Upload folder using huggingface_hub

.gitattributes

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

added_tokens.json

@@ -0,0 +1,24 @@
+{
+  "</tool_call>": 151658,
+  "<tool_call>": 151657,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}

config.json

@@ -0,0 +1,38 @@
+{
+  "architectures": [
+    "C2LLMForEmbedding"
+  ],
+  "attention_dropout": 0.0,
+  "attn_implementation": "flash_attention_2",
+  "auto_map": {
+    "AutoConfig": "configuration_c2llm.C2LLMConfig",
+    "AutoModel": "modeling_c2llm.C2LLMForEmbedding"
+  },
+  "compressed_dim": 3584,
+  "embedding_method": "pma",
+  "hidden_act": "silu",
+  "hidden_size": 3584,
+  "initializer_range": 0.02,
+  "intermediate_size": 18944,
+  "max_position_embeddings": 32768,
+  "max_window_layers": 28,
+  "model_type": "c2llm",
+  "num_attention_heads": 28,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 4,
+  "padding_side": "left",
+  "pma_ln": true,
+  "pma_norm": false,
+  "pma_norm_mode": "post_normal",
+  "pma_num_heads": 32,
+  "rms_norm_eps": 1e-06,
+  "rope_theta": 1000000.0,
+  "sliding_window": null,
+  "tie_word_embeddings": false,
+  "tokenizer_name_or_path": "codefuse-ai/C2LLM-7B",
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.51.3",
+  "use_cache": true,
+  "use_sliding_window": false,
+  "vocab_size": 152064
+}

config_sentence_transformers.json

@@ -0,0 +1,8 @@
+{
+  "prompts": {
+    "query": "Given a query from a web search, retrieve code that is helpful in addressing the query.\n",
+    "document": "Retrieved Answer:"
+  },
+  "default_prompt_name": null,
+  "similarity_fn_name": "cosine"
+}

configuration_c2llm.py

@@ -0,0 +1,87 @@
+from transformers import PretrainedConfig
+
+class C2LLMConfig(PretrainedConfig):
+    model_type = "c2llm"
+    keys_to_ignore_at_inference = ["past_key_values"]
+
+    base_model_tp_plan = {
+        "layers.*.self_attn.q_proj": "colwise",
+        "layers.*.self_attn.k_proj": "colwise",
+        "layers.*.self_attn.v_proj": "colwise",
+        "layers.*.self_attn.o_proj": "rowwise",
+        "layers.*.mlp.gate_proj": "colwise",
+        "layers.*.mlp.up_proj": "colwise",
+        "layers.*.mlp.down_proj": "rowwise",
+    }
+    base_model_pp_plan = {
+        "embed_tokens": (["input_ids"], ["inputs_embeds"]),
+        "layers": (["hidden_states", "attention_mask"], ["hidden_states"]),
+        "norm": (["hidden_states"], ["hidden_states"]),
+    }
+
+
+    def __init__(
+        self,
+        attention_dropout=0.0,
+        bos_token_id=151643,
+        eos_token_id=151645,
+        hidden_act="silu",
+        hidden_size=3584,
+        initializer_range=0.02,
+        intermediate_size=18944,
+        max_position_embeddings=32768,
+        max_window_layers=28,
+        model_type="c2llm",
+        num_attention_heads=28,
+        num_hidden_layers=28,
+        num_key_value_heads=4,
+        rms_norm_eps=1e-6,
+        rope_theta=1000000.0,
+        sliding_window=131072,
+        tie_word_embeddings=False,
+        torch_dtype="bfloat16",
+        transformers_version="4.43.1",
+        use_cache=True,
+        use_sliding_window=False,
+        vocab_size=152064,
+        **kwargs,
+    ):
+        self.vocab_size = vocab_size
+        self.max_position_embeddings = max_position_embeddings
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.use_sliding_window = use_sliding_window
+        self.sliding_window = sliding_window if use_sliding_window else None
+        self.max_window_layers = max_window_layers
+
+        # for backward compatibility
+        if num_key_value_heads is None:
+            num_key_value_heads = num_attention_heads
+
+        self.num_key_value_heads = num_key_value_heads
+        self.hidden_act = hidden_act
+        self.initializer_range = initializer_range
+        self.rms_norm_eps = rms_norm_eps
+        self.use_cache = use_cache
+        self.rope_theta = rope_theta
+        self.attention_dropout = attention_dropout
+
+        super().__init__(
+            tie_word_embeddings=tie_word_embeddings,
+            **kwargs,
+        )
+    def to_dict(self):
+        output = super().to_dict()
+        
+        keys_to_remove = [
+            "base_model"
+        ]
+        
+        for key in keys_to_remove:
+            output.pop(key, None)
+            
+        return output
+        
+__all__ = ["C2LLMConfig"]

custom_transformer.py

@@ -0,0 +1,45 @@
+from __future__ import annotations
+
+import inspect
+import logging
+import os
+from collections.abc import Callable
+from pathlib import Path
+from typing import TYPE_CHECKING, Any
+
+from sentence_transformers.backend import load_onnx_model, load_openvino_model
+
+try:
+    from typing import Self
+except ImportError:
+    from typing_extensions import Self
+
+import torch
+from transformers import AutoConfig, AutoModel, AutoTokenizer, MT5Config, PretrainedConfig, T5Config
+from transformers.utils.import_utils import is_peft_available
+from transformers.utils.peft_utils import find_adapter_config_file
+
+from sentence_transformers.models.InputModule import InputModule
+
+logger = logging.getLogger(__name__)
+
+if TYPE_CHECKING and is_peft_available():
+    from peft import PeftConfig
+from sentence_transformers.models import Transformer
+
+
+class C2LLMTransformer(Transformer):
+    config_file_name: str = "sentence_bert_config.json"
+    config_keys: list[str] = ["max_seq_length", "do_lower_case"]
+    save_in_root: bool = True
+
+    
+    def forward(self, features: dict[str, torch.Tensor], **kwargs) -> dict[str, torch.Tensor]:
+        trans_features = {key: value for key, value in features.items() if key in self.model_forward_params}
+
+        outputs = self.auto_model(**trans_features, **kwargs, return_dict=True)
+      
+        sentence_embedding = outputs["sentence_embedding"]
+        features["sentence_embedding"] = sentence_embedding
+
+        return features

model-00003-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2766a6d5a6f3a0c00676b1705aa971928704f4b408456dbbaef0642f5d20bc9b
+size 4330866320

model-00004-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d7113672e683e2af68b90adaeb06e102f242a773af8c96e70a8d2da5d3ea8c44
+size 1192821112

model.safetensors.index.json

@@ -0,0 +1,359 @@
+{
+  "metadata": {
+    "total_size": 15334057984
+  },
+  "weight_map": {
+    "mha_pma_disc.S": "model-00004-of-00004.safetensors",
+    "mha_pma_disc.mab.fc_k.bias": "model-00004-of-00004.safetensors",
+    "mha_pma_disc.mab.fc_k.weight": "model-00004-of-00004.safetensors",
+    "mha_pma_disc.mab.fc_o.bias": "model-00004-of-00004.safetensors",
+    "mha_pma_disc.mab.fc_o.weight": "model-00004-of-00004.safetensors",
+    "mha_pma_disc.mab.fc_q.bias": "model-00004-of-00004.safetensors",
+    "mha_pma_disc.mab.fc_q.weight": "model-00004-of-00004.safetensors",
+    "mha_pma_disc.mab.fc_v.bias": "model-00004-of-00004.safetensors",
+    "mha_pma_disc.mab.fc_v.weight": "model-00004-of-00004.safetensors",
+    "mha_pma_disc.mab.ln0.bias": "model-00004-of-00004.safetensors",
+    "mha_pma_disc.mab.ln0.weight": "model-00004-of-00004.safetensors",
+    "mha_pma_disc.mab.ln1.bias": "model-00004-of-00004.safetensors",
+    "mha_pma_disc.mab.ln1.weight": "model-00004-of-00004.safetensors",
+    "plm_model.lm_head.weight": "model-00004-of-00004.safetensors",
+    "plm_model.model.embed_tokens.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.0.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.0.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.0.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.0.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.0.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.0.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.0.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.0.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.0.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.0.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.0.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.0.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.1.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.1.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.1.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.1.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.1.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.1.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.1.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.1.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.1.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.1.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.1.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.1.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.10.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.10.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.10.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.10.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.10.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.10.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.10.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.10.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.10.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.10.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.10.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.10.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.11.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.11.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.11.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.11.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.11.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.11.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.11.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.11.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.11.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.11.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.11.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.11.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.12.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.12.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.12.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.12.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.12.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.12.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.12.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.12.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.12.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.12.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.12.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.12.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.13.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.13.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.13.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.13.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.13.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.13.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.13.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.13.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.13.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.13.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.13.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.13.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.14.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.14.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.14.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.14.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.14.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.14.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.14.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.14.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.14.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.14.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.14.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.14.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.15.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.15.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.15.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.15.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.15.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.15.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.15.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.15.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.15.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.15.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.15.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.15.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.16.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.16.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.16.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.16.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.16.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.16.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.16.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.16.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.16.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.16.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.16.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.16.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.17.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.17.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.17.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.17.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.17.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.17.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.17.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.17.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.17.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.17.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.17.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.17.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.18.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.18.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.18.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.18.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.18.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.18.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.18.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.18.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.18.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.18.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.18.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.18.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.19.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.19.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.19.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.19.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.19.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.19.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.19.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.19.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.19.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.19.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.19.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.19.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.2.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.2.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.2.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.2.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.2.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.2.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.2.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.2.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.2.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.2.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.2.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.2.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.20.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.20.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.20.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.20.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.20.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.20.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.20.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.20.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.20.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.20.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.20.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.20.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.21.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.21.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.21.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.21.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.21.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.21.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.21.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.21.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.21.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.21.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.21.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.21.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.22.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.22.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.22.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.22.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.22.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.22.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.22.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.22.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.22.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.22.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.22.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.22.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.23.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.23.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.23.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.23.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.23.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.23.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.23.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.23.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.23.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.23.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.23.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.23.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.24.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.24.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.24.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.24.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.24.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.24.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.24.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.24.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.24.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.24.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.24.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.24.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.25.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.25.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.25.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.25.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.25.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.25.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.25.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.25.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.25.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.25.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.25.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.25.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.26.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.26.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.26.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.26.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.26.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.26.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.26.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.26.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.26.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.26.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.26.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.26.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.27.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.27.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.27.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.27.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.27.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.27.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.27.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.27.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.27.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.27.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.27.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.27.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "plm_model.model.layers.3.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.3.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.3.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.3.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.3.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.3.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.3.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.3.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.3.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.3.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.3.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.3.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.4.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.4.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.4.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.4.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.4.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.4.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.4.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.4.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.4.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.4.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.4.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.4.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.5.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.5.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.5.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.5.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.5.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.5.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.5.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.5.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.5.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.5.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.5.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.5.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.6.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.6.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.6.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.6.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.6.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.6.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.6.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.6.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.6.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.6.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.6.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.6.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.7.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.7.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.7.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.7.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.7.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.7.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.7.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.7.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.7.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.7.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.7.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.7.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.8.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.8.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.8.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.8.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.8.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.8.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.8.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.8.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.8.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.8.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.8.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.8.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "plm_model.model.layers.9.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.9.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.9.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.9.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.9.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.9.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.9.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.9.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.9.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.9.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.9.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "plm_model.model.layers.9.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "plm_model.model.norm.weight": "model-00003-of-00004.safetensors"
+  }
+}

modeling_c2llm.py

@@ -0,0 +1,641 @@
+import os
+import gc
+import inspect
+import math
+import multiprocessing as mp
+import queue
+from multiprocessing import Queue
+import warnings
+from typing import Any, Union, List, Dict, Literal, Optional
+import torch
+import torch.nn.functional as F
+import torch.utils.checkpoint
+from torch import nn
+from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
+from transformers import PretrainedConfig
+
+from transformers import Qwen2Config
+from transformers.activations import ACT2FN
+from transformers.cache_utils import Cache, DynamicCache
+from transformers.modeling_attn_mask_utils import _prepare_4d_causal_attention_mask, _prepare_4d_causal_attention_mask_for_sdpa, _prepare_4d_attention_mask, _prepare_4d_attention_mask_for_sdpa
+from transformers.modeling_outputs import BaseModelOutputWithPast, CausalLMOutputWithPast, SequenceClassifierOutputWithPast
+from transformers.modeling_utils import PreTrainedModel
+from transformers.utils import (
+    add_start_docstrings,
+    add_start_docstrings_to_model_forward,
+    is_flash_attn_2_available,
+    is_flash_attn_greater_or_equal_2_10,
+    logging,
+    replace_return_docstrings,
+)
+import numpy as np
+from transformers import Qwen2Config
+from transformers import Qwen2ForCausalLM
+import inspect
+import math
+import os
+import warnings
+from typing import List, Optional, Tuple, Union
+from tqdm import tqdm, trange
+import torch
+import torch.nn.functional as F
+import torch.utils.checkpoint
+from torch import nn
+from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
+
+from transformers.activations import ACT2FN
+from transformers.cache_utils import Cache, DynamicCache
+from transformers.modeling_attn_mask_utils import _prepare_4d_causal_attention_mask, _prepare_4d_causal_attention_mask_for_sdpa, _prepare_4d_attention_mask, _prepare_4d_attention_mask_for_sdpa
+from transformers.modeling_outputs import BaseModelOutputWithPast, CausalLMOutputWithPast, SequenceClassifierOutputWithPast
+from transformers.modeling_utils import PreTrainedModel
+from transformers.utils import (
+    add_start_docstrings,
+    add_start_docstrings_to_model_forward,
+    is_flash_attn_2_available,
+    is_flash_attn_greater_or_equal_2_10,
+    logging,
+    replace_return_docstrings,
+)
+import numpy as np
+import torch
+import os
+import argparse
+import json
+from tqdm import tqdm
+from typing import cast, List, Union, Tuple
+from transformers import AutoTokenizer, AutoModel  # pylint: disable=C0413
+from peft import LoraConfig, get_peft_model, TaskType
+import time
+import torch.nn.functional as F
+import sys
+import time
+import torch 
+import torch.nn as nn
+import torch.nn.functional as F
+import numpy as np
+from tqdm import tqdm, trange
+from collections import defaultdict
+from transformers import AutoTokenizer, AutoModel, AutoModelForCausalLM, AutoConfig
+import torch.distributed as dist
+from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+import sys
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import math
+import re
+import logging
+logging.getLogger().setLevel(logging.INFO)
+from .configuration_c2llm import C2LLMConfig
+from transformers.models.qwen2.modeling_qwen2 import Qwen2DecoderLayer, Qwen2Attention
+
+class MAB_POST(nn.Module):
+    def __init__(self, dim_Q, dim_K, dim_V, num_heads, ln=False):
+        super(MAB_POST, self).__init__()
+        self.dim_V = dim_V
+        self.num_heads = num_heads
+        self.fc_q = nn.Linear(dim_Q, dim_V)
+        self.fc_k = nn.Linear(dim_K, dim_V)
+        self.fc_v = nn.Linear(dim_K, dim_V)
+        if ln:
+            self.ln0 = nn.LayerNorm(dim_V)
+            self.ln1 = nn.LayerNorm(dim_V)
+        self.fc_o = nn.Linear(dim_V, dim_V)
+        nn.init.xavier_uniform_(self.fc_q.weight)
+        nn.init.xavier_uniform_(self.fc_k.weight)
+        nn.init.xavier_uniform_(self.fc_v.weight)
+        nn.init.xavier_uniform_(self.fc_o.weight)
+
+    def forward(self, Q, K, pad_mask=None):
+
+        Q_ = self.fc_q(Q)
+        K_, V_ = self.fc_k(K), self.fc_v(K)
+
+        dim_split = self.dim_V // self.num_heads
+        Q_ = torch.cat(Q_.split(dim_split, 2), 0)
+        K_ = torch.cat(K_.split(dim_split, 2), 0)
+        V_ = torch.cat(V_.split(dim_split, 2), 0)
+
+        pad_mask = pad_mask.unsqueeze(1).repeat(self.num_heads, Q.size(1), 1)
+        score = Q_.bmm(K_.transpose(1,2))/math.sqrt(self.dim_V)
+        score = score.masked_fill(pad_mask == 0, -1e12)
+        A = torch.softmax(score, 2)
+        A = A * pad_mask
+        O = torch.cat(A.bmm(V_).split(Q.size(0), 0), 2) 
+        O = Q + O
+        O = O if getattr(self, 'ln0', None) is None else self.ln0(O)
+        O = O + F.relu(self.fc_o(O))
+        O = O if getattr(self, 'ln1', None) is None else self.ln1(O)
+        return O
+
+
+class PMA(nn.Module):
+    def __init__(self, dim, compressed_dim, num_heads, num_seeds, ln=False, pma_mode=None):
+        super(PMA, self).__init__()
+        self.S = nn.Parameter(torch.Tensor(1, num_seeds, compressed_dim))
+        nn.init.xavier_uniform_(self.S)
+        if pma_mode == 'post_normal':
+            self.mab = MAB_POST(compressed_dim, dim, compressed_dim, num_heads, ln=ln)
+        else:
+            raise ValueError(f"Error, the pma_mode {pma_mode} is not implemented !")
+
+    def forward(self, X, pad_mask):
+        if self.S.dtype != torch.bfloat16:
+            X = X.float()
+        return self.mab(self.S.repeat(X.size(0), 1, 1), X, pad_mask)
+
+
+
+class MAB_POST_v2(nn.Module):
+    def __init__(self, dim_Q, dim_K, dim_V, num_heads, ln=False):
+        super(MAB_POST_v2, self).__init__()
+        self.dim_V = dim_V
+        self.num_heads = num_heads
+        self.fc_q = nn.Linear(dim_Q, dim_V)
+        self.fc_k = nn.Linear(dim_K, dim_V)
+        self.fc_v = nn.Linear(dim_K, dim_V)
+
+        if ln:
+            self.ln0 = nn.LayerNorm(dim_V)
+            self.ln1 = nn.LayerNorm(dim_V)
+        self.fc_o = nn.Linear(dim_V, dim_V)
+        nn.init.xavier_uniform_(self.fc_q.weight)
+        nn.init.xavier_uniform_(self.fc_k.weight)
+        nn.init.xavier_uniform_(self.fc_v.weight)
+        nn.init.xavier_uniform_(self.fc_o.weight)
+
+
+
+    # Q(B, num_seed, D), pad_mask (bs, seq) Post-LN
+    def forward(self, Q, K, pad_mask=None):
+
+        Q_tmp = self.fc_q(Q) # B, num_seed, C
+        K_, V_ = self.fc_k(K), self.fc_v(K) # B, L, C
+
+        dim_split = self.dim_V // self.num_heads
+        Q_ = torch.cat(Q_tmp.split(dim_split, 2), 0) # (B* num_head, num_seed, C)
+        K_ = torch.cat(K_.split(dim_split, 2), 0) # (B* num_head, L, C)
+        V_ = torch.cat(V_.split(dim_split, 2), 0) # (B* num_head,L, C)
+
+        pad_mask = pad_mask.unsqueeze(1).repeat(self.num_heads, Q.size(1), 1) # (B*num_head, num_seed, L)
+        score = Q_.bmm(K_.transpose(1,2))/math.sqrt(self.dim_V) # (B*num_head, num_seed, L) 
+        score = score.masked_fill(pad_mask == 0, -1e12) # B,num_seed,L
+        A = torch.softmax(score, 2)  # (B*num_head, num_seed, L)
+        A = A * pad_mask
+        O = torch.cat(A.bmm(V_).split(Q.size(0), 0), 2) # (B, num_seed, D)
+        O = Q_tmp + O
+        # O = torch.cat((Q_ + A.bmm(V_)).split(Q.size(0), 0), 2)
+        O = O if getattr(self, 'ln0', None) is None else self.ln0(O)
+        O = O + F.relu(self.fc_o(O))
+        O = O if getattr(self, 'ln1', None) is None else self.ln1(O)
+        return O
+
+
+
+
+class PMA_v2(nn.Module):
+    def __init__(self, dim, compressed_dim, num_heads, num_seeds, ln=False):
+        super(PMA_v2, self).__init__()
+        self.S = nn.Parameter(torch.Tensor(1, num_seeds, dim))
+        nn.init.xavier_uniform_(self.S)
+        # if pma_mode == 'post_normal':
+        self.mab = MAB_POST_v2(dim, dim, compressed_dim, num_heads, ln=ln)
+        # elif pma_mode == 'pre_normal':
+            # self.mab = MAB_PRE_NORMAL(dim, dim, compressed_dim, num_heads, ln=ln)
+        # elif pma_mode == 'pre_gptj':
+            # self.mab = MAB_PRE_GPTJ(dim, dim, compressed_dim, num_heads, ln=ln)
+        # else:
+            # raise ValueError(f"Error, the pma_mode {pma_mode} is not implemented !")
+    # X: (bs, seq, emb), pad_mask: (bs, seq)
+    def forward(self, X, pad_mask):
+        if self.S.dtype != torch.bfloat16:
+            X = X.float()
+        return self.mab(self.S.expand(X.size(0), -1, -1), X, pad_mask)
+
+
+class C2LLMModel(PreTrainedModel):
+    config_class = C2LLMConfig 
+    config: C2LLMConfig
+    base_model_prefix = "model"
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["Qwen2DecoderLayer"]
+    _skip_keys_device_placement = ["past_key_values"]
+    _supports_flash_attn = True
+    _supports_sdpa = True
+    _supports_flex_attn = True
+
+    _can_compile_fullgraph = True
+    _supports_attention_backend = True
+    _can_record_outputs = {
+        "hidden_states": Qwen2DecoderLayer,
+        "attentions": Qwen2Attention,
+    }
+
+
+class C2LLMForEmbedding(C2LLMModel):
+
+    config_class = C2LLMConfig
+    model_type   = "c2llm"
+
+    def __init__(self, config):
+        super().__init__(config)
+        qwen_cfg = Qwen2Config.from_dict(config.to_dict())
+        self.plm_model = AutoModelForCausalLM.from_config(qwen_cfg) 
+        self.embedding_method = config.embedding_method
+        self.inf_seq_length = 2048
+        self.padding_side = config.padding_side
+
+        self.emb_dim = self.plm_model.model.embed_tokens.weight.size(1)
+        self.keep_max_layer = self.plm_model.config.num_hidden_layers
+        self.num_heads = config.pma_num_heads
+        self.ln = config.pma_ln
+        self.norm = config.pma_norm
+        self.pma_mode = config.pma_norm_mode
+        self.compressed_dim = config.compressed_dim
+        
+        self.mha_pma_disc = PMA_v2(self.emb_dim, self.compressed_dim, self.num_heads, 1, ln=self.ln)
+        self.pool = None
+        self.target_devices = self.get_target_devices(None)
+        self.tokenizer = AutoTokenizer.from_pretrained(config.tokenizer_name_or_path, padding_side=config.padding_side) if config.tokenizer_name_or_path is not None else None
+        self.config_class = C2LLMConfig
+
+
+    def pma_embedding(self, mha_pma, A, mask):
+        res = mha_pma(A, mask).squeeze(1)
+        return res
+
+    def get_hidden_states(self, **inputs):
+        outputs = self.plm_model(inputs['input_ids'], inputs['attention_mask'], output_hidden_states=True)
+        return outputs.hidden_states[self.keep_max_layer]
+
+    def get_sentence_embedding(self, embedding_method, hidden_states, emb_type, attention_mask):
+
+        
+        if embedding_method == 'pma':
+            
+            if emb_type == 'disc':
+                res_embedding = self.pma_embedding(self.mha_pma_disc, hidden_states, attention_mask)
+                if self.norm:
+                    res_embedding = torch.nn.functional.normalize(res_embedding, p=2.0, dim=-1, eps=1e-12, out=None)
+                return res_embedding
+            else:
+                raise NotImplementedError(f"emb type {emb_type} hasn't been implemented")
+        else:
+            raise NotImplementedError(f"embedding method {embedding_method} hasn't been implemented")
+
+
+    @staticmethod
+    def get_target_devices(devices: Union[str, int, List[str], List[int]]) -> List[str]:
+        """
+
+        Args:
+            devices (Union[str, int, List[str], List[int]]): specified devices, can be `str`, `int`, list of `str`, or list of `int`.
+
+        Raises:
+            ValueError: Devices should be a string or an integer or a list of strings or a list of integers.
+
+        Returns:
+            List[str]: A list of target devices in format.
+        """
+        if devices is None:
+            if torch.cuda.is_available():
+                return [f"cuda:{i}" for i in range(torch.cuda.device_count())]
+            elif is_torch_npu_available():
+                return [f"npu:{i}" for i in range(torch.npu.device_count())]
+            elif hasattr(torch, "musa") and torch.musa.is_available():
+                return [f"musa:{i}" for i in range(torch.musa.device_count())]
+            elif torch.backends.mps.is_available():
+                try:
+                    return [f"mps:{i}" for i in range(torch.mps.device_count())]
+                except:
+                    return ["mps"]
+            else:
+                return ["cpu"]
+        elif isinstance(devices, str):
+            return [devices]
+        elif isinstance(devices, int):
+            if hasattr(torch, "musa") and torch.musa.is_available():
+                return [f"musa:{devices}"]
+            else:
+                return [f"cuda:{devices}"]
+        elif isinstance(devices, list):
+            if isinstance(devices[0], str):
+                return devices
+            elif isinstance(devices[0], int):
+                if hasattr(torch, "musa") and torch.musa.is_available():
+                    return [f"musa:{device}" for device in devices]
+                else:
+                    return [f"cuda:{device}" for device in devices]
+            else:
+                raise ValueError("devices should be a string or an integer or a list of strings or a list of integers.")
+        else:
+            raise ValueError("devices should be a string or an integer or a list of strings or a list of integers.")
+
+        # adapted from https://github.com/UKPLab/sentence-transformers/blob/1802076d4eae42ff0a5629e1b04e75785d4e193b/sentence_transformers/SentenceTransformer.py#L807
+    
+    
+    def start_multi_process_pool(
+        self,
+        process_target_func: Any,
+    ) -> Dict[Literal["input", "output", "processes"], Any]:
+        """
+        Starts a multi-process pool to process the encoding with several independent processes
+        via :meth:`SentenceTransformer.encode_multi_process <sentence_transformers.SentenceTransformer.encode_multi_process>`.
+
+        This method is recommended if you want to encode on multiple GPUs or CPUs. It is advised
+        to start only one process per GPU. This method works together with encode_multi_process
+        and stop_multi_process_pool.
+
+        Returns:
+            Dict[str, Any]: A dictionary with the target processes, an input queue, and an output queue.
+        """
+        if self.plm_model is None or self.mha_pma_disc is None:
+            raise ValueError("Model is not initialized.")
+
+        logging.info("Start multi-process pool on devices: {}".format(", ".join(map(str, self.target_devices))))
+
+        self.to("cpu")
+        self.share_memory()
+        ctx = mp.get_context("spawn")
+        input_queue = ctx.Queue()
+        output_queue = ctx.Queue()
+        processes = []
+
+        for device_id in tqdm(self.target_devices, desc='initial target device'):
+            p = ctx.Process(
+                target=process_target_func,
+                args=(device_id, self, input_queue, output_queue),
+                daemon=True,
+            )
+            p.start()
+            processes.append(p)
+
+        return {"input": input_queue, "output": output_queue, "processes": processes}
+
+
+
+    @staticmethod
+    def _encode_multi_process_worker(
+        target_device: str, model: 'C2LLMForEmbedding', input_queue: Queue, results_queue: Queue
+    ) -> None:
+        model = model.to(target_device)
+        while True:
+            try:
+                chunk_id, sentences, kwargs = (
+                    input_queue.get()
+                )
+                embeddings = model.encode_single_device(
+                    sentences,
+                    device=target_device,
+                    **kwargs
+                )
+
+                results_queue.put([chunk_id, embeddings])
+            except queue.Empty:
+                break
+    
+    def encode_multi_process(
+        self,
+        sentences: List[str],
+        pool: Dict[Literal["input", "output", "processes"], Any],
+        **kwargs
+    ):
+
+        chunk_size = math.ceil(len(sentences) / len(pool["processes"]))
+
+        input_queue = pool["input"]
+        last_chunk_id = 0
+        chunk = []
+
+        for sentence in sentences:
+            chunk.append(sentence)
+            if len(chunk) >= chunk_size:
+                input_queue.put(
+                    [last_chunk_id, chunk, kwargs]
+                )
+                last_chunk_id += 1
+                chunk = []
+
+        if len(chunk) > 0:
+            input_queue.put([last_chunk_id, chunk, kwargs])
+            last_chunk_id += 1
+
+        output_queue = pool["output"]
+        results_list = sorted(
+            [output_queue.get() for _ in trange(last_chunk_id, desc="")],
+            key=lambda x: x[0],
+        )
+        embeddings = self._concatenate_results_from_multi_process([result[1] for result in results_list])
+        return embeddings
+
+    def _concatenate_results_from_multi_process(self, results_list: List[Union[torch.Tensor, np.ndarray, Any]]):
+        """concatenate and return the results from all the processes
+
+        Args:
+            results_list (List[Union[torch.Tensor, np.ndarray, Any]]): A list of results from all the processes.
+
+        Raises:
+            NotImplementedError: Unsupported type for results_list
+
+        Returns:
+            Union[torch.Tensor, np.ndarray]: return the embedding vectors in a numpy array or tensor.
+        """
+        if isinstance(results_list[0], torch.Tensor):
+            # move all tensors to the same device
+            results_list = [res.to(self.target_devices[0]) for res in results_list]
+            return torch.cat(results_list, dim=0)
+        elif isinstance(results_list[0], np.ndarray):
+            return np.concatenate(results_list, axis=0)
+        else:
+            raise NotImplementedError("Unsupported type for results_list")
+
+
+    def forward(self, input_ids: torch.Tensor, attention_mask: torch.Tensor, return_dict: bool=True, **kwargs):
+        outputs = self.plm_model(input_ids, attention_mask, output_hidden_states=True)
+        hidden_states = outputs.hidden_states[self.keep_max_layer]
+        embeddings = self.get_sentence_embedding(self.embedding_method, hidden_states, 'disc', attention_mask)
+        if not return_dict:
+            return (embeddings,)
+        return {"sentence_embedding": embeddings}
+
+    def encode_single_device(
+        self, 
+        sentences: Union[List[str], str],
+        batch_size: int = 16, 
+        convert_to_numpy: bool = False,        
+        convert_to_tensor: bool = True, 
+        show_progress_bar: bool = True, 
+        max_seq_length: int = 2048, 
+        device: Optional[str] = None,
+        **kwargs: Any
+    ):
+        if max_seq_length is None:
+            max_seq_length = self.inf_seq_length
+
+        input_is_string = False        
+        if isinstance(sentences, str) or not hasattr(sentences, "__len__"):
+            sentences = [sentences]
+            input_is_string = True
+        all_embeddings = []
+        length_sorted_idx = np.argsort([-len(s) for s in sentences])
+        sentences_sorted = [sentences[idx] for idx in length_sorted_idx] # 大到小重排
+        with torch.no_grad():
+            for start_index in trange(0, len(sentences), batch_size, desc="Batches", disable=not show_progress_bar):
+                sentences_batch = sentences_sorted[start_index: start_index + batch_size]
+                inputs = self.tokenizer(sentences_batch, padding=True, truncation=True, max_length=max_seq_length, return_tensors='pt').to(self.plm_model.device)
+                hidden_states = self.get_hidden_states(**inputs)
+                embeddings = self.get_sentence_embedding(self.embedding_method, hidden_states, 'disc', inputs['attention_mask'])
+                embeddings = embeddings.detach()
+                if convert_to_numpy:
+                    if embeddings.dtype == torch.bfloat16:
+                        embeddings = embeddings.cpu().to(torch.float32)
+                    else:
+                        embeddings = embeddings.cpu()
+                all_embeddings.extend(embeddings)
+        all_embeddings = [all_embeddings[idx] for idx in np.argsort(length_sorted_idx)]
+        if convert_to_tensor:
+            all_embeddings = torch.stack(all_embeddings)
+        elif convert_to_numpy:
+            all_embeddings = np.asarray([emb.numpy() for emb in all_embeddings])
+
+        if input_is_string:
+            all_embeddings = all_embeddings[0]
+        return all_embeddings
+
+
+    def encode(self, sentences, batch_size=16, convert_to_numpy=False,
+            convert_to_tensor=True, show_progress_bar=True, max_seq_length=None, **kwargs):
+
+        if max_seq_length is None:
+            max_seq_length = self.inf_seq_length
+        
+        if convert_to_tensor == convert_to_numpy:
+            convert_to_tensor=True
+            convert_to_numpy=False
+            
+        if isinstance(sentences, str) or len(self.target_devices) == 1:
+            return self.encode_single_device(
+                sentences,
+                batch_size=batch_size,
+                convert_to_numpy=convert_to_numpy,
+                convert_to_tensor=convert_to_tensor,
+                show_progress_bar=show_progress_bar,
+                max_seq_length=max_seq_length,
+                device=self.target_devices[0],
+                **kwargs
+            )
+        if self.pool is None:
+            self.pool = self.start_multi_process_pool(C2LLMForEmbedding._encode_multi_process_worker)
+        
+
+        all_embeddings = []
+        length_sorted_idx = np.argsort([-len(s) for s in sentences])
+        sentences_sorted = [sentences[idx] for idx in length_sorted_idx] # 大到小重排
+        with torch.no_grad():
+            for start_index in trange(0, len(sentences), batch_size, desc="Batches", disable=not show_progress_bar):
+                sentences_batch = sentences_sorted[start_index: start_index + batch_size]
+                embeddings_batch = self.encode_multi_process(
+                    sentences_batch,
+                    self.pool,
+                    convert_to_numpy=convert_to_numpy,
+                    convert_to_tensor=convert_to_tensor,
+                    show_progress_bar=show_progress_bar,
+                    max_seq_length=max_seq_length,
+                    **kwargs
+                )
+                embeddings_batch = embeddings_batch.detach()
+                if convert_to_numpy:
+                    if embeddings_batch.dtype == torch.bfloat16:
+                        embeddings_batch = embeddings_batch.cpu().to(torch.float32)
+                    else:
+                        embeddings_batch = embeddings_batch.cpu()
+                all_embeddings.extend(embeddings_batch)
+        all_embeddings = [all_embeddings[idx] for idx in np.argsort(length_sorted_idx)]
+        if convert_to_tensor:
+            all_embeddings = torch.stack(all_embeddings)
+        elif convert_to_numpy:
+            all_embeddings = np.asarray([emb.numpy() for emb in all_embeddings])
+
+
+        return all_embeddings
+
+
+    def encode_queries(self, sentences, batch_size=16, convert_to_numpy=False,
+            convert_to_tensor=True, show_progress_bar=True, max_seq_length=None, **kwargs):
+        if max_seq_length is None:
+            max_seq_length = self.inf_seq_length
+        
+        if convert_to_tensor == convert_to_numpy:
+            convert_to_tensor=True
+            convert_to_numpy=False
+
+        return self.encode(
+            sentences=sentences,
+            batch_size=batch_size,
+            convert_to_numpy=convert_to_numpy,
+            convert_to_tensor=convert_to_tensor,
+            show_progress_bar=show_progress_bar,
+            max_seq_length=max_seq_length,
+            **kwargs
+        )
+
+    
+    def encode_corpus(self, sentences, batch_size=16, convert_to_numpy=False,
+            convert_to_tensor=True, show_progress_bar=True, max_seq_length=None, **kwargs):
+
+        if max_seq_length is None:
+            max_seq_length = self.inf_seq_length
+        
+        if convert_to_tensor == convert_to_numpy:
+            convert_to_tensor=True
+            convert_to_numpy=False
+        sentences = [sentence['title']+' '+sentence['text'] for sentence in sentences]
+
+        return self.encode(
+            sentences=sentences,
+            batch_size=batch_size,
+            convert_to_numpy=convert_to_numpy,
+            convert_to_tensor=convert_to_tensor,
+            show_progress_bar=show_progress_bar,
+            max_seq_length=max_seq_length,
+            **kwargs
+    
+        )
+
+    @staticmethod
+    def stop_multi_process_pool(pool: Dict[Literal["input", "output", "processes"], Any]) -> None:
+        """
+        Stops all processes started with start_multi_process_pool.
+
+        Args:
+            pool (Dict[str, object]): A dictionary containing the input queue, output queue, and process list.
+
+        Returns:
+            None
+        """
+        for p in pool["processes"]:
+            p.terminate()
+
+        for p in pool["processes"]:
+            p.join()
+            p.close()
+
+        pool["input"].close()
+        pool["output"].close()
+        pool = None
+
+    def stop_self_pool(self):
+        if self.pool is not None:
+            self.stop_multi_process_pool(self.pool)
+            self.pool = None
+        try:
+            self.model.to('cpu')
+            torch.cuda.empty_cache()
+        except:
+            pass
+        if gc is not None and callable(gc.collect):
+            gc.collect()
+
+    def __del__(self):
+        self.stop_self_pool()
+

modules.json

@@ -0,0 +1,8 @@
+[
+  {
+    "idx": 0,
+    "name": "0",
+    "path": "",
+    "type": "custom_transformer.C2LLMTransformer"
+  }
+]

sentence_bert_config.json

@@ -0,0 +1,4 @@
+{
+  "max_seq_length": 32768,
+  "do_lower_case": false
+}

special_tokens_map.json

@@ -0,0 +1,31 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c5ae00e602b8860cbd784ba82a8aa14e8feecec692e7076590d014d7b7fdafa
+size 11421896

tokenizer_config.json

@@ -0,0 +1,208 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "chat_template": "{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- messages[0]['content'] }}\n    {%- else %}\n        {{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }}\n    {%- endif %}\n    {{- \"\\n\\n# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n    {%- for tool in tools %}\n        {{- \"\\n\" }}\n        {{- tool | tojson }}\n    {%- endfor %}\n    {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- '<|im_start|>system\\n' + messages[0]['content'] + '<|im_end|>\\n' }}\n    {%- else %}\n        {{- '<|im_start|>system\\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- for message in messages %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) or (message.role == \"assistant\" and not message.tool_calls) %}\n        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {{- '<|im_start|>' + message.role }}\n        {%- if message.content %}\n            {{- '\\n' + message.content }}\n        {%- endif %}\n        {%- for tool_call in message.tool_calls %}\n            {%- if tool_call.function is defined %}\n                {%- set tool_call = tool_call.function %}\n            {%- endif %}\n            {{- '\\n<tool_call>\\n{\"name\": \"' }}\n            {{- tool_call.name }}\n            {{- '\", \"arguments\": ' }}\n            {{- tool_call.arguments | tojson }}\n            {{- '}\\n</tool_call>' }}\n        {%- endfor %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- message.content }}\n        {{- '\\n</tool_response>' }}\n        {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n            {{- '<|im_end|>\\n' }}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\\n' }}\n{%- endif %}\n",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "model_max_length": 32768,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}