient_checkpointing_func(
-                    partial(decoder_layer.__call__, **flash_attn_kwargs),
-                    hidden_states,
-                    causal_mask,
-                    position_ids,
-                    past_key_values,
-                    output_attentions,
-                    use_cache,
-                    cache_position,
-                    position_embeddings,
-                )
-            else:
-                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: torch.Tensor,
-        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:
-                valid_rows = attention_mask.sum(dim=1) > 0
-
-                if valid_rows.any():
-                    # Only check right-padding on non-empty rows
-                    right_padded_rows = attention_mask[valid_rows, -1] == 0
-                    is_padding_right = right_padded_rows.any().item()
-                    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 Qwen3. "
-                            "Make sure to call `tokenizer.padding_side = 'left'` before tokenizing the input."
-                        )
-                # 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 Qwen3. 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
-
-        # For SDPA, when possible, we will rely on its `is_causal` argument instead of its `attn_mask` argument, in
-        # order to dispatch on Flash Attention 2. This feature is not compatible with static cache, as SDPA will fail
-        # to infer the 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)
-
-        # When output attentions is True, sdpa implementation's forward method calls the eager implementation's forward
-        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,
-                is_training=self.training,
-            ):
-                return None
-
-        dtype, device = input_tensor.dtype, input_tensor.device
-        min_dtype = torch.finfo(dtype).min
-        sequence_length = input_tensor.shape[1]
-        # SlidingWindowCache or StaticCache
-        if using_sliding_window_cache or using_static_cache:
-            target_length = past_key_values.get_max_cache_shape()
-        # DynamicCache or no cache
-        else:
-            target_length = (
-                attention_mask.shape[-1]
-                if isinstance(attention_mask, torch.Tensor)
-                else past_seen_tokens + sequence_length + 1
-            )
-
-        # In case the provided `attention` mask is 2D, we generate a causal mask here (4D).
-        causal_mask = self._prepare_4d_causal_attention_mask_with_cache_position(
-            attention_mask,
-            sequence_length=sequence_length,
-            target_length=target_length,
-            dtype=dtype,
-            device=device,
-            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"]
-            and not output_attentions
-        ):
-            # Attend to all tokens in fully masked rows in the causal_mask, for example the relevant first rows when
-            # using left padding. This is required by F.scaled_dot_product_attention memory-efficient attention path.
-            # Details: https://github.com/pytorch/pytorch/issues/110213
-            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,
-        device: torch.device,
-        cache_position: torch.Tensor,
-        batch_size: int,
-        config: Qwen3Config,
-        past_key_values: Cache,
-    ):
-        """
-        Creates a causal 4D mask of shape `(batch_size, 1, query_length, key_value_length)` from a 2D mask of shape
-        `(batch_size, key_value_length)`, or if the input `attention_mask` is already 4D, do nothing.
-        Args:
-            attention_mask (`torch.Tensor`):
-                A 2D attention mask of shape `(batch_size, key_value_length)` or a 4D attention mask of shape `(batch_size, 1, query_length, key_value_length)`.
-            sequence_length (`int`):
-                The sequence length being processed.
-            target_length (`int`):
-                The target length: when generating with static cache, the mask should be as long as the static cache, to account for the 0 padding, the part of the cache that is not filled yet.
-            dtype (`torch.dtype`):
-                The dtype to use for the 4D attention mask.
-            device (`torch.device`):
-                The device to place the 4D attention mask on.
-            cache_position (`torch.Tensor`):
-                Indices depicting the position of the input sequence tokens in the sequence.
-            batch_size (`torch.Tensor`):
-                Batch size.
-            config (`Qwen3Config`):
-                The model's configuration class
-            past_key_values (`Cache`):
-                The cache class that is being used currently to generate
-        """
-        if attention_mask is not None and attention_mask.dim() == 4:
-            # In this case we assume that the mask comes already in inverted form and requires no inversion or slicing.
-            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=device,
-            )
-            diagonal_attend_mask = torch.arange(
-                target_length, device=device
-            ) > cache_position.reshape(-1, 1)
-            if config.sliding_window is not None:
-                # if we have sliding window, we should not attend to tokens beyond sliding window length, so we mask them out also
-                # the check is needed to verify is current checkpoint was trained with sliding window or not
-                if (
-                    not isinstance(past_key_values, SlidingWindowCache)
-                    or sequence_length > target_length
-                ):
-                    sliding_attend_mask = torch.arange(
-                        target_length, device=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()
-                )  # copy to contiguous memory for in-place edit
-                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
-
-
-class KwargsForCausalLM(FlashAttentionKwargs, LossKwargs): ...
-
-
-class Qwen3BidirForCausalLM(Qwen3PreTrainedModel, GenerationMixin):
-    _tied_weights_keys = ["lm_head.weight"]
-    _tp_plan = {"lm_head": "colwise_rep"}
-    _pp_plan = {"lm_head": (["hidden_states"], ["logits"])}
-
-    def __init__(self, config):
-        super().__init__(config)
-        self.model = Qwen3BidirModel(config)
-        self.vocab_size = config.vocab_size
-        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
-
-        # Initialize weights and apply final processing
-        self.post_init()
-
-    def get_input_embeddings(self):
-        return self.model.embed_tokens
-
-    def set_input_embeddings(self, value):
-        self.model.embed_tokens = value
-
-    def get_output_embeddings(self):
-        return self.lm_head
-
-    def set_output_embeddings(self, new_embeddings):
-        self.lm_head = new_embeddings
-
-    def set_decoder(self, decoder):
-        self.model = decoder
-
-    def get_decoder(self):
-        return self.model
-
-    @can_return_tuple
-    @deprecate_kwarg("num_logits_to_keep", version="4.50", new_name="logits_to_keep")
-    @add_start_docstrings_to_model_forward(QWEN3_INPUTS_DOCSTRING)
-    @replace_return_docstrings(
-        output_type=CausalLMOutputWithPast, config_class=_CONFIG_FOR_DOC
-    )
-    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"""
-            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]`.
-            logits_to_keep (`int` or `torch.Tensor`, *optional*):
-                If an `int`, compute logits for the last `logits_to_keep` tokens. If `0`, calculate logits for all
-                `input_ids` (special case). Only last token logits are needed for generation, and calculating them only for that
-                token can save memory, which becomes pretty significant for long sequences or large vocabulary size.
-                If a `torch.Tensor`, must be 1D corresponding to the indices to keep in the sequence length dimension.
-                This is useful when using packed tensor format (single dimension for batch and sequence length).
-        Returns:
-        Example:
-        ```python
-        >>> from transformers import AutoTokenizer, Qwen3ForCausalLM
-        >>> model = Qwen3ForCausalLM.from_pretrained("Qwen/Qwen3-8B")
-        >>> tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3-8B")
-        >>> 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,
-        )

modeling_splade.py

@@ -0,0 +1,28 @@
+"""
+This file exists solely to allow loading the Qwen3ForCausalLM via the AutoModelForMaskedLM class.
+Compared to standard Qwen3, we're using bidirectional attention and not causal attention, but it's specified
+with `is_causal=False` in the config.
+"""
+
+from transformers import Qwen3ForCausalLM as _Qwen3ForCausalLM
+
+
+class Qwen3ForCausalLM(_Qwen3ForCausalLM):
+    def tie_weights(self, *args, **kwargs):
+        """Explicitly re-tie lm_head to embed_tokens to hopefully avoid meta tensor errors."""
+        super().tie_weights(*args, **kwargs)
+        if (
+            self.config.tie_word_embeddings
+            and hasattr(self, "lm_head")
+            and hasattr(self, "model")
+        ):
+            self.lm_head.weight = self.model.embed_tokens.weight
+
+    def _init_weights(self, module):
+        """Skip lm_head init when it will be tied to embed_tokens later."""
+        if module is getattr(self, "lm_head", None) and self.config.tie_word_embeddings:
+            return
+        super()._init_weights(module)
+
+
+__all__ = ["Qwen3ForCausalLM"]

modules.json

@@ -0,0 +1,14 @@
+[
+  {
+    "idx": 0,
+    "name": "0",
+    "path": "",
+    "type": "sentence_transformers.sparse_encoder.models.MLMTransformer"
+  },
+  {
+    "idx": 1,
+    "name": "1",
+    "path": "1_SpladePooling",
+    "type": "sentence_transformers.sparse_encoder.models.SpladePooling"
+  }
+]

splade.py

@@ -1,109 +0,0 @@
-import os
-from transformers import (
-    PretrainedConfig,
-    PreTrainedModel,
-    AutoConfig,
-)
-from huggingface_hub import snapshot_download
-from typing import Optional
-from transformers.utils import is_flash_attn_2_available
-from .utils import (
-    get_decoder_model,
-    prepare_tokenizer,
-    splade_max,
-    similarity,
-    encode,
-)
-from peft import PeftModel
-
-
-class SpladeConfig(PretrainedConfig):
-    model_type = "splade"
-
-    def __init__(
-        self,
-        model_name_or_path: str = "Qwen/Qwen3-0.6B",
-        attn_implementation: str = "flash_attention_2",
-        bidirectional: bool = True,  # only for decoder models
-        padding_side: str = "left",
-        **kwargs,
-    ):
-        super().__init__(**kwargs)
-        self.model_name_or_path = model_name_or_path
-        self.attn_implementation = attn_implementation
-        self.bidirectional = bidirectional
-        self.padding_side = padding_side
-
-
-class Splade(PreTrainedModel):
-    config_class = SpladeConfig
-
-    # methods for MTEB's interface
-    similarity = similarity
-    encode = encode
-
-    def __init__(self, config, weights_path=None, token=None):
-        super().__init__(config)
-        self.name = "splade"
-    
-        base_cfg = AutoConfig.from_pretrained(
-            config.model_name_or_path,
-            attn_implementation=config.attn_implementation,
-            torch_dtype="auto",
-        )
-    
-        self.tokenizer = prepare_tokenizer(
-            config.model_name_or_path, padding_side=config.padding_side
-        )
-    
-        if is_flash_attn_2_available():
-            config.attn_implementation = "flash_attention_2"
-        else:
-            config.attn_implementation = "sdpa"
-
-        source = weights_path or config.model_name_or_path
-    
-        self.model = get_decoder_model(
-            model_name_or_path=source,
-            attn_implementation=config.attn_implementation,
-            bidirectional=getattr(config, "bidirectional", False),
-            base_cfg=base_cfg,
-            token=token
-        )
-
-    def save_pretrained(self, save_directory, *args, **kwargs):
-        self.model.save_pretrained(os.path.join(save_directory, "lora"))
-        self.config.save_pretrained(save_directory)
-
-    @classmethod
-    def from_pretrained(cls, model_name_or_path, *args, **kwargs):
-        token = kwargs.get("token", None)
-    
-        config = SpladeConfig.from_pretrained(
-            model_name_or_path,
-            token=token,
-        )
-    
-        model = cls(config, weights_path=model_name_or_path, token=token)
-    
-        model.reverse_voc = {v: k for k, v in model.tokenizer.vocab.items()}
-        return model
-
-    def forward(self, **tokens):
-        output = self.model(**tokens)
-        splade_reps, _ = splade_max(output.logits, tokens["attention_mask"])
-        return (splade_reps,)
-
-    def get_width(self):
-        return self.model.config.vocab_size
-
-    def create_batch_dict(self, input_texts, max_length):
-        return self.tokenizer(
-            input_texts,
-            add_special_tokens=True,
-            padding="longest",
-            truncation=True,
-            max_length=max_length,
-            return_attention_mask=True,
-            return_tensors="pt",
-        )

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:be75606093db2094d7cd20f3c2f385c212750648bd6ea4fb2bf507a6a4c55506
+size 11422650

tokenizer_config.json

@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "model_max_length": 131072,
+  "padding_side": "left",
+  "pad_token": "<|endoftext|>",
+  "chat_template": null,
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

utils.py

@@ -1,152 +0,0 @@
-import numpy as np
-import torch
-
-from typing import Any
-from transformers import AutoTokenizer
-
-
-def splade_max(features, attention_mask):
-    """
-    SPLADE pooling operation
-    """
-    relu = torch.nn.ReLU(inplace=False)
-    values, ids_ = torch.max(
-        torch.log(1 + relu(features)) * attention_mask.unsqueeze(-1), dim=1
-    )
-    return values, ids_
-
-
-def encode(
-    self,
-    sentences: list[str],
-    max_length: int = 1024,
-    prompt_type: str = "document",
-    return_dict: bool = False,
-    print_dict: bool = False,
-    batch_size: int = 8,
-    top_k_q: int = -1,
-    top_k_d: int = -1,
-    **kwargs: Any,
-) -> np.ndarray:
-    all_embeddings = []
-    for i in range(0, len(sentences), batch_size):
-        batch_texts = sentences[i : i + batch_size]
-        batch_dict = self.create_batch_dict(batch_texts, max_length)
-        batch_dict = {
-            key: value.to(self.model.device) for key, value in batch_dict.items()
-        }
-        with torch.no_grad():
-            splare_reps = self(**batch_dict)[0]
-            if prompt_type == "query" and top_k_q > 0:
-                splare_reps = top_k(splare_reps, top_k_q)
-            if prompt_type == "document" and top_k_d > 0:
-                splare_reps = top_k(splare_reps, top_k_d)
-            all_embeddings.append(splare_reps.cpu().float().numpy())
-    if return_dict:
-        d = bow_dict(self, np.concatenate(all_embeddings, axis=0))
-        if print_dict:
-            print_bow_bars(sentences, d)
-        return d
-    else:
-        return np.concatenate(all_embeddings, axis=0)
-
-
-def bow_dict(self, embeddings):
-    out = []
-    for vector in embeddings:
-        idx = np.nonzero(vector)[0]
-        weights = vector[idx]
-        d = {k: v for k, v in zip(idx.tolist(), weights.tolist())}
-        sorted_d = {
-            self.reverse_voc[k]: float(v)
-            for k, v in sorted(d.items(), key=lambda item: item[1], reverse=True)
-        }
-        out.append(sorted_d)
-    return out
-
-
-def print_bow_bars(sentences, bow_list, width=20):
-    ascii_header("TOP ACTIVATED WORDS")
-    for sent, bow in zip(sentences, bow_list):
-        print(f"* INPUT: {sent}\n")
-        max_w = max(bow.values())
-        for k, v in sorted(bow.items(), key=lambda x: x[1], reverse=True):
-            bar = "█" * int(v / max_w * width)
-            print(f"{k[:25]:25} | {bar} {v:.2f}")
-        print("\n")
-
-
-def ascii_header(title, width=70):
-    title = f" {title} "
-    print("+" + "-" * (width - 2) + "+")
-    print("|" + title.center(width - 2) + "|")
-    print("+" + "-" * (width - 2) + "+")
-    print("\n")
-
-
-def similarity(self, a, b) -> torch.Tensor:
-    """
-    MTEB eval requires this
-    """
-    if not isinstance(a, torch.Tensor):
-        a = torch.tensor(a)
-    if not isinstance(b, torch.Tensor):
-        b = torch.tensor(b)
-
-    def _dot_score_core(a_tensor, b_tensor):
-        if len(a_tensor.shape) == 1:
-            a_tensor = a_tensor.unsqueeze(0)
-        if len(b_tensor.shape) == 1:
-            b_tensor = b_tensor.unsqueeze(0)
-        return a_tensor @ b_tensor.transpose(0, 1)
-
-    return _dot_score_core(a, b)
-
-
-def prepare_tokenizer(tokenizer_name: str, padding_side="right"):
-    """
-    loads and prepares tokenizer
-    """
-    tokenizer = AutoTokenizer.from_pretrained(tokenizer_name)
-    tokenizer.pad_token = (
-        tokenizer.bos_token or tokenizer.pad_token or tokenizer.eos_token
-    )
-    tokenizer.padding_side = padding_side
-    return tokenizer
-
-
-def get_decoder_model(
-    model_name_or_path: str, attn_implementation: str, bidirectional: bool, base_cfg, token=None
-):
-    """
-    base_cfg is the pretrained config of the underlying model
-    """
-    print("WARNING: bidirectional only tested for transformer 4.51.2")
-    assert (
-        bidirectional is True
-    ), "the model has been trained with bi-directional attention!"
-    assert (
-        attn_implementation == "flash_attention_2"
-    ), f"bidir models only support flash_attention_2 for now, not {attn_implementation}!"
-    from .modeling_qwen3_bidir import Qwen3BidirForCausalLM
-
-    return Qwen3BidirForCausalLM.from_pretrained(
-        model_name_or_path,
-        config=base_cfg,
-        torch_dtype=torch.bfloat16,
-        attn_implementation=attn_implementation,
-        token=token,
-    )
-
-
-def top_k(x: torch.Tensor, k: int) -> torch.Tensor:
-    """
-    zeroes out all but the top-k values in the last dimension of x
-    """
-    _, topk_indices = x.topk(k, dim=-1)
-    # create a zero tensor of the same shape as x
-    mask = torch.zeros_like(x, dtype=torch.bool)
-    # use scatter along the last dimension
-    mask.scatter_(-1, topk_indices, True)
-    # zero out all but the top-k
-    return x * mask