Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Support qwen2.5 #2054

Merged
merged 5 commits into from
Sep 18, 2024
Merged

Support qwen2.5 #2054

Changes from 1 commit
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
275b967
update
Jintao-Huang Sep 18, 2024
7c583df
Merge branch 'main' into support_qwen2_5
Jintao-Huang Sep 18, 2024
d680d40
update model.py
Jintao-Huang Sep 18, 2024
433562c
update
Jintao-Huang Sep 18, 2024
c213746
update
Jintao-Huang Sep 18, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
Next Next commit
update model.py
  • Loading branch information
@Jintao-Huang
Jintao-Huang committed Sep 18, 2024
commit d680d4048ca56382cbee9fd53296f7c4156dbb86
135 changes: 55 additions & 80 deletions swift/llm/utils/model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2625,76 +2625,6 @@ def get_model_tokenizer_glm4v(model_dir: str,
support_vllm=True,
tags=['moe'],
hf_model_id='databricks/dbrx-instruct')
@register_model(
ModelType.qwen2_5_0_5b,
'qwen/Qwen2.5-0.5B',
LoRATM.llama,
TemplateType.default_generation,
support_flash_attn=True,
support_vllm=True,
support_lmdeploy=True,
requires=['transformers>=4.37'],
hf_model_id='Qwen/Qwen2.5-0.5B')
@register_model(
ModelType.qwen2_5_1_5b,
'qwen/Qwen2.5-1.5B',
LoRATM.llama,
TemplateType.default_generation,
support_flash_attn=True,
support_vllm=True,
support_lmdeploy=True,
requires=['transformers>=4.37'],
hf_model_id='Qwen/Qwen2.5-1.5B')
@register_model(
ModelType.qwen2_5_3b,
'qwen/Qwen2.5-3B',
LoRATM.llama,
TemplateType.default_generation,
support_flash_attn=True,
support_vllm=True,
support_lmdeploy=True,
requires=['transformers>=4.37'],
hf_model_id='Qwen/Qwen2.5-3B')
@register_model(
ModelType.qwen2_5_7b,
'qwen/Qwen2.5-7B',
LoRATM.llama,
TemplateType.default_generation,
support_flash_attn=True,
support_vllm=True,
support_lmdeploy=True,
requires=['transformers>=4.37'],
hf_model_id='Qwen/Qwen2.5-7B')
@register_model(
ModelType.qwen2_5_14b,
'qwen/Qwen2.5-14B',
LoRATM.llama,
TemplateType.default_generation,
support_flash_attn=True,
support_vllm=True,
support_lmdeploy=True,
requires=['transformers>=4.37'],
hf_model_id='Qwen/Qwen2.5-14B')
@register_model(
ModelType.qwen2_5_32b,
'qwen/Qwen2.5-32B',
LoRATM.llama,
TemplateType.default_generation,
support_flash_attn=True,
support_vllm=True,
support_lmdeploy=True,
requires=['transformers>=4.37'],
hf_model_id='Qwen/Qwen2.5-32B')
@register_model(
ModelType.qwen2_5_72b,
'qwen/Qwen2.5-72B',
LoRATM.llama,
TemplateType.default_generation,
support_flash_attn=True,
support_vllm=True,
support_lmdeploy=True,
requires=['transformers>=4.37'],
hf_model_id='Qwen/Qwen2.5-72B')
def get_model_tokenizer_with_flash_attn(model_dir: str,
torch_dtype: torch.dtype,
model_kwargs: Dict[str, Any],
Expand Down Expand Up @@ -3474,16 +3404,6 @@ def rotary_emb(self, query_states, key_states, **kwargs):
support_lmdeploy=True,
requires=['transformers>=4.37'],
hf_model_id='Qwen/CodeQwen1.5-7B-Chat')
@register_model(
ModelType.qwen2_5_0_5b_instruct,
'qwen/Qwen2.5-0.5B-Instruct',
LoRATM.llama,
TemplateType.qwen,
support_flash_attn=True,
support_vllm=True,
support_lmdeploy=True,
requires=['transformers>=4.37'],
hf_model_id='Qwen/Qwen2-0.5B-Instruct')
def get_model_tokenizer_qwen2_chat(model_dir: str,
torch_dtype: torch.dtype,
model_kwargs: Dict[str, Any],
Expand All @@ -3493,6 +3413,61 @@ def get_model_tokenizer_qwen2_chat(model_dir: str,
return get_model_tokenizer_with_flash_attn(model_dir, torch_dtype, model_kwargs, load_model, **kwargs)


for model_size in ['0.5B', '1.5B', '3B', '7B', '14B', '32B', '72B']:
model_size_lower = model_size.lower().replace('.', '_')
register_model(
f'qwen2_5-{model_size_lower}-instruct',
f'qwen/Qwen2.5-{model_size}-Instruct',
LoRATM.llama,
TemplateType.qwen,
get_model_tokenizer_qwen2_chat,
support_flash_attn=True,
support_vllm=True,
support_lmdeploy=True,
requires=['transformers>=4.37'],
hf_model_id=f'Qwen/Qwen2.5-{model_size}-Instruct')
register_model(
f'qwen2_5-{model_size_lower}',
f'qwen/Qwen2.5-{model_size}',
LoRATM.llama,
TemplateType.default_generation,
get_model_tokenizer_with_flash_attn,
support_flash_attn=True,
support_vllm=True,
support_lmdeploy=True,
requires=['transformers>=4.37'],
hf_model_id=f'Qwen/Qwen2.5-{model_size}')

for quant_bits in [4, 8]:
quant_type = f'GPTQ-Int{quant_bits}'
quant_type_lower = quant_type.lower()
register_model(
f'qwen2_5-{model_size_lower}-{quant_type_lower}',
f'qwen/Qwen2.5-{model_size}-Instruct-{quant_type}',
LoRATM.llama,
TemplateType.qwen,
get_model_tokenizer_qwen2_chat,
support_flash_attn=True,
support_vllm=True,
function_kwargs={'gptq_bits': quant_bits},
torch_dtype=torch.float16,
requires=['auto_gptq>=0.5', 'transformers>=4.37'],
hf_model_id=f'Qwen/Qwen2.5-{model_size}-Instruct-{quant_type}')

register_model(
f'qwen2_5-{model_size_lower}-awq',
f'qwen/Qwen2.5-{model_size}-Instruct-AWQ',
LoRATM.llama,
TemplateType.qwen,
get_model_tokenizer_qwen2_chat,
support_flash_attn=True,
support_vllm=True,
function_kwargs={'is_awq': True},
torch_dtype=torch.float16,
requires=['transformers>=4.37', 'autoawq'],
hf_model_id=f'Qwen/Qwen2.5-{model_size}-Instruct-AWQ')


@register_model(
ModelType.qwen2_audio_7b_instruct,
'qwen/Qwen2-Audio-7B-Instruct',
Expand Down
Loading