[CI/Build] Enable tests for recent day-0 new models (#34585)
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
This commit is contained in:
@@ -102,13 +102,13 @@ def glmasr_patch_mm_data(mm_data: MultiModalDataDict) -> MultiModalDataDict:
|
|||||||
# incorrect token ids. So we need use `add_special_tokens=False` here
|
# incorrect token ids. So we need use `add_special_tokens=False` here
|
||||||
# to leave bos_token to be added by the processor.
|
# to leave bos_token to be added by the processor.
|
||||||
_ADD_SPECIAL_TOKENS_OVERRIDES = {
|
_ADD_SPECIAL_TOKENS_OVERRIDES = {
|
||||||
|
"lfm2_vl": False,
|
||||||
"nemotron_parse": False,
|
"nemotron_parse": False,
|
||||||
"ovis": False,
|
"ovis": False,
|
||||||
"ovis2_5": False,
|
"ovis2_5": False,
|
||||||
"paligemma": False,
|
"paligemma": False,
|
||||||
"ultravox": False,
|
"ultravox": False,
|
||||||
"whisper": False,
|
"whisper": False,
|
||||||
"lfm2_vl": False,
|
|
||||||
}
|
}
|
||||||
|
|
||||||
_IGNORE_MM_KEYS = {
|
_IGNORE_MM_KEYS = {
|
||||||
@@ -450,6 +450,8 @@ def test_processing_correctness(
|
|||||||
num_batches: int,
|
num_batches: int,
|
||||||
simplify_rate: float,
|
simplify_rate: float,
|
||||||
):
|
):
|
||||||
|
if model_id == "allendou/Fun-ASR-Nano-2512-vllm":
|
||||||
|
pytest.skip("Cached audio `input_features` not matched. Fix later.")
|
||||||
if model_id == "google/gemma-3n-E2B-it":
|
if model_id == "google/gemma-3n-E2B-it":
|
||||||
pytest.skip("Fix later")
|
pytest.skip("Fix later")
|
||||||
if model_id == "OpenGVLab/InternVL2-2B":
|
if model_id == "OpenGVLab/InternVL2-2B":
|
||||||
@@ -468,9 +470,6 @@ def test_processing_correctness(
|
|||||||
"correctness test as is. Let's revisit adapting this "
|
"correctness test as is. Let's revisit adapting this "
|
||||||
"test once more realtime models exist."
|
"test once more realtime models exist."
|
||||||
)
|
)
|
||||||
if model_id == "internlm/Intern-S1-Pro":
|
|
||||||
# FIXME(Isotr0py): Fix later.
|
|
||||||
pytest.skip("Tokenization issue. Fix later")
|
|
||||||
|
|
||||||
_test_processing_correctness(
|
_test_processing_correctness(
|
||||||
model_id,
|
model_id,
|
||||||
|
|||||||
@@ -160,9 +160,6 @@ def test_model_tensor_schema(model_id: str):
|
|||||||
pytest.skip(
|
pytest.skip(
|
||||||
"Kimi-K2.5's offline inference has issues about vision chunks. Fix later."
|
"Kimi-K2.5's offline inference has issues about vision chunks. Fix later."
|
||||||
)
|
)
|
||||||
if model_id == "internlm/Intern-S1-Pro":
|
|
||||||
# FIXME(Isotr0py): Fix later.
|
|
||||||
pytest.skip("Intern-S1-Pro has issue to pass the test.")
|
|
||||||
|
|
||||||
model_info = HF_EXAMPLE_MODELS.find_hf_info(model_id)
|
model_info = HF_EXAMPLE_MODELS.find_hf_info(model_id)
|
||||||
model_info.check_available_online(on_fail="skip")
|
model_info.check_available_online(on_fail="skip")
|
||||||
|
|||||||
@@ -730,7 +730,6 @@ _MULTIMODAL_EXAMPLE_MODELS = {
|
|||||||
),
|
),
|
||||||
"FunASRForConditionalGeneration": _HfExamplesInfo(
|
"FunASRForConditionalGeneration": _HfExamplesInfo(
|
||||||
"allendou/Fun-ASR-Nano-2512-vllm",
|
"allendou/Fun-ASR-Nano-2512-vllm",
|
||||||
is_available_online=False,
|
|
||||||
),
|
),
|
||||||
"FunAudioChatForConditionalGeneration": _HfExamplesInfo(
|
"FunAudioChatForConditionalGeneration": _HfExamplesInfo(
|
||||||
"funaudiochat", is_available_online=False
|
"funaudiochat", is_available_online=False
|
||||||
@@ -755,7 +754,6 @@ _MULTIMODAL_EXAMPLE_MODELS = {
|
|||||||
"Glm4vMoeForConditionalGeneration": _HfExamplesInfo("zai-org/GLM-4.5V"),
|
"Glm4vMoeForConditionalGeneration": _HfExamplesInfo("zai-org/GLM-4.5V"),
|
||||||
"GlmOcrForConditionalGeneration": _HfExamplesInfo(
|
"GlmOcrForConditionalGeneration": _HfExamplesInfo(
|
||||||
"zai-org/GLM-OCR",
|
"zai-org/GLM-OCR",
|
||||||
is_available_online=False,
|
|
||||||
min_transformers_version="5.1.0",
|
min_transformers_version="5.1.0",
|
||||||
),
|
),
|
||||||
"H2OVLChatModel": _HfExamplesInfo(
|
"H2OVLChatModel": _HfExamplesInfo(
|
||||||
|
|||||||
@@ -85,11 +85,7 @@ class InternS1ProProcessingInfo(Qwen3VLProcessingInfo):
|
|||||||
return self.ctx.get_hf_config()
|
return self.ctx.get_hf_config()
|
||||||
|
|
||||||
def get_hf_processor(self, **kwargs: object) -> AutoProcessor:
|
def get_hf_processor(self, **kwargs: object) -> AutoProcessor:
|
||||||
return AutoProcessor.from_pretrained(
|
return self.ctx.get_hf_processor(**kwargs)
|
||||||
self.ctx.model_config.model,
|
|
||||||
trust_remote_code=True,
|
|
||||||
**kwargs,
|
|
||||||
)
|
|
||||||
|
|
||||||
|
|
||||||
class InternS1ProMoeMLP(nn.Module):
|
class InternS1ProMoeMLP(nn.Module):
|
||||||
@@ -497,7 +493,7 @@ class InternS1ProMoeLLMForCausalLM(Qwen3MoeForCausalLM):
|
|||||||
)
|
)
|
||||||
|
|
||||||
|
|
||||||
class Qwen3VLMoeMixtureOfExperts(MixtureOfExperts):
|
class InternS1ProMoeMixtureOfExperts(MixtureOfExperts):
|
||||||
def update_physical_experts_metadata(
|
def update_physical_experts_metadata(
|
||||||
self,
|
self,
|
||||||
num_physical_experts: int,
|
num_physical_experts: int,
|
||||||
@@ -547,7 +543,7 @@ class Qwen3VLMoeMixtureOfExperts(MixtureOfExperts):
|
|||||||
dummy_inputs=Qwen3VLDummyInputsBuilder,
|
dummy_inputs=Qwen3VLDummyInputsBuilder,
|
||||||
)
|
)
|
||||||
class InternS1ProForConditionalGeneration(
|
class InternS1ProForConditionalGeneration(
|
||||||
Qwen3VLForConditionalGeneration, Qwen3VLMoeMixtureOfExperts
|
Qwen3VLForConditionalGeneration, InternS1ProMoeMixtureOfExperts
|
||||||
):
|
):
|
||||||
is_3d_moe_weight: bool = True
|
is_3d_moe_weight: bool = True
|
||||||
packed_modules_mapping = {
|
packed_modules_mapping = {
|
||||||
|
|||||||
Reference in New Issue
Block a user