support more models

README.md

@@ -45,6 +45,7 @@ Additionally, we are expanding capabilities for other modalities. Currently, we
 SWIFT has rich documentations for users, please check [here](https://github.com/modelscope/swift/tree/main/docs/source_en/LLM).
 
 ## 🎉 News
+- 2024.05.20: Supports TeleChat-12B-v2 model, model_type are `telechat-12b-v2`
 - 🔥2024.05.20: Support for inferencing and fine-tuning cogvlm2-llama3-chinese-chat-19B, cogvlm2-llama3-chat-19B. you can refer to [cogvlm2 Best Practices](docs/source_en/Multi-Modal/cogvlm2-best-practice.md).
 - 🔥2024.05.17: Support peft=0.11.0. Meanwhile support 3 new tuners: `BOFT`, `Vera` and `Pissa`. use `--sft_type boft/vera` to use BOFT or Vera, use `--init_lora_weights pissa` with `--sft_type lora` to use Pissa.
 - 2024.05.16: Supports Llava-Next (Stronger) series models. For best practice, you can refer to [here](https://github.com/modelscope/swift/tree/main/docs/source_en/Multi-Modal/llava-best-practice.md).

README_CN.md

@@ -46,6 +46,7 @@ SWIFT支持近**200种LLM和MLLM**（多模态大模型）的训练、推理、
 SWIFT具有丰富的文档体系，如有使用问题请请查看[这里](https://github.com/modelscope/swift/tree/main/docs/source/LLM).
 
 ## 🎉 新闻
+- 2024.05.21: 支持TeleChat-12b-v2模型, model_type为 `telechat-12b-v2`
 - 🔥2024.05.20: 支持 cogvlm2-llama3-chinese-chat-19B, cogvlm2-llama3-chat-19B 的推理与微调, 可以查看[cogvlm2最佳实践](docs/source/Multi-Modal/cogvlm2最佳实践.md).
 - 🔥2024.05.17: 支持peft=0.11.0. 同时支持了三个新的tuner方法： `BOFT`, `Vera` 和 `Pissa`. 使用 `--sft_type boft/vera` 开启BOFT或者Vera, 使用 `--init_lora_weights pissa` 以及 `--sft_type lora` 来使用 Pissa.
 - 2024.05.16: 支持Llava-Next (Stronger)系列模型，最佳实践可以查看[这里](https://github.com/modelscope/swift/tree/main/docs/source/Multi-Modal/llava最佳实践.md).

swift/llm/utils/model.py

@@ -244,6 +244,8 @@ class ModelType:
     deepseek_vl_7b_chat = 'deepseek-vl-7b-chat'
     # deepseek-v2
     deepseek_v2_chat = 'deepseek-v2-chat'
+    deepseek_v2_lite = 'deepseek-v2-lite'
+    deepseek_v2_lite_chat = 'deepseek-v2-lite-chat'
     # gemma
     gemma_2b = 'gemma-2b'
     gemma_7b = 'gemma-7b'
@@ -357,6 +359,8 @@ class ModelType:
     # teleAI
     telechat_7b = 'telechat-7b'
     telechat_12b = 'telechat-12b'
+    telechat_12b_v2 = 'telechat-12b-v2'
+    telechat_12b_v2_gptq_int4 = 'telechat-12b-v2-gptq-int4'
     # grok-1
     grok_1 = 'grok-1'
     # dbrx
@@ -2580,7 +2584,26 @@ def get_model_tokenizer_internlm2(model_dir: str,
 
     return model, tokenizer
 
-
+@register_model(
+    ModelType.deepseek_v2_lite,
+    'deepseek-ai/DeepSeek-V2-Lite',
+    LoRATM.deepseek2,
+    TemplateType.default_generation,
+    support_gradient_checkpointing=False,
+    support_flash_attn=True,
+    support_vllm=True,
+    requires=['transformers>=4.39.3'],
+    hf_model_id='deepseek-ai/DeepSeek-V2-Lite')
+@register_model(
+    ModelType.deepseek_v2_lite_chat,
+    'deepseek-ai/DeepSeek-V2-Lite-Chat',
+    LoRATM.deepseek2,
+    TemplateType.deepseek2,
+    support_gradient_checkpointing=False,
+    support_flash_attn=True,
+    support_vllm=True,
+    requires=['transformers>=4.39.3'],
+    hf_model_id='deepseek-ai/DeepSeek-V2-Lite-Chat')
 @register_model(
     ModelType.deepseek_v2_chat,
     'deepseek-ai/DeepSeek-V2-Chat',
@@ -3643,6 +3666,24 @@ def get_model_tokenizer_codellama(model_dir: str,
     TemplateType.telechat,
     support_flash_attn=True,
     hf_model_id='Tele-AI/TeleChat-12B')
+@register_model(
+    ModelType.telechat_12b_v2,
+    'TeleAI/TeleChat-12B-v2',
+    LoRATM.telechat,
+    TemplateType.telechat_v2,
+    eos_token=2,
+    support_flash_attn=True,
+    hf_model_id='Tele-AI/TeleChat-12B-v2')
+@register_model(
+    ModelType.telechat_12b_v2_gptq_int4,
+    'swift/TeleChat-12B-V2-GPTQ-Int4',
+    LoRATM.telechat,
+    TemplateType.telechat_v2,
+    eos_token=2,
+    requires=['auto_gptq>=0.5'],
+    support_flash_attn=True,
+    function_kwargs={'gptq_bits': 4},
+    hf_model_id='Tele-AI/TeleChat-12B-v2')
 def get_model_tokenizer_phi(model_dir: str,
                             torch_dtype: Dtype,
                             model_kwargs: Dict[str, Any],
@@ -3651,9 +3692,11 @@ def get_model_tokenizer_phi(model_dir: str,
     model_config = AutoConfig.from_pretrained(model_dir, trust_remote_code=True)
     use_flash_attn = kwargs.pop('use_flash_attn', False)
     model_config.flash_attn = use_flash_attn
-    return get_model_tokenizer_from_repo(
+    model, tokenizer = get_model_tokenizer_from_repo(
         model_dir, torch_dtype, model_kwargs, load_model, model_config=model_config, **kwargs)
 
+    return model, tokenizer
+
 
 @register_model(
     ModelType.telechat_7b,

swift/llm/utils/template.py

@@ -75,6 +75,7 @@ class TemplateType:
     atom = 'atom'
     phi3 = 'phi3'
     telechat = 'telechat'
+    telechat_v2 = 'telechat-v2'
     dbrx = 'dbrx'
     mengzi = 'mengzi'
     c4ai = 'c4ai'
@@ -1374,6 +1375,8 @@ def get_generate_ids(generate_ids: Tensor, input_token_len: int) -> List[int]:
 
 register_template(TemplateType.telechat, Template([], ['<_user>{{QUERY}}<_bot>'], ['<_end>'], ['<_end>']))
 
+register_template(TemplateType.telechat_v2, Template([], ['<_user> {{QUERY}}<_bot>'], [], ['<_end>']))
+
 DBRX_SYSTEM = (
     'You are DBRX, created by Databricks. You were last updated in December 2023. '
     'You answer questions based on information available up to that point.\n'