初始化项目，由ModelHub XC社区提供模型

Model: recogna-nlp/internlm-chatbode-7b
Source: Original Platform

.gitattributes

@@ -0,0 +1,35 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz 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

README.md

@@ -0,0 +1,242 @@
+---
+library_name: transformers
+model-index:
+- name: internlm-chatbode-7b
+  results:
+  - task:
+      type: text-generation
+      name: Text Generation
+    dataset:
+      name: ENEM Challenge (No Images)
+      type: eduagarcia/enem_challenge
+      split: train
+      args:
+        num_few_shot: 3
+    metrics:
+    - type: acc
+      value: 63.05
+      name: accuracy
+    source:
+      url: >-
+        https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=recogna-nlp/internlm-chatbode-7b
+      name: Open Portuguese LLM Leaderboard
+  - task:
+      type: text-generation
+      name: Text Generation
+    dataset:
+      name: BLUEX (No Images)
+      type: eduagarcia-temp/BLUEX_without_images
+      split: train
+      args:
+        num_few_shot: 3
+    metrics:
+    - type: acc
+      value: 51.46
+      name: accuracy
+    source:
+      url: >-
+        https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=recogna-nlp/internlm-chatbode-7b
+      name: Open Portuguese LLM Leaderboard
+  - task:
+      type: text-generation
+      name: Text Generation
+    dataset:
+      name: OAB Exams
+      type: eduagarcia/oab_exams
+      split: train
+      args:
+        num_few_shot: 3
+    metrics:
+    - type: acc
+      value: 42.32
+      name: accuracy
+    source:
+      url: >-
+        https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=recogna-nlp/internlm-chatbode-7b
+      name: Open Portuguese LLM Leaderboard
+  - task:
+      type: text-generation
+      name: Text Generation
+    dataset:
+      name: Assin2 RTE
+      type: assin2
+      split: test
+      args:
+        num_few_shot: 15
+    metrics:
+    - type: f1_macro
+      value: 91.33
+      name: f1-macro
+    source:
+      url: >-
+        https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=recogna-nlp/internlm-chatbode-7b
+      name: Open Portuguese LLM Leaderboard
+  - task:
+      type: text-generation
+      name: Text Generation
+    dataset:
+      name: Assin2 STS
+      type: eduagarcia/portuguese_benchmark
+      split: test
+      args:
+        num_few_shot: 15
+    metrics:
+    - type: pearson
+      value: 80.69
+      name: pearson
+    source:
+      url: >-
+        https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=recogna-nlp/internlm-chatbode-7b
+      name: Open Portuguese LLM Leaderboard
+  - task:
+      type: text-generation
+      name: Text Generation
+    dataset:
+      name: FaQuAD NLI
+      type: ruanchaves/faquad-nli
+      split: test
+      args:
+        num_few_shot: 15
+    metrics:
+    - type: f1_macro
+      value: 79.8
+      name: f1-macro
+    source:
+      url: >-
+        https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=recogna-nlp/internlm-chatbode-7b
+      name: Open Portuguese LLM Leaderboard
+  - task:
+      type: text-generation
+      name: Text Generation
+    dataset:
+      name: HateBR Binary
+      type: ruanchaves/hatebr
+      split: test
+      args:
+        num_few_shot: 25
+    metrics:
+    - type: f1_macro
+      value: 87.99
+      name: f1-macro
+    source:
+      url: >-
+        https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=recogna-nlp/internlm-chatbode-7b
+      name: Open Portuguese LLM Leaderboard
+  - task:
+      type: text-generation
+      name: Text Generation
+    dataset:
+      name: PT Hate Speech Binary
+      type: hate_speech_portuguese
+      split: test
+      args:
+        num_few_shot: 25
+    metrics:
+    - type: f1_macro
+      value: 68.09
+      name: f1-macro
+    source:
+      url: >-
+        https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=recogna-nlp/internlm-chatbode-7b
+      name: Open Portuguese LLM Leaderboard
+  - task:
+      type: text-generation
+      name: Text Generation
+    dataset:
+      name: tweetSentBR
+      type: eduagarcia/tweetsentbr_fewshot
+      split: test
+      args:
+        num_few_shot: 25
+    metrics:
+    - type: f1_macro
+      value: 61.11
+      name: f1-macro
+    source:
+      url: >-
+        https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=recogna-nlp/internlm-chatbode-7b
+      name: Open Portuguese LLM Leaderboard
+language:
+- pt
+pipeline_tag: text-generation
+---
+
+# internlm-chatbode-7b
+
+<!--- PROJECT LOGO -->
+<p align="center">
+  <img src="https://huggingface.co/recogna-nlp/internlm-chatbode-7b/resolve/main/_1add1e52-f428-4c7c-bab2-3c6958e029fa.jpeg" alt="ChatBode Logo" width="400" style="margin-left:'auto' margin-right:'auto' display:'block'"/>
+</p>
+
+O InternLm-ChatBode é um modelo de linguagem ajustado para o idioma português, desenvolvido a partir do modelo [InternLM2](https://huggingface.co/internlm/internlm2-chat-7b). Este modelo foi refinado através do processo de fine-tuning utilizando o dataset UltraAlpaca.
+
+## Características Principais
+
+- **Modelo Base:** [internlm/internlm2-chat-7b](internlm/internlm2-chat-7b)
+- **Dataset para Fine-tuning:** UltraAlpaca
+- **Treinamento:** O treinamento foi realizado a partir do fine-tuning, usando QLoRA, do internlm2-chat-7b.
+
+## Exemplo de uso
+
+A seguir um exemplo de código de como carregar e utilizar o modelo:
+
+```python
+import torch
+from transformers import AutoTokenizer, AutoModelForCausalLM
+tokenizer = AutoTokenizer.from_pretrained("recogna-nlp/internlm-chatbode-7b", trust_remote_code=True)
+model = AutoModelForCausalLM.from_pretrained("recogna-nlp/internlm-chatbode-7b", torch_dtype=torch.float16, trust_remote_code=True).cuda()
+model = model.eval()
+response, history = model.chat(tokenizer, "Olá", history=[])
+print(response)
+response, history = model.chat(tokenizer, "O que é o Teorema de Pitágoras? Me dê um exemplo", history=history)
+print(response)
+```
+
+As respostas podem ser geradas via stream utilizando o método `stream_chat`:
+
+```python
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+model_path = "recogna-nlp/internlm-chatbode-7b"
+model = AutoModelForCausalLM.from_pretrained(model_path, torch_dtype=torch.float16, trust_remote_code=True).cuda()
+tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
+
+model = model.eval()
+length = 0
+for response, history in model.stream_chat(tokenizer, "Olá", history=[]):
+    print(response[length:], flush=True, end="")
+    length = len(response)
+```
+
+# Open Portuguese LLM Leaderboard Evaluation Results  
+
+Detailed results can be found [here](https://huggingface.co/datasets/eduagarcia-temp/llm_pt_leaderboard_raw_results/tree/main/recogna-nlp/internlm-chatbode-7b) and on the [🚀 Open Portuguese LLM Leaderboard](https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard)
+
+|          Metric          |  Value  |
+|--------------------------|---------|
+|Average                   |**69.54**|
+|ENEM Challenge (No Images)|    63.05|
+|BLUEX (No Images)         |    51.46|
+|OAB Exams                 |    42.32|
+|Assin2 RTE                |    91.33|
+|Assin2 STS                |    80.69|
+|FaQuAD NLI                |    79.80|
+|HateBR Binary             |    87.99|
+|PT Hate Speech Binary     |    68.09|
+|tweetSentBR               |    61.11|
+
+
+## Citação
+Se você deseja utilizar o Chatbode em sua pesquisa, cite-o da seguinte maneira:
+
+```
+@misc {chatbode_2024,
+	author       = { Gabriel Lino Garcia, Pedro Henrique Paiola and  and João Paulo Papa},
+	title        = { Chatbode },
+	year         = {2024},
+	url          = { https://huggingface.co/recogna-nlp/internlm-chatbode-7b/ },
+	doi          = { 10.57967/hf/3317 },
+	publisher    = { Hugging Face }
+}
+```

config.json

@@ -0,0 +1,36 @@
+{
+  "_name_or_path": "recogna-nlp/internlm-chatbode-7b",
+  "architectures": [
+    "InternLM2ForCausalLM"
+  ],
+  "attn_implementation": "eager",
+  "auto_map": {
+    "AutoConfig": "configuration_internlm2.InternLM2Config",
+    "AutoModel": "imodeling_internlm2.InternLM2ForCausalLM",
+    "AutoModelForCausalLM": "modeling_internlm2.InternLM2ForCausalLM"
+  },
+  "bias": false,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "intermediate_size": 14336,
+  "max_position_embeddings": 32768,
+  "model_type": "internlm2",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "num_key_value_heads": 8,
+  "pad_token_id": 2,
+  "rms_norm_eps": 1e-05,
+  "rope_scaling": {
+    "factor": 2.0,
+    "type": "dynamic"
+  },
+  "rope_theta": 1000000,
+  "tie_word_embeddings": false,
+  "torch_dtype": "float16",
+  "transformers_version": "4.40.0.dev0",
+  "use_cache": true,
+  "vocab_size": 92544
+}

configuration_internlm2.py

@@ -0,0 +1,151 @@
+# coding=utf-8
+# Copyright (c) The InternLM team and The HuggingFace Inc. team. All rights reserved.
+#
+# This code is based on transformers/src/transformers/models/llama/configuration_llama.py
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+""" InternLM2 model configuration"""
+
+from transformers.configuration_utils import PretrainedConfig
+from transformers.utils import logging
+
+logger = logging.get_logger(__name__)
+
+INTERNLM2_PRETRAINED_CONFIG_ARCHIVE_MAP = {}
+
+
+# Modified from transformers.model.llama.configuration_llama.LlamaConfig
+class InternLM2Config(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`InternLM2Model`]. It is used to instantiate
+    an InternLM2 model according to the specified arguments, defining the model architecture. Instantiating a
+    configuration with the defaults will yield a similar configuration to that of the InternLM2-7B.
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+
+    Args:
+        vocab_size (`int`, *optional*, defaults to 32000):
+            Vocabulary size of the InternLM2 model. Defines the number of different tokens that can be represented by the
+            `inputs_ids` passed when calling [`InternLM2Model`]
+        hidden_size (`int`, *optional*, defaults to 4096):
+            Dimension of the hidden representations.
+        intermediate_size (`int`, *optional*, defaults to 11008):
+            Dimension of the MLP representations.
+        num_hidden_layers (`int`, *optional*, defaults to 32):
+            Number of hidden layers in the Transformer encoder.
+        num_attention_heads (`int`, *optional*, defaults to 32):
+            Number of attention heads for each attention layer in the Transformer encoder.
+        num_key_value_heads (`int`, *optional*):
+            This is the number of key_value heads that should be used to implement Grouped Query Attention. If
+            `num_key_value_heads=num_attention_heads`, the model will use Multi Head Attention (MHA), if
+            `num_key_value_heads=1 the model will use Multi Query Attention (MQA) otherwise GQA is used. When
+            converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed
+            by meanpooling all the original heads within that group. For more details checkout [this
+            paper](https://arxiv.org/pdf/2305.13245.pdf). If it is not specified, will default to
+            `num_attention_heads`.
+        hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
+            The non-linear activation function (function or string) in the decoder.
+        max_position_embeddings (`int`, *optional*, defaults to 2048):
+            The maximum sequence length that this model might ever be used with. Typically set this to something large
+            just in case (e.g., 512 or 1024 or 2048).
+        initializer_range (`float`, *optional*, defaults to 0.02):
+            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
+        rms_norm_eps (`float`, *optional*, defaults to 1e-12):
+            The epsilon used by the rms normalization layers.
+        use_cache (`bool`, *optional*, defaults to `True`):
+            Whether or not the model should return the last key/values attentions (not used by all models). Only
+            relevant if `config.is_decoder=True`.
+        tie_word_embeddings(`bool`, *optional*, defaults to `False`):
+            Whether to tie weight embeddings
+        Example:
+
+    """
+    model_type = "internlm2"
+    _auto_class = "AutoConfig"
+
+    def __init__(  # pylint: disable=W0102
+        self,
+        vocab_size=103168,
+        hidden_size=4096,
+        intermediate_size=11008,
+        num_hidden_layers=32,
+        num_attention_heads=32,
+        num_key_value_heads=None,
+        hidden_act="silu",
+        max_position_embeddings=2048,
+        initializer_range=0.02,
+        rms_norm_eps=1e-6,
+        use_cache=True,
+        pad_token_id=0,
+        bos_token_id=1,
+        eos_token_id=2,
+        tie_word_embeddings=False,
+        bias=True,
+        rope_theta=10000,
+        rope_scaling=None,
+        attn_implementation="eager",
+        **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.bias = bias
+
+        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.rope_scaling = rope_scaling
+        self._rope_scaling_validation()
+
+        self.attn_implementation = attn_implementation
+        if self.attn_implementation is None:
+            self.attn_implementation = "eager"
+        super().__init__(
+            pad_token_id=pad_token_id,
+            bos_token_id=bos_token_id,
+            eos_token_id=eos_token_id,
+            tie_word_embeddings=tie_word_embeddings,
+            **kwargs,
+        )
+
+    def _rope_scaling_validation(self):
+        """
+        Validate the `rope_scaling` configuration.
+        """
+        if self.rope_scaling is None:
+            return
+
+        if not isinstance(self.rope_scaling, dict) or len(self.rope_scaling) != 2:
+            raise ValueError(
+                "`rope_scaling` must be a dictionary with with two fields, `type` and `factor`, "
+                f"got {self.rope_scaling}"
+            )
+        rope_scaling_type = self.rope_scaling.get("type", None)
+        rope_scaling_factor = self.rope_scaling.get("factor", None)
+        if rope_scaling_type is None or rope_scaling_type not in ["linear", "dynamic"]:
+            raise ValueError(
+                f"`rope_scaling`'s type field must be one of ['linear', 'dynamic'], got {rope_scaling_type}"
+            )
+        if rope_scaling_factor is None or not isinstance(rope_scaling_factor, float) or rope_scaling_factor < 1.0:
+            raise ValueError(f"`rope_scaling`'s factor field must be a float >= 1, got {rope_scaling_factor}")

generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "pad_token_id": 2,
+  "transformers_version": "4.40.0.dev0"
+}

model-00001-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c635ff22d4fb7519699e70d804ab7a6a173609d0f44739afa0042b75c1ec140
+size 4885470560

model-00002-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:994b32f8085f044c644e9391c0ea9da395de83fb538ee9cc28502f7c489c6222
+size 4915913344

model-00003-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba016d4abdc7c00a96055208d02bda3eb78c64512a13a36ae45edbae2b9365f5
+size 4915938224

model-00004-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8e4b668307c5bf6f5fc63b45d90bb8e4894723a478c49d41d4e18a605db4b80d
+size 758120576

model.safetensors.index.json

@@ -0,0 +1,234 @@
+{
+  "metadata": {
+    "total_size": 15475417088
+  },
+  "weight_map": {
+    "model.layers.0.attention.wo.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.attention.wqkv.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.attention_norm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.feed_forward.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.feed_forward.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.feed_forward.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.ffn_norm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.attention.wo.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.attention.wqkv.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.attention_norm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.feed_forward.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.feed_forward.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.feed_forward.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.ffn_norm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.10.attention.wo.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.attention.wqkv.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.attention_norm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.feed_forward.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.feed_forward.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.feed_forward.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.ffn_norm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.attention.wo.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.attention.wqkv.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.attention_norm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.feed_forward.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.feed_forward.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.feed_forward.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.ffn_norm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.attention.wo.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.attention.wqkv.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.attention_norm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.feed_forward.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.feed_forward.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.feed_forward.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.ffn_norm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.attention.wo.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.attention.wqkv.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.attention_norm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.feed_forward.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.feed_forward.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.feed_forward.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.ffn_norm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.attention.wo.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.attention.wqkv.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.attention_norm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.feed_forward.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.feed_forward.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.feed_forward.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.ffn_norm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.attention.wo.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.attention.wqkv.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.attention_norm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.feed_forward.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.feed_forward.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.feed_forward.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.ffn_norm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.attention.wo.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.attention.wqkv.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.attention_norm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.feed_forward.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.feed_forward.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.feed_forward.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.ffn_norm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.attention.wo.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.attention.wqkv.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.attention_norm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.feed_forward.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.feed_forward.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.feed_forward.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.ffn_norm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.attention.wo.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.attention.wqkv.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.attention_norm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.feed_forward.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.feed_forward.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.feed_forward.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.ffn_norm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.19.attention.wo.weight": "model-00002-of-00004.safetensors",
+    "model.layers.19.attention.wqkv.weight": "model-00002-of-00004.safetensors",
+    "model.layers.19.attention_norm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.19.feed_forward.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.19.feed_forward.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.19.feed_forward.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.19.ffn_norm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.2.attention.wo.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.attention.wqkv.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.attention_norm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.feed_forward.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.feed_forward.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.feed_forward.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.ffn_norm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.20.attention.wo.weight": "model-00002-of-00004.safetensors",
+    "model.layers.20.attention.wqkv.weight": "model-00002-of-00004.safetensors",
+    "model.layers.20.attention_norm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.feed_forward.w1.weight": "model-00002-of-00004.safetensors",
+    "model.layers.20.feed_forward.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.feed_forward.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.20.ffn_norm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.attention.wo.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.attention.wqkv.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.attention_norm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.feed_forward.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.feed_forward.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.feed_forward.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.ffn_norm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.attention.wo.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.attention.wqkv.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.attention_norm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.feed_forward.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.feed_forward.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.feed_forward.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.ffn_norm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.attention.wo.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.attention.wqkv.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.attention_norm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.feed_forward.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.feed_forward.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.feed_forward.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.ffn_norm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.attention.wo.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.attention.wqkv.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.attention_norm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.feed_forward.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.feed_forward.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.feed_forward.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.ffn_norm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.attention.wo.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.attention.wqkv.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.attention_norm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.feed_forward.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.feed_forward.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.feed_forward.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.ffn_norm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.attention.wo.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.attention.wqkv.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.attention_norm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.feed_forward.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.feed_forward.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.feed_forward.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.ffn_norm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.attention.wo.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.attention.wqkv.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.attention_norm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.feed_forward.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.feed_forward.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.feed_forward.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.ffn_norm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.28.attention.wo.weight": "model-00003-of-00004.safetensors",
+    "model.layers.28.attention.wqkv.weight": "model-00003-of-00004.safetensors",
+    "model.layers.28.attention_norm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.28.feed_forward.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.28.feed_forward.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.28.feed_forward.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.28.ffn_norm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.29.attention.wo.weight": "model-00003-of-00004.safetensors",
+    "model.layers.29.attention.wqkv.weight": "model-00003-of-00004.safetensors",
+    "model.layers.29.attention_norm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.29.feed_forward.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.29.feed_forward.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.29.feed_forward.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.29.ffn_norm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.3.attention.wo.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.attention.wqkv.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.attention_norm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.feed_forward.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.feed_forward.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.feed_forward.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.ffn_norm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.30.attention.wo.weight": "model-00003-of-00004.safetensors",
+    "model.layers.30.attention.wqkv.weight": "model-00003-of-00004.safetensors",
+    "model.layers.30.attention_norm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.30.feed_forward.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.30.feed_forward.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.30.feed_forward.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.30.ffn_norm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.31.attention.wo.weight": "model-00003-of-00004.safetensors",
+    "model.layers.31.attention.wqkv.weight": "model-00003-of-00004.safetensors",
+    "model.layers.31.attention_norm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.31.feed_forward.w1.weight": "model-00003-of-00004.safetensors",
+    "model.layers.31.feed_forward.w2.weight": "model-00003-of-00004.safetensors",
+    "model.layers.31.feed_forward.w3.weight": "model-00003-of-00004.safetensors",
+    "model.layers.31.ffn_norm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.4.attention.wo.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.attention.wqkv.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.attention_norm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.feed_forward.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.feed_forward.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.feed_forward.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.ffn_norm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.attention.wo.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.attention.wqkv.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.attention_norm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.feed_forward.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.feed_forward.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.feed_forward.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.ffn_norm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.attention.wo.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.attention.wqkv.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.attention_norm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.feed_forward.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.feed_forward.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.feed_forward.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.ffn_norm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.attention.wo.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.attention.wqkv.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.attention_norm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.feed_forward.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.feed_forward.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.feed_forward.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.ffn_norm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.8.attention.wo.weight": "model-00001-of-00004.safetensors",
+    "model.layers.8.attention.wqkv.weight": "model-00001-of-00004.safetensors",
+    "model.layers.8.attention_norm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.8.feed_forward.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.8.feed_forward.w2.weight": "model-00001-of-00004.safetensors",
+    "model.layers.8.feed_forward.w3.weight": "model-00001-of-00004.safetensors",
+    "model.layers.8.ffn_norm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.9.attention.wo.weight": "model-00001-of-00004.safetensors",
+    "model.layers.9.attention.wqkv.weight": "model-00001-of-00004.safetensors",
+    "model.layers.9.attention_norm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.feed_forward.w1.weight": "model-00001-of-00004.safetensors",
+    "model.layers.9.feed_forward.w2.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.feed_forward.w3.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.ffn_norm.weight": "model-00002-of-00004.safetensors",
+    "model.norm.weight": "model-00003-of-00004.safetensors",
+    "model.tok_embeddings.weight": "model-00001-of-00004.safetensors",
+    "output.weight": "model-00004-of-00004.safetensors"
+  }
+}

special_tokens_map.json

@@ -0,0 +1,38 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|action_start|>",
+    "<|action_end|>",
+    "<|interpreter|>",
+    "<|plugin|>"
+  ],
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenization_internlm2.py

@@ -0,0 +1,236 @@
+# coding=utf-8
+# Copyright (c) The InternLM team and The HuggingFace Inc. team. All rights reserved.
+#
+# This code is based on transformers/src/transformers/models/llama/tokenization_llama.py
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""Tokenization classes for InternLM."""
+import os
+from shutil import copyfile
+from typing import Any, Dict, List, Optional, Tuple
+
+import sentencepiece as spm
+from transformers.tokenization_utils import PreTrainedTokenizer
+from transformers.utils import logging
+
+logger = logging.get_logger(__name__)
+
+VOCAB_FILES_NAMES = {"vocab_file": "./tokenizer.model"}
+
+PRETRAINED_VOCAB_FILES_MAP = {}
+
+
+# Modified from transformers.model.llama.tokenization_llama.LlamaTokenizer
+class InternLM2Tokenizer(PreTrainedTokenizer):
+    """
+    Construct a InternLM2 tokenizer. Based on byte-level Byte-Pair-Encoding.
+
+    Args:
+        vocab_file (`str`):
+            Path to the vocabulary file.
+    """
+
+    vocab_files_names = VOCAB_FILES_NAMES
+    pretrained_vocab_files_map = PRETRAINED_VOCAB_FILES_MAP
+    model_input_names = ["input_ids", "attention_mask"]
+    _auto_class = "AutoTokenizer"
+
+    def __init__(
+        self,
+        vocab_file,
+        unk_token="<unk>",
+        bos_token="<s>",
+        eos_token="</s>",
+        pad_token="</s>",
+        sp_model_kwargs: Optional[Dict[str, Any]] = None,
+        add_bos_token=True,
+        add_eos_token=False,
+        decode_with_prefix_space=False,
+        clean_up_tokenization_spaces=False,
+        **kwargs,
+    ):
+        self.sp_model_kwargs = {} if sp_model_kwargs is None else sp_model_kwargs
+        self.vocab_file = vocab_file
+        self.add_bos_token = add_bos_token
+        self.add_eos_token = add_eos_token
+        self.decode_with_prefix_space = decode_with_prefix_space
+        self.sp_model = spm.SentencePieceProcessor(**self.sp_model_kwargs)
+        self.sp_model.Load(vocab_file)
+        self._no_prefix_space_tokens = None
+        super().__init__(
+            bos_token=bos_token,
+            eos_token=eos_token,
+            unk_token=unk_token,
+            pad_token=pad_token,
+            clean_up_tokenization_spaces=clean_up_tokenization_spaces,
+            **kwargs,
+        )
+
+    @property
+    def no_prefix_space_tokens(self):
+        if self._no_prefix_space_tokens is None:
+            vocab = self.convert_ids_to_tokens(list(range(self.vocab_size)))
+            self._no_prefix_space_tokens = {i for i, tok in enumerate(vocab) if not tok.startswith("▁")}
+        return self._no_prefix_space_tokens
+
+    @property
+    def vocab_size(self):
+        """Returns vocab size"""
+        return self.sp_model.get_piece_size()
+
+    @property
+    def bos_token_id(self) -> Optional[int]:
+        return self.sp_model.bos_id()
+
+    @property
+    def eos_token_id(self) -> Optional[int]:
+        return self.sp_model.eos_id()
+
+    def get_vocab(self):
+        """Returns vocab as a dict"""
+        vocab = {self.convert_ids_to_tokens(i): i for i in range(self.vocab_size)}
+        vocab.update(self.added_tokens_encoder)
+        return vocab
+
+    def _tokenize(self, text):
+        """Returns a tokenized string."""
+        return self.sp_model.encode(text, out_type=str)
+
+    def _convert_token_to_id(self, token):
+        """Converts a token (str) in an id using the vocab."""
+        return self.sp_model.piece_to_id(token)
+
+    def _convert_id_to_token(self, index):
+        """Converts an index (integer) in a token (str) using the vocab."""
+        token = self.sp_model.IdToPiece(index)
+        return token
+
+    def _maybe_add_prefix_space(self, tokens, decoded):
+        if tokens and tokens[0] not in self.no_prefix_space_tokens:
+            return " " + decoded
+        else:
+            return decoded
+
+    def convert_tokens_to_string(self, tokens):
+        """Converts a sequence of tokens (string) in a single string."""
+        current_sub_tokens = []
+        out_string = ""
+        prev_is_special = False
+        for token in tokens:
+            # make sure that special tokens are not decoded using sentencepiece model
+            if token in self.all_special_tokens:
+                if not prev_is_special:
+                    out_string += " "
+                out_string += self.sp_model.decode(current_sub_tokens) + token
+                prev_is_special = True
+                current_sub_tokens = []
+            else:
+                current_sub_tokens.append(token)
+                prev_is_special = False
+        out_string += self.sp_model.decode(current_sub_tokens)
+        out_string = self.clean_up_tokenization(out_string)
+        out_string = self._maybe_add_prefix_space(tokens=tokens, decoded=out_string)
+        return out_string[1:]
+
+    def save_vocabulary(self, save_directory, filename_prefix: Optional[str] = None) -> Tuple[str]:
+        """
+        Save the vocabulary and special tokens file to a directory.
+
+        Args:
+            save_directory (`str`):
+                The directory in which to save the vocabulary.
+
+        Returns:
+            `Tuple(str)`: Paths to the files saved.
+        """
+        if not os.path.isdir(save_directory):
+            logger.error(f"Vocabulary path ({save_directory}) should be a directory")
+            return
+        out_vocab_file = os.path.join(
+            save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
+        )
+
+        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file) and os.path.isfile(self.vocab_file):
+            copyfile(self.vocab_file, out_vocab_file)
+        elif not os.path.isfile(self.vocab_file):
+            with open(out_vocab_file, "wb") as fi:
+                content_spiece_model = self.sp_model.serialized_model_proto()
+                fi.write(content_spiece_model)
+
+        return (out_vocab_file,)
+
+    def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=None):
+        if self.add_bos_token:
+            bos_token_ids = [self.bos_token_id]
+        else:
+            bos_token_ids = []
+
+        output = bos_token_ids + token_ids_0
+
+        if token_ids_1 is not None:
+            output = output + token_ids_1
+
+        if self.add_eos_token:
+            output = output + [self.eos_token_id]
+
+        return output
+
+    def get_special_tokens_mask(
+        self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None, already_has_special_tokens: bool = False
+    ) -> List[int]:
+        """
+        Retrieve sequence ids from a token list that has no special tokens added. This method is called when adding
+        special tokens using the tokenizer `prepare_for_model` method.
+
+        Args:
+            token_ids_0 (`List[int]`):
+                List of IDs.
+            token_ids_1 (`List[int]`, *optional*):
+                Optional second list of IDs for sequence pairs.
+            already_has_special_tokens (`bool`, *optional*, defaults to `False`):
+                Whether or not the token list is already formatted with special tokens for the model.
+
+        Returns:
+            `List[int]`: A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token.
+        """
+        if already_has_special_tokens:
+            return super().get_special_tokens_mask(
+                token_ids_0=token_ids_0, token_ids_1=token_ids_1, already_has_special_tokens=True
+            )
+
+        if token_ids_1 is None:
+            return [1] + ([0] * len(token_ids_0)) + [1]
+        return [1] + ([0] * len(token_ids_0)) + [1, 1] + ([0] * len(token_ids_1)) + [1]
+
+    def create_token_type_ids_from_sequences(
+        self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None
+    ) -> List[int]:
+        """
+        Create a mask from the two sequences passed to be used in a sequence-pair classification task. T5 does not make
+        use of token type ids, therefore a list of zeros is returned.
+
+        Args:
+            token_ids_0 (`List[int]`):
+                List of IDs.
+            token_ids_1 (`List[int]`, *optional*):
+                Optional second list of IDs for sequence pairs.
+
+        Returns:
+            `List[int]`: List of zeros.
+        """
+        eos = [self.eos_token_id]
+
+        if token_ids_1 is None:
+            return len(token_ids_0 + eos) * [0]
+        return len(token_ids_0 + eos + token_ids_1 + eos) * [0]

tokenization_internlm2_fast.py

@@ -0,0 +1,214 @@
+# coding=utf-8
+# Copyright (c) The InternLM team and The HuggingFace Inc. team. All rights reserved.
+#
+# This code is based on transformers/src/transformers/models/llama/tokenization_llama_fast.py
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""Tokenization Fast class for InternLM."""
+import os
+from shutil import copyfile
+from typing import Any, Dict, Optional, Tuple
+
+from tokenizers import processors, decoders, Tokenizer, normalizers
+from tokenizers.models import BPE
+
+from transformers.tokenization_utils_fast import PreTrainedTokenizerFast
+from transformers.utils import logging
+
+from transformers.convert_slow_tokenizer import (
+    SLOW_TO_FAST_CONVERTERS,
+    SpmConverter,
+    SentencePieceExtractor,
+)
+
+from .tokenization_internlm2 import InternLM2Tokenizer
+
+logger = logging.get_logger(__name__)
+
+VOCAB_FILES_NAMES = {"vocab_file": "./tokenizer.model"}
+
+# Modified from transformers.convert_slow_tokenizer.LlamaConverter
+class InternLM2Converter(SpmConverter):
+    handle_byte_fallback = True
+
+    def vocab(self, proto):
+        vocab = [
+            ("<unk>", 0.0),
+            ("<s>", 0.0),
+            ("</s>", 0.0),
+        ]
+        vocab += [(piece.piece, piece.score) for piece in proto.pieces[3:]]
+        return vocab
+
+    def unk_id(self, proto):
+        unk_id = 0
+        return unk_id
+
+    def decoder(self, replacement, add_prefix_space):
+        decoders_sequence = [
+            decoders.Replace("▁", " "),
+            decoders.ByteFallback(),
+            decoders.Fuse(),
+        ]
+        if self.proto.normalizer_spec.add_dummy_prefix:
+            decoders_sequence.append(decoders.Strip(content=" ", left=1))
+        return decoders.Sequence(decoders_sequence)
+
+    def tokenizer(self, proto):
+        model_type = proto.trainer_spec.model_type
+        vocab_scores = self.vocab(proto)
+        # special tokens
+        added_tokens = self.original_tokenizer.added_tokens_decoder
+        for i in range(len(vocab_scores)):
+            piece, score = vocab_scores[i]
+            if i in added_tokens:
+                vocab_scores[i] = (added_tokens[i].content, score)
+        if model_type == 1:
+            raise RuntimeError("InternLM2 is supposed to be a BPE model!")
+
+        elif model_type == 2:
+            _, merges = SentencePieceExtractor(self.original_tokenizer.vocab_file).extract(vocab_scores)
+            bpe_vocab = {word: i for i, (word, _score) in enumerate(vocab_scores)}
+            tokenizer = Tokenizer(
+                BPE(bpe_vocab, merges, unk_token=proto.trainer_spec.unk_piece, fuse_unk=True, byte_fallback=True)
+            )
+            tokenizer.add_special_tokens(
+                [ added_token for index, added_token in added_tokens.items()]
+            )
+        else:
+            raise Exception(
+                "You're trying to run a `Unigram` model but you're file was trained with a different algorithm"
+            )
+
+        return tokenizer
+
+    def normalizer(self, proto):
+        normalizers_list = []
+        if proto.normalizer_spec.add_dummy_prefix:
+            normalizers_list.append(normalizers.Prepend(prepend="▁"))
+        normalizers_list.append(normalizers.Replace(pattern=" ", content="▁"))
+        return normalizers.Sequence(normalizers_list)
+
+    def pre_tokenizer(self, replacement, add_prefix_space):
+        return None
+
+SLOW_TO_FAST_CONVERTERS["InternLM2Tokenizer"] = InternLM2Converter
+
+
+# Modified from transformers.model.llama.tokenization_llama_fast.LlamaTokenizerFast -> InternLM2TokenizerFast
+class InternLM2TokenizerFast(PreTrainedTokenizerFast):
+    vocab_files_names = VOCAB_FILES_NAMES
+    slow_tokenizer_class = InternLM2Tokenizer
+    padding_side = "left"
+    model_input_names = ["input_ids", "attention_mask"]
+    _auto_class = "AutoTokenizer"
+
+    def __init__(
+        self,
+        vocab_file,
+        unk_token="<unk>",
+        bos_token="<s>",
+        eos_token="</s>",
+        pad_token="</s>",
+        sp_model_kwargs: Optional[Dict[str, Any]] = None,
+        add_bos_token=True,
+        add_eos_token=False,
+        decode_with_prefix_space=False,
+        clean_up_tokenization_spaces=False,
+        **kwargs,
+    ):
+        super().__init__(
+            vocab_file=vocab_file,
+            unk_token=unk_token,
+            bos_token=bos_token,
+            eos_token=eos_token,
+            pad_token=pad_token,
+            sp_model_kwargs=sp_model_kwargs,
+            add_bos_token=add_bos_token,
+            add_eos_token=add_eos_token,
+            decode_with_prefix_space=decode_with_prefix_space,
+            clean_up_tokenization_spaces=clean_up_tokenization_spaces,
+            **kwargs,
+        )
+        self._add_bos_token = add_bos_token
+        self._add_eos_token = add_eos_token
+        self.update_post_processor()
+        self.vocab_file = vocab_file
+
+    @property
+    def can_save_slow_tokenizer(self) -> bool:
+        return os.path.isfile(self.vocab_file) if self.vocab_file else False
+
+    def update_post_processor(self):
+        """
+        Updates the underlying post processor with the current `bos_token` and `eos_token`.
+        """
+        bos = self.bos_token
+        bos_token_id = self.bos_token_id
+        if bos is None and self.add_bos_token:
+            raise ValueError("add_bos_token = True but bos_token = None")
+
+        eos = self.eos_token
+        eos_token_id = self.eos_token_id
+        if eos is None and self.add_eos_token:
+            raise ValueError("add_eos_token = True but eos_token = None")
+
+        single = f"{(bos+':0 ') if self.add_bos_token else ''}$A:0{(' '+eos+':0') if self.add_eos_token else ''}"
+        pair = f"{single}{(' '+bos+':1') if self.add_bos_token else ''} $B:1{(' '+eos+':1') if self.add_eos_token else ''}"
+
+        special_tokens = []
+        if self.add_bos_token:
+            special_tokens.append((bos, bos_token_id))
+        if self.add_eos_token:
+            special_tokens.append((eos, eos_token_id))
+        self._tokenizer.post_processor = processors.TemplateProcessing(
+            single=single, pair=pair, special_tokens=special_tokens
+        )
+
+    @property
+    def add_eos_token(self):
+        return self._add_eos_token
+
+    @property
+    def add_bos_token(self):
+        return self._add_bos_token
+
+    @add_eos_token.setter
+    def add_eos_token(self, value):
+        self._add_eos_token = value
+        self.update_post_processor()
+
+    @add_bos_token.setter
+    def add_bos_token(self, value):
+        self._add_bos_token = value
+        self.update_post_processor()
+
+    def save_vocabulary(self, save_directory: str, filename_prefix: Optional[str] = None) -> Tuple[str]:
+        if not self.can_save_slow_tokenizer:
+            raise ValueError(
+                "Your fast tokenizer does not have the necessary information to save the vocabulary for a slow "
+                "tokenizer."
+            )
+
+        if not os.path.isdir(save_directory):
+            logger.error(f"Vocabulary path ({save_directory}) should be a directory")
+            return
+        out_vocab_file = os.path.join(
+            save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
+        )
+
+        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file):
+            copyfile(self.vocab_file, out_vocab_file)
+
+        return (out_vocab_file,)

tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f868398fc4e05ee1e8aeba95ddf18ddcc45b8bce55d5093bead5bbf80429b48b
+size 1477754

tokenizer_config.json

@@ -0,0 +1,102 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "92538": {
+      "content": "<|plugin|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "92539": {
+      "content": "<|interpreter|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "92540": {
+      "content": "<|action_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "92541": {
+      "content": "<|action_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "92542": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "92543": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|action_start|>",
+    "<|action_end|>",
+    "<|interpreter|>",
+    "<|plugin|>"
+  ],
+  "auto_map": {
+    "AutoTokenizer": [
+      "tokenization_internlm2.InternLM2Tokenizer",
+      "tokenization_internlm2_fast.InternLM2TokenizerFast"
+    ]
+  },
+  "bos_token": "<s>",
+  "chat_template": "{{ bos_token }}{% for message in messages %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\n' }}{% endif %}",
+  "clean_up_tokenization_spaces": false,
+  "decode_with_prefix_space": false,
+  "eos_token": "</s>",
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "</s>",
+  "sp_model_kwargs": null,
+  "tokenizer_class": "InternLM2Tokenizer",
+  "unk_token": "<unk>"
+}