Skip to content

support qwen3 on nvidia #3302

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

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
Open

support qwen3 on nvidia #3302

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
bb61a23
support qwen3 on nvidia
icyxp Jul 23, 2025
5d44bdd
Update __init__.py
icyxp Jul 29, 2025
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
42 changes: 42 additions & 0 deletions server/text_generation_server/models/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -152,6 +152,9 @@
from text_generation_server.models.custom_modeling.flash_qwen2_modeling import (
Qwen2ForCausalLM,
)
from text_generation_server.models.custom_modeling.flash_qwen3_modeling import (
Qwen3ForCausalLM,
)
from text_generation_server.models.custom_modeling.flash_mistral_modeling import (
FlashMistralForCausalLM,
)
Expand Down Expand Up @@ -348,6 +351,11 @@ class ModelType(enum.Enum):
"name": "Qwen 2",
"url": "https://huggingface.co/collections/Qwen/qwen2-6659360b33528ced941e557f",
}
QWEN3 = {
"type": "qwen3",
"name": "Qwen 3",
"url": "https://huggingface.co/collections/Qwen/qwen3-67dd247413f0e2e4f653967f",
}
QWEN2_VL = {
"type": "qwen2_vl",
"name": "Qwen 2 VL",
Expand Down Expand Up @@ -1470,6 +1478,40 @@ def get_model(
trust_remote_code=trust_remote_code,
)

if model_type == QWEN3:
if FLASH_ATTENTION:
return FlashCausalLM(
model_id=model_id,
model_class=Qwen3ForCausalLM,
revision=revision,
quantize=quantize,
speculator=speculator,
dtype=dtype,
kv_cache_dtype=kv_cache_dtype,
trust_remote_code=trust_remote_code,
lora_adapter_ids=lora_adapter_ids,
)
elif FLASH_TRANSFORMERS_BACKEND:
return TransformersFlashCausalLM.fallback(
model_id,
revision,
quantize=quantize,
speculator=speculator,
dtype=dtype,
trust_remote_code=trust_remote_code,
)
elif sharded:
raise NotImplementedError(FLASH_ATT_ERROR_MESSAGE.format("Sharded Qwen3"))
else:
return CausalLM.fallback(
model_id,
revision,
quantize=quantize,
speculator=speculator,
dtype=dtype,
trust_remote_code=trust_remote_code,
)

if model_type == OPT:
return CausalLM(
model_id=model_id,
Expand Down
Loading