Bump Transformers version to 5.13.0 (#47867)
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
@@ -962,7 +962,7 @@ s3transfer==0.10.3
|
|||||||
# via boto3
|
# via boto3
|
||||||
sacrebleu==2.4.3
|
sacrebleu==2.4.3
|
||||||
# via lm-eval
|
# via lm-eval
|
||||||
safetensors==0.7.0
|
safetensors==0.8.0
|
||||||
# via
|
# via
|
||||||
# -r requirements/test/../common.txt
|
# -r requirements/test/../common.txt
|
||||||
# accelerate
|
# accelerate
|
||||||
@@ -1159,7 +1159,7 @@ tqdm==4.67.3
|
|||||||
# segmentation-models-pytorch
|
# segmentation-models-pytorch
|
||||||
# sentence-transformers
|
# sentence-transformers
|
||||||
# transformers
|
# transformers
|
||||||
transformers==5.10.4
|
transformers @ git+https://github.com/huggingface/transformers.git@bef013c4a2180aeef83355bd200a7383ce1ac7e3
|
||||||
# via
|
# via
|
||||||
# -r requirements/test/../common.txt
|
# -r requirements/test/../common.txt
|
||||||
# -r requirements/test/cuda.in
|
# -r requirements/test/cuda.in
|
||||||
|
|||||||
@@ -39,7 +39,7 @@ open_clip_torch==2.32.0 # Required for nemotron_vl test, Nemotron Parse in test_
|
|||||||
datamodel_code_generator # required for minicpm3 test
|
datamodel_code_generator # required for minicpm3 test
|
||||||
lm-eval[api]>=0.4.12 # required for model evaluation test
|
lm-eval[api]>=0.4.12 # required for model evaluation test
|
||||||
mteb[bm25s]>=2, <3 # required for mteb test
|
mteb[bm25s]>=2, <3 # required for mteb test
|
||||||
transformers==5.10.4
|
transformers @ git+https://github.com/huggingface/transformers.git@v5.13-release
|
||||||
tokenizers==0.22.2
|
tokenizers==0.22.2
|
||||||
schemathesis>=4.0.0 # Required for openai schema test.
|
schemathesis>=4.0.0 # Required for openai schema test.
|
||||||
# quantization
|
# quantization
|
||||||
|
|||||||
@@ -1053,7 +1053,7 @@ s3transfer==0.10.3
|
|||||||
# via boto3
|
# via boto3
|
||||||
sacrebleu==2.4.3
|
sacrebleu==2.4.3
|
||||||
# via lm-eval
|
# via lm-eval
|
||||||
safetensors==0.7.0
|
safetensors==0.8.0
|
||||||
# via
|
# via
|
||||||
# -c requirements/common.txt
|
# -c requirements/common.txt
|
||||||
# -r requirements/test/../common.txt
|
# -r requirements/test/../common.txt
|
||||||
@@ -1261,7 +1261,7 @@ tqdm==4.67.3
|
|||||||
# segmentation-models-pytorch
|
# segmentation-models-pytorch
|
||||||
# sentence-transformers
|
# sentence-transformers
|
||||||
# transformers
|
# transformers
|
||||||
transformers==5.10.4
|
transformers @ git+https://github.com/huggingface/transformers.git@bef013c4a2180aeef83355bd200a7383ce1ac7e3
|
||||||
# via
|
# via
|
||||||
# -c requirements/common.txt
|
# -c requirements/common.txt
|
||||||
# -r requirements/test/../common.txt
|
# -r requirements/test/../common.txt
|
||||||
|
|||||||
@@ -29,7 +29,7 @@ opencv-python-headless >= 4.13.0 # required for video test
|
|||||||
datamodel_code_generator # required for minicpm3 test
|
datamodel_code_generator # required for minicpm3 test
|
||||||
lm-eval[api]>=0.4.12 # required for model evaluation test
|
lm-eval[api]>=0.4.12 # required for model evaluation test
|
||||||
mteb[bm25s]>=2, <3 # required for mteb test
|
mteb[bm25s]>=2, <3 # required for mteb test
|
||||||
transformers==5.10.4
|
transformers @ git+https://github.com/huggingface/transformers.git@v5.13-release
|
||||||
tokenizers==0.22.2
|
tokenizers==0.22.2
|
||||||
schemathesis>=4.0.0 # Required for openai schema test.
|
schemathesis>=4.0.0 # Required for openai schema test.
|
||||||
# quantization
|
# quantization
|
||||||
|
|||||||
@@ -35,7 +35,7 @@ open_clip_torch==2.32.0 # Required for nemotron_vl test, Nemotron Parse in test_
|
|||||||
datamodel_code_generator # required for minicpm3 test
|
datamodel_code_generator # required for minicpm3 test
|
||||||
lm-eval[api]>=0.4.12 # required for model evaluation test
|
lm-eval[api]>=0.4.12 # required for model evaluation test
|
||||||
mteb[bm25s]>=2, <3 # required for mteb test
|
mteb[bm25s]>=2, <3 # required for mteb test
|
||||||
transformers==5.10.4
|
transformers @ git+https://github.com/huggingface/transformers.git@v5.13-release
|
||||||
tokenizers==0.22.2
|
tokenizers==0.22.2
|
||||||
schemathesis>=4.0.0 # Required for openai schema test
|
schemathesis>=4.0.0 # Required for openai schema test
|
||||||
# quantization
|
# quantization
|
||||||
|
|||||||
@@ -1035,7 +1035,7 @@ s3transfer==0.16.0
|
|||||||
# via boto3
|
# via boto3
|
||||||
sacrebleu==2.6.0
|
sacrebleu==2.6.0
|
||||||
# via lm-eval
|
# via lm-eval
|
||||||
safetensors==0.7.0
|
safetensors==0.8.0
|
||||||
# via
|
# via
|
||||||
# -c requirements/common.txt
|
# -c requirements/common.txt
|
||||||
# -r requirements/test/../common.txt
|
# -r requirements/test/../common.txt
|
||||||
@@ -1218,7 +1218,7 @@ tqdm==4.67.3
|
|||||||
# sentence-transformers
|
# sentence-transformers
|
||||||
# tilelang
|
# tilelang
|
||||||
# transformers
|
# transformers
|
||||||
transformers==5.10.4
|
transformers @ git+https://github.com/huggingface/transformers.git@bef013c4a2180aeef83355bd200a7383ce1ac7e3
|
||||||
# via
|
# via
|
||||||
# -c requirements/common.txt
|
# -c requirements/common.txt
|
||||||
# -r requirements/test/../common.txt
|
# -r requirements/test/../common.txt
|
||||||
|
|||||||
@@ -17,7 +17,7 @@ accelerate
|
|||||||
arctic-inference
|
arctic-inference
|
||||||
lm_eval[api]>=0.4.12
|
lm_eval[api]>=0.4.12
|
||||||
modelscope<1.38
|
modelscope<1.38
|
||||||
transformers==5.10.4
|
transformers @ git+https://github.com/huggingface/transformers.git@v5.13-release
|
||||||
|
|
||||||
# --- Audio Processing ---
|
# --- Audio Processing ---
|
||||||
librosa
|
librosa
|
||||||
|
|||||||
@@ -779,7 +779,7 @@ rpds-py==0.30.0
|
|||||||
# referencing
|
# referencing
|
||||||
sacrebleu==2.6.0
|
sacrebleu==2.6.0
|
||||||
# via lm-eval
|
# via lm-eval
|
||||||
safetensors==0.7.0
|
safetensors==0.8.0
|
||||||
# via
|
# via
|
||||||
# -c requirements/common.txt
|
# -c requirements/common.txt
|
||||||
# -r requirements/test/../common.txt
|
# -r requirements/test/../common.txt
|
||||||
@@ -940,7 +940,7 @@ tqdm==4.67.3
|
|||||||
# pqdm
|
# pqdm
|
||||||
# sentence-transformers
|
# sentence-transformers
|
||||||
# transformers
|
# transformers
|
||||||
transformers==5.10.4
|
transformers @ git+https://github.com/huggingface/transformers.git@bef013c4a2180aeef83355bd200a7383ce1ac7e3
|
||||||
# via
|
# via
|
||||||
# -c requirements/common.txt
|
# -c requirements/common.txt
|
||||||
# -r requirements/test/../common.txt
|
# -r requirements/test/../common.txt
|
||||||
|
|||||||
@@ -316,8 +316,12 @@ class GraniteMoeHybridAttention(nn.Module):
|
|||||||
|
|
||||||
|
|
||||||
ALL_DECODER_LAYER_TYPES = {
|
ALL_DECODER_LAYER_TYPES = {
|
||||||
|
# Transformers < 5.13.0
|
||||||
"attention": GraniteMoeHybridAttentionDecoderLayer,
|
"attention": GraniteMoeHybridAttentionDecoderLayer,
|
||||||
"mamba": GraniteMoeHybridMambaDecoderLayer,
|
"mamba": GraniteMoeHybridMambaDecoderLayer,
|
||||||
|
# Transformers >= 5.13.0
|
||||||
|
"full_attention": GraniteMoeHybridAttentionDecoderLayer,
|
||||||
|
"linear_attention": GraniteMoeHybridMambaDecoderLayer,
|
||||||
}
|
}
|
||||||
|
|
||||||
|
|
||||||
|
|||||||
@@ -73,6 +73,7 @@ from vllm.multimodal.processing import (
|
|||||||
BaseProcessingInfo,
|
BaseProcessingInfo,
|
||||||
PromptReplacement,
|
PromptReplacement,
|
||||||
PromptUpdate,
|
PromptUpdate,
|
||||||
|
PromptUpdateDetails,
|
||||||
)
|
)
|
||||||
from vllm.sequence import IntermediateTensors
|
from vllm.sequence import IntermediateTensors
|
||||||
from vllm.transformers_utils.configs.hunyuan_vl import (
|
from vllm.transformers_utils.configs.hunyuan_vl import (
|
||||||
@@ -751,8 +752,10 @@ class HunYuanVLMultiModalProcessor(BaseMultiModalProcessor[HunYuanVLProcessingIn
|
|||||||
hf_processor = self.info.get_hf_processor(**hf_processor_mm_kwargs)
|
hf_processor = self.info.get_hf_processor(**hf_processor_mm_kwargs)
|
||||||
image_processor = self.info.get_image_processor(**hf_processor_mm_kwargs)
|
image_processor = self.info.get_image_processor(**hf_processor_mm_kwargs)
|
||||||
|
|
||||||
placeholder = {
|
token_ids = {
|
||||||
"image": hf_processor.image_token_id,
|
"image": hf_processor.image_token_id,
|
||||||
|
"image_start": hf_processor.image_start_token_id,
|
||||||
|
"image_end": hf_processor.image_end_token_id,
|
||||||
}
|
}
|
||||||
|
|
||||||
merge_size = image_processor.merge_size
|
merge_size = image_processor.merge_size
|
||||||
@@ -766,12 +769,17 @@ class HunYuanVLMultiModalProcessor(BaseMultiModalProcessor[HunYuanVLProcessingIn
|
|||||||
num_tokens = (int(grid_h) // merge_size) * (
|
num_tokens = (int(grid_h) // merge_size) * (
|
||||||
int(grid_w) // merge_size + 1
|
int(grid_w) // merge_size + 1
|
||||||
) + 2
|
) + 2
|
||||||
return [placeholder[modality]] * num_tokens
|
tokens = (
|
||||||
|
[token_ids[f"{modality}_start"]]
|
||||||
|
+ [token_ids[modality]] * num_tokens
|
||||||
|
+ [token_ids[f"{modality}_end"]]
|
||||||
|
)
|
||||||
|
return PromptUpdateDetails.select_token_id(tokens, token_ids[modality])
|
||||||
|
|
||||||
return [
|
return [
|
||||||
PromptReplacement(
|
PromptReplacement(
|
||||||
modality=modality,
|
modality=modality,
|
||||||
target=[placeholder[modality]],
|
target=[token_ids[modality]],
|
||||||
replacement=partial(get_replacement_hunyuan_vl, modality=modality),
|
replacement=partial(get_replacement_hunyuan_vl, modality=modality),
|
||||||
)
|
)
|
||||||
for modality in ("image",)
|
for modality in ("image",)
|
||||||
|
|||||||
@@ -137,11 +137,9 @@ class Olmo3Attention(nn.Module):
|
|||||||
)
|
)
|
||||||
|
|
||||||
# Rotary embeddings. Rope scaling is only applied on full attention layers.
|
# Rotary embeddings. Rope scaling is only applied on full attention layers.
|
||||||
if sliding_window is None:
|
rope_parameters = self.config.rope_parameters
|
||||||
rope_parameters = self.config.rope_parameters
|
attn_type = "full_attention" if sliding_window is None else "sliding_attention"
|
||||||
else:
|
rope_parameters = rope_parameters.get(attn_type, rope_parameters)
|
||||||
rope_theta = self.config.rope_parameters["rope_theta"]
|
|
||||||
rope_parameters = {"rope_type": "default", "rope_theta": rope_theta}
|
|
||||||
self.rotary_emb = get_rope(
|
self.rotary_emb = get_rope(
|
||||||
self.head_dim,
|
self.head_dim,
|
||||||
max_position=self.max_position_embeddings,
|
max_position=self.max_position_embeddings,
|
||||||
|
|||||||
@@ -21,8 +21,6 @@ __all__ = [
|
|||||||
"GLM4VProcessor",
|
"GLM4VProcessor",
|
||||||
"Granite4VisionProcessor",
|
"Granite4VisionProcessor",
|
||||||
"H2OVLProcessor",
|
"H2OVLProcessor",
|
||||||
"HunYuanVLProcessor",
|
|
||||||
"HunYuanVLImageProcessor",
|
|
||||||
"Moondream3Processor",
|
"Moondream3Processor",
|
||||||
"InternVLProcessor",
|
"InternVLProcessor",
|
||||||
"IsaacProcessor",
|
"IsaacProcessor",
|
||||||
@@ -58,8 +56,6 @@ _CLASS_TO_MODULE: dict[str, str] = {
|
|||||||
"GLM4VProcessor": "vllm.transformers_utils.processors.glm4v",
|
"GLM4VProcessor": "vllm.transformers_utils.processors.glm4v",
|
||||||
"Granite4VisionProcessor": "vllm.transformers_utils.processors.granite4_vision",
|
"Granite4VisionProcessor": "vllm.transformers_utils.processors.granite4_vision",
|
||||||
"H2OVLProcessor": "vllm.transformers_utils.processors.h2ovl",
|
"H2OVLProcessor": "vllm.transformers_utils.processors.h2ovl",
|
||||||
"HunYuanVLProcessor": "vllm.transformers_utils.processors.hunyuan_vl",
|
|
||||||
"HunYuanVLImageProcessor": "vllm.transformers_utils.processors.hunyuan_vl_image",
|
|
||||||
"InternVLProcessor": "vllm.transformers_utils.processors.internvl",
|
"InternVLProcessor": "vllm.transformers_utils.processors.internvl",
|
||||||
"IsaacProcessor": "vllm.transformers_utils.processors.isaac",
|
"IsaacProcessor": "vllm.transformers_utils.processors.isaac",
|
||||||
"KimiAudioProcessor": "vllm.transformers_utils.processors.kimi_audio",
|
"KimiAudioProcessor": "vllm.transformers_utils.processors.kimi_audio",
|
||||||
|
|||||||
Reference in New Issue
Block a user