[Model] Gemma3n MM (#20495)

Signed-off-by: ShriKode <shrikode@gmail.com>
Signed-off-by: NickLucche <nlucches@redhat.com>
Signed-off-by: Roger Wang <hey@rogerw.me>
Co-authored-by: ShriKode <shrikode@gmail.com>
Co-authored-by: Roger Wang <hey@rogerw.me>
This commit is contained in:
Nicolò Lucchesi
2025-08-09 09:56:25 -07:00
committed by GitHub
co-authored by ShriKode Roger Wang
parent 2d18256e47
commit 5a16fa614c
11 changed files with 864 additions and 55 deletions
@@ -211,7 +211,33 @@ def run_gemma3(questions: list[str], modality: str) -> ModelRequestData:
)
for question in questions
]
return ModelRequestData(
engine_args=engine_args,
prompts=prompts,
)
# Gemma3N
def run_gemma3n(questions: list[str], modality: str) -> ModelRequestData:
assert modality == "image"
model_name = "google/gemma-3n-E2B-it"
engine_args = EngineArgs(
model=model_name,
max_model_len=2048,
max_num_seqs=2,
limit_mm_per_prompt={modality: 1},
enforce_eager=True,
)
prompts = [
(
"<start_of_turn>user\n"
f"<image_soft_token>{question}<end_of_turn>\n"
"<start_of_turn>model\n"
)
for question in questions
]
return ModelRequestData(
engine_args=engine_args,
prompts=prompts,
@@ -1395,6 +1421,7 @@ model_example_map = {
"florence2": run_florence2,
"fuyu": run_fuyu,
"gemma3": run_gemma3,
"gemma3n": run_gemma3n,
"glm4v": run_glm4v,
"glm4_1v": run_glm4_1v,
"h2ovl_chat": run_h2ovl,