Skip to content

Commit 8d33e42

Browse files
anzr299AlexanderDokuchaev
anzr299
and
AlexanderDokuchaev
authored
[Torch] Bump PyTorch Version to 2.6.0 (openvinotoolkit#3253)
### Changes 1. Bump Torch version up to 2.6.0 and torchvision to 0.21.0 2. use weights_only=False for all model loading in pytorch 3. modify multi_head_attention_forward code in pt2 function hook 4. xfail torch fx backend tests in conformance test issue-162009 5. do not export pytorch backend models before ov.convert in conformance test. 6. transformers version update from 4.38.2 -> 4.48.3, optimum 1.23.3 -> 1.24.0, optimum-intel 1.21.0 -> 1.22.0, accelerate 0.28.0 -> 1.1.0 7. make attn_implementation=eager to ensure old transformers SDPA attention is used instead of the new torch SDPA API 8. add xfail for swin model in test_model_saving due to issue-162383 <html xmlns:v="urn:schemas-microsoft-com:vml" xmlns:o="urn:schemas-microsoft-com:office:office" xmlns:x="urn:schemas-microsoft-com:office:excel" xmlns="http://www.w3.org/TR/REC-html40"> Category | Job | Status | Job Number | Notes -- | -- | -- | -- | -- Nightly | install_pt_cpu | Pass | 637 |   Nightly | install_pt_gpu | Pass | 638 |   Nightly | torch_nightly | Pass | 463 |   Nightly | e2e_pytorch_full | Pass | 962 |   Nightly/Windows | install_torch_cpu | Pass | 526 |   Nightly/Windows | install_torch_gpu | - | - | No Build info Manual | post_training_quantization | Pass | 623 |   Manual | post_training_weights_compression | Pass | 326 |   GH WC | WC test on GHA | Pass | <a href="https://github.com/openvinotoolkit/nncf/actions/runs/13456801808">33</a> | Example | Example | Pass | <a href="https://github.com/openvinotoolkit/nncf/actions/runs/13456804856">238</a> |   </body> </html> --------- Co-authored-by: Alexander Dokuchaev <alexander.dokuchaev@intel.com>
1 parent f4c9693 commit 8d33e42

File tree

103 files changed

+244
-181
lines changed

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

103 files changed

+244
-181
lines changed

‎constraints.txt

+2-2
Original file line numberDiff line numberDiff line change
@@ -2,8 +2,8 @@
22
openvino==2025.0.0
33

44
# Pytorch
5-
torch==2.5.1
6-
torchvision==0.20.1
5+
torch==2.6.0
6+
torchvision==0.21.0
77

88
# ONNX
99
onnx==1.17.0

‎docs/Installation.md

+1-1
Original file line numberDiff line numberDiff line change
@@ -49,7 +49,7 @@ as well as the supported versions of Python:
4949

5050
| NNCF | OpenVINO | PyTorch | ONNX | TensorFlow | Python |
5151
|-----------|------------|----------|----------|------------|--------|
52-
| `develop` | `2025.0.0` | `2.5.1` | `1.17.0` | `2.15.1` | `3.10` |
52+
| `develop` | `2025.0.0` | `2.6.0` | `1.17.0` | `2.15.1` | `3.10` |
5353
| `2.15.0` | `2025.0.0` | `2.5.1` | `1.17.0` | `2.15.1` | `3.10` |
5454
| `2.14.1` | `2024.6.0` | `2.5.1` | `1.17.0` | `2.15.1` | `3.10` |
5555
| `2.14.0` | `2024.5.0` | `2.5.1` | `1.17.0` | `2.15.1` | `3.10` |

‎examples/experimental/torch/classification/bootstrap_nas_search.py

+2-2
Original file line numberDiff line numberDiff line change
@@ -140,9 +140,9 @@ def validate_model_fn_top1(model_, loader_):
140140
nncf_network = create_nncf_network(model, nncf_config)
141141

142142
if config.search_mode_active:
143-
compression_state = torch.load(config.search_elasticity_state_path)
143+
compression_state = torch.load(config.search_elasticity_state_path, weights_only=False)
144144
model, elasticity_ctrl = resume_compression_from_state(nncf_network, compression_state)
145-
model_weights = torch.load(config.search_supernet_weights)
145+
model_weights = torch.load(config.search_supernet_weights, weights_only=False)
146146

147147
load_state(model, model_weights, is_resume=True)
148148

‎examples/llm_compression/openvino/tiny_llama_synthetic_data/requirements.txt

+1-1
Original file line numberDiff line numberDiff line change
@@ -1,4 +1,4 @@
1-
torch==2.5.1
1+
torch==2.6.0
22
datasets==3.0.1
33
numpy>=1.23.5,<2
44
openvino==2025.0

‎examples/post_training_quantization/torch/mobilenet_v2/requirements.txt

+2-2
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,6 @@
11
fastdownload==0.0.7
22
openvino==2025.0
33
scikit-learn
4-
torch==2.5.1
5-
torchvision==0.20.1
4+
torch==2.6.0
5+
torchvision==0.21.0
66
setuptools<=72.1.0

‎examples/post_training_quantization/torch/ssd300_vgg16/requirements.txt

+2-2
Original file line numberDiff line numberDiff line change
@@ -2,8 +2,8 @@ fastdownload==0.0.7
22
onnx==1.17.0
33
openvino==2025.0
44
pycocotools==2.0.7
5-
torch==2.5.1
5+
torch==2.6.0
66
torchmetrics==1.0.1
7-
torchvision==0.20.1
7+
torchvision==0.21.0
88
numpy<2
99
setuptools<=72.1.0

‎examples/post_training_quantization/torch_fx/resnet18/requirements.txt

+2-2
Original file line numberDiff line numberDiff line change
@@ -1,4 +1,4 @@
11
fastdownload==0.0.7
22
openvino==2025.0
3-
torch==2.5.1
4-
torchvision==0.20.1
3+
torch==2.6.0
4+
torchvision==0.21.0

‎examples/quantization_aware_training/torch/resnet18/main.py

+1-1
Original file line numberDiff line numberDiff line change
@@ -293,7 +293,7 @@ def transform_fn(data_item):
293293
acc1_int8_best = acc1_int8
294294

295295
# Load quantization modules and parameters from best checkpoint to the source model.
296-
ckpt = torch.load(ROOT / BEST_CKPT_NAME)
296+
ckpt = torch.load(ROOT / BEST_CKPT_NAME, weights_only=False)
297297
quantized_model = nncf.torch.load_from_config(
298298
deepcopy(model), ckpt["compression_config"], torch.ones((1, 3, IMAGE_SIZE, IMAGE_SIZE)).to(device)
299299
)

‎examples/quantization_aware_training/torch/resnet18/requirements.txt

+2-2
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,5 @@
11
fastdownload==0.0.7
22
openvino==2025.0
3-
torch==2.5.1
4-
torchvision==0.20.1
3+
torch==2.6.0
4+
torchvision==0.21.0
55
setuptools<=72.1.0

‎examples/torch/common/model_loader.py

+4-2
Original file line numberDiff line numberDiff line change
@@ -57,7 +57,7 @@ def load_model(
5757
# Check if provided path is a url and download the checkpoint if yes
5858
if is_url(weights_path):
5959
weights_path = download_checkpoint(weights_path)
60-
sd = torch.load(weights_path, map_location="cpu", pickle_module=restricted_pickle_module)
60+
sd = torch.load(weights_path, map_location="cpu", pickle_module=restricted_pickle_module, weights_only=False)
6161
if MODEL_STATE_ATTR in sd:
6262
sd = sd[MODEL_STATE_ATTR]
6363
load_state(loaded_model, sd, is_resume=False)
@@ -71,7 +71,9 @@ def load_model(
7171
def load_resuming_checkpoint(resuming_checkpoint_path: str):
7272
if osp.isfile(resuming_checkpoint_path):
7373
logger.info(f"=> loading checkpoint '{resuming_checkpoint_path}'")
74-
checkpoint = torch.load(resuming_checkpoint_path, map_location="cpu", pickle_module=restricted_pickle_module)
74+
checkpoint = torch.load(
75+
resuming_checkpoint_path, map_location="cpu", pickle_module=restricted_pickle_module, weights_only=False
76+
)
7577
return checkpoint
7678
msg = f"no checkpoint found at '{resuming_checkpoint_path}'"
7779
raise FileNotFoundError(msg)

‎examples/torch/common/models/classification/mobilenet_v2_cifar10.py

+1-1
Original file line numberDiff line numberDiff line change
@@ -143,6 +143,6 @@ def mobilenet_v2_cifar10(pretrained=False, progress=True, device="cpu", **kwargs
143143
model = MobileNetV2(**kwargs)
144144
if pretrained:
145145
script_dir = os.path.dirname(__file__)
146-
state_dict = torch.load(script_dir + "/state_dicts/mobilenet_v2.pt", map_location=device)
146+
state_dict = torch.load(script_dir + "/state_dicts/mobilenet_v2.pt", map_location=device, weights_only=False)
147147
model.load_state_dict(state_dict)
148148
return model

‎examples/torch/common/models/classification/resnet_cifar10.py

+1-1
Original file line numberDiff line numberDiff line change
@@ -270,7 +270,7 @@ def _resnet(arch, block, layers, pretrained, progress, device, **kwargs):
270270
model = ResNet(block, layers, **kwargs)
271271
if pretrained:
272272
script_dir = os.path.dirname(__file__)
273-
state_dict = torch.load(script_dir + "/state_dicts/" + arch + ".pt", map_location=device)
273+
state_dict = torch.load(script_dir + "/state_dicts/" + arch + ".pt", map_location=device, weights_only=False)
274274
model.load_state_dict(state_dict)
275275
return model
276276

‎examples/torch/common/models/classification/vgg11_bn_cifar10.py

+1-1
Original file line numberDiff line numberDiff line change
@@ -136,7 +136,7 @@ def _vgg(arch, cfg, batch_norm, pretrained, progress, device, **kwargs):
136136
model = VGG(make_layers(cfgs[cfg], batch_norm=batch_norm), **kwargs)
137137
if pretrained:
138138
script_dir = os.path.dirname(__file__)
139-
state_dict = torch.load(script_dir + "/state_dicts/" + arch + ".pt", map_location=device)
139+
state_dict = torch.load(script_dir + "/state_dicts/" + arch + ".pt", map_location=device, weights_only=False)
140140
model.load_state_dict(state_dict)
141141
return model
142142

‎examples/torch/object_detection/main.py

+1-1
Original file line numberDiff line numberDiff line change
@@ -361,7 +361,7 @@ def create_model(config: SampleConfig):
361361
ssd_net = build_ssd(config.model, config.ssd_params, image_size, config.num_classes, config)
362362
weights = config.get("weights")
363363
if weights:
364-
sd = torch.load(weights, map_location="cpu", pickle_module=restricted_pickle_module)
364+
sd = torch.load(weights, map_location="cpu", pickle_module=restricted_pickle_module, weights_only=False)
365365
sd = sd["state_dict"]
366366
load_state(ssd_net, sd)
367367

‎examples/torch/object_detection/models/ssd_mobilenet.py

+3-1
Original file line numberDiff line numberDiff line change
@@ -109,7 +109,9 @@ def build_ssd_mobilenet(cfg, size, num_classes, config):
109109
# may be used to perform arbitrary code execution during unpickling. Only load the data you
110110
# trust.
111111
#
112-
basenet_weights = torch.load(config.basenet, pickle_module=restricted_pickle_module)["state_dict"]
112+
basenet_weights = torch.load(config.basenet, pickle_module=restricted_pickle_module, weights_only=False)[
113+
"state_dict"
114+
]
113115
new_weights = {}
114116
for wn, wv in basenet_weights.items():
115117
wn = wn.replace("model.", "")

‎examples/torch/object_detection/models/ssd_vgg.py

+7-2
Original file line numberDiff line numberDiff line change
@@ -81,7 +81,12 @@ def load_weights(self, base_file):
8181
# trust.
8282
#
8383
self.load_state_dict(
84-
torch.load(base_file, map_location=lambda storage, loc: storage, pickle_module=restricted_pickle_module)
84+
torch.load(
85+
base_file,
86+
weights_only=False,
87+
map_location=lambda storage, loc: storage,
88+
pickle_module=restricted_pickle_module,
89+
)
8590
)
8691
logger.debug("Finished!")
8792
else:
@@ -170,7 +175,7 @@ def build_ssd_vgg(cfg, size, num_classes, config):
170175

171176
if config.basenet and (config.resuming_checkpoint_path is None) and (config.weights is None):
172177
logger.debug("Loading base network...")
173-
basenet_weights = torch.load(config.basenet, pickle_module=restricted_pickle_module)
178+
basenet_weights = torch.load(config.basenet, pickle_module=restricted_pickle_module, weights_only=False)
174179
new_weights = {}
175180
for wn, wv in basenet_weights.items():
176181
wn = wn.replace("features.", "")

‎nncf/experimental/torch/nas/bootstrapNAS/elasticity/elastic_width.py

+1-1
Original file line numberDiff line numberDiff line change
@@ -556,7 +556,7 @@ def __init__(
556556
self._filter_importance_fn = filter_importance_fn
557557
self._external_importance = None
558558
if external_importance_path is not None:
559-
self._external_importance = torch.load(external_importance_path)
559+
self._external_importance = torch.load(external_importance_path, weights_only=False)
560560
nncf_logger.debug("Loaded custom external weight importance.")
561561
self._weights_normalizer_fn = weights_normalizer_fn
562562
self._add_dynamic_inputs = add_dynamic_inputs

‎nncf/experimental/torch/nas/bootstrapNAS/search/supernet.py

+6-2
Original file line numberDiff line numberDiff line change
@@ -60,9 +60,13 @@ def from_checkpoint(
6060
:return: SuperNetwork with wrapped functionality.
6161
"""
6262
nncf_network = create_nncf_network(model, nncf_config)
63-
compression_state = torch.load(supernet_elasticity_path, map_location=torch.device(nncf_config.device))
63+
compression_state = torch.load(
64+
supernet_elasticity_path, map_location=torch.device(nncf_config.device), weights_only=False
65+
)
6466
model, elasticity_ctrl = resume_compression_from_state(nncf_network, compression_state)
65-
model_weights = torch.load(supernet_weights_path, map_location=torch.device(nncf_config.device))
67+
model_weights = torch.load(
68+
supernet_weights_path, map_location=torch.device(nncf_config.device), weights_only=False
69+
)
6670
load_state(model, model_weights, is_resume=True)
6771
elasticity_ctrl.multi_elasticity_handler.activate_maximum_subnet()
6872
return TrainedSuperNet(elasticity_ctrl, model)

‎nncf/experimental/torch/nas/bootstrapNAS/training/training_algorithm.py

+1-1
Original file line numberDiff line numberDiff line change
@@ -255,7 +255,7 @@ def from_checkpoint(
255255
msg = f"no checkpoint found at '{resuming_checkpoint_path}'"
256256
raise FileNotFoundError(msg)
257257
nncf_logger.info(f"=> loading checkpoint '{resuming_checkpoint_path}'")
258-
checkpoint = torch.load(resuming_checkpoint_path, map_location="cpu")
258+
checkpoint = torch.load(resuming_checkpoint_path, map_location="cpu", weights_only=False)
259259

260260
training_state = checkpoint[cls._state_names.TRAINING_ALGO_STATE]
261261
nncf_config = NNCFConfig()

‎nncf/experimental/torch2/function_hook/handle_inner_functions.py

+15-9
Original file line numberDiff line numberDiff line change
@@ -64,7 +64,6 @@ def multi_head_attention_forward(
6464
) -> Tuple[Tensor, Optional[Tensor]]:
6565

6666
is_batched = _mha_shape_check(query, key, value, key_padding_mask, attn_mask, num_heads)
67-
6867
if not is_batched:
6968
query = query.unsqueeze(1)
7069
key = key.unsqueeze(1)
@@ -91,7 +90,6 @@ def multi_head_attention_forward(
9190
)
9291

9392
if is_causal and key_padding_mask is None and not need_weights:
94-
9593
attn_mask = None
9694
else:
9795
attn_mask = _canonical_mask(
@@ -104,7 +102,6 @@ def multi_head_attention_forward(
104102
)
105103

106104
if key_padding_mask is not None:
107-
108105
is_causal = False
109106

110107
assert (
@@ -133,7 +130,17 @@ def multi_head_attention_forward(
133130
b_q = b_k = b_v = None
134131
else:
135132
b_q, b_k, b_v = in_proj_bias.chunk(3)
136-
q, k, v = _in_projection(query, key, value, q_proj_weight, k_proj_weight, v_proj_weight, b_q, b_k, b_v)
133+
q, k, v = _in_projection(
134+
query,
135+
key,
136+
value,
137+
q_proj_weight,
138+
k_proj_weight,
139+
v_proj_weight,
140+
b_q,
141+
b_k,
142+
b_v,
143+
)
137144

138145
if attn_mask is not None:
139146
if attn_mask.dim() == 2:
@@ -195,10 +202,9 @@ def multi_head_attention_forward(
195202
src_len = k.size(1)
196203

197204
if key_padding_mask is not None:
198-
assert key_padding_mask.shape == (
199-
bsz,
200-
src_len,
201-
), f"expecting key_padding_mask shape of {(bsz, src_len)}, but got {key_padding_mask.shape}"
205+
if not torch.jit.is_scripting() and not torch.jit.is_tracing(): # type: ignore
206+
_check_key_padding_mask(key_padding_mask, src_len, bsz) # type: ignore
207+
202208
key_padding_mask = (
203209
key_padding_mask.view(bsz, 1, 1, src_len).expand(-1, num_heads, -1, -1).reshape(bsz * num_heads, 1, src_len)
204210
)
@@ -211,7 +217,7 @@ def multi_head_attention_forward(
211217
dropout_p = 0.0
212218

213219
if need_weights:
214-
B, Nt, E = q.shape # noqa: F841
220+
_B, _Nt, E = q.shape # noqa: F841
215221
q_scaled = q * math.sqrt(1.0 / float(E))
216222

217223
assert not (is_causal and attn_mask is None), "FIXME: is_causal not implemented for need_weights"

‎nncf/torch/accuracy_aware_training/runner.py

+1-1
Original file line numberDiff line numberDiff line change
@@ -122,7 +122,7 @@ def _load_checkpoint(self, model, checkpoint_path):
122122
if self._load_checkpoint_fn is not None:
123123
self._load_checkpoint_fn(model, checkpoint_path)
124124
else:
125-
resuming_checkpoint = torch.load(checkpoint_path, map_location="cpu")
125+
resuming_checkpoint = torch.load(checkpoint_path, map_location="cpu", weights_only=False)
126126
resuming_model_state_dict = resuming_checkpoint.get("state_dict", resuming_checkpoint)
127127
load_state(model, resuming_model_state_dict, is_resume=True)
128128

‎nncf/torch/automl/agent/ddpg/ddpg.py

+2-2
Original file line numberDiff line numberDiff line change
@@ -269,9 +269,9 @@ def load_weights(self, output):
269269
if output is None:
270270
return
271271

272-
self.actor.load_state_dict(torch.load(f"{output}/actor.pkl"))
272+
self.actor.load_state_dict(torch.load(f"{output}/actor.pkl", weights_only=False))
273273

274-
self.critic.load_state_dict(torch.load(f"{output}/critic.pkl"))
274+
self.critic.load_state_dict(torch.load(f"{output}/critic.pkl", weights_only=False))
275275

276276
def save_model(self, output):
277277
torch.save(self.actor.state_dict(), f"{output}/actor.pkl")

‎nncf/torch/extensions/src/quantization/cpu/functions_cpu.cpp

+2-2
Original file line numberDiff line numberDiff line change
@@ -81,7 +81,7 @@ at::Tensor q_forward(
8181

8282
at::Tensor output;
8383

84-
DISPATCH_TENSOR_DATA_TYPES(input.type(), "q_cpu_forward", ([&] {
84+
DISPATCH_TENSOR_DATA_TYPES(input.scalar_type(), "q_cpu_forward", ([&] {
8585
output = q_cpu_forward<scalar_t>(input, input_low, input_range, levels);
8686
}));
8787

@@ -103,7 +103,7 @@ std::vector<at::Tensor> q_backward(
103103
CHECK_INPUT(input_range);
104104

105105
std::vector<at::Tensor> results;
106-
DISPATCH_TENSOR_DATA_TYPES(input.type(), "q_cpu_backward", ([&] {
106+
DISPATCH_TENSOR_DATA_TYPES(input.scalar_type(), "q_cpu_backward", ([&] {
107107
results = q_cpu_backward<scalar_t>(grad_output, input, input_low, input_range, levels, level_low, level_high, is_asymmetric);
108108
}));
109109

‎nncf/torch/quantization/precision_init/hawq_init.py

+1-1
Original file line numberDiff line numberDiff line change
@@ -478,7 +478,7 @@ def _calc_traces(
478478
tolerance: float,
479479
) -> TracesPerLayer:
480480
if self._traces_per_layer_path:
481-
return TracesPerLayer(torch.load(self._traces_per_layer_path).to(self._init_device))
481+
return TracesPerLayer(torch.load(self._traces_per_layer_path, weights_only=False).to(self._init_device))
482482

483483
quantizers_switcher = QuantizersSwitcher(list(self._all_quantizers_per_scope.values()))
484484
params_to_restore = self.disable_all_gradients_except_weights_of_quantized_modules(

‎nncf/version.py

+1-1
Original file line numberDiff line numberDiff line change
@@ -12,6 +12,6 @@
1212
__version__ = "2.16.0"
1313

1414

15-
BKC_TORCH_SPEC = "==2.5.*"
15+
BKC_TORCH_SPEC = "==2.6.*"
1616
BKC_TF_SPEC = "==2.15.*"
1717
STRICT_TF_SPEC = ">=2.9.3,<2.16.0"

‎tests/post_training/data/ptq_reference_data.yaml

+32
Original file line numberDiff line numberDiff line change
@@ -42,8 +42,16 @@ torchvision/resnet18_backend_CUDA_TORCH:
4242
metric_value: 0.69152
4343
torchvision/resnet18_backend_FX_TORCH:
4444
metric_value: 0.6946
45+
exception_xfail_reason:
46+
type: "OpConversionFailure"
47+
error_message: "Check 'is_conversion_successful' failed at src/frontends/pytorch/src/frontend.cpp:171:\nFrontEnd API failed with OpConversionFailure:\nModel wasn't fully converted. Failed operations detailed log:\n-- torch::None with a message:\nNone constant cannot be converted to OpenVINO opset and should be removed by consuming operation.\nSummary:\n-- No conversion rule found for operations: aten.adaptive_avg_pool2d.default, aten.conv2d.default, aten.linear.default, aten.max_pool2d.default\n-- Conversion is failed for: torch::None\n"
48+
message: "Issue-162009"
4549
torchvision/resnet18_backend_CUDA_FX_TORCH:
4650
metric_value: 0.6946
51+
exception_xfail_reason:
52+
type: "OpConversionFailure"
53+
error_message: "Check 'is_conversion_successful' failed at src/frontends/pytorch/src/frontend.cpp:171:\nFrontEnd API failed with OpConversionFailure:\nModel wasn't fully converted. Failed operations detailed log:\n-- torch::None with a message:\nNone constant cannot be converted to OpenVINO opset and should be removed by consuming operation.\nSummary:\n-- No conversion rule found for operations: aten.adaptive_avg_pool2d.default, aten.conv2d.default, aten.linear.default, aten.max_pool2d.default\n-- Conversion is failed for: torch::None\n"
54+
message: "Issue-162009"
4755
torchvision/mobilenet_v3_small_BC_backend_FP32:
4856
metric_value: 0.6766
4957
torchvision/mobilenet_v3_small_BC_backend_OV:
@@ -52,24 +60,48 @@ torchvision/mobilenet_v3_small_BC_backend_ONNX:
5260
metric_value: 0.6679
5361
torchvision/mobilenet_v3_small_BC_backend_FX_TORCH:
5462
metric_value: 0.6679
63+
exception_xfail_reason:
64+
type: "OpConversionFailure"
65+
error_message: "Check 'is_conversion_successful' failed at src/frontends/pytorch/src/frontend.cpp:171:\nFrontEnd API failed with OpConversionFailure:\nModel wasn't fully converted. Failed operations detailed log:\n-- torch::None with a message:\nNone constant cannot be converted to OpenVINO opset and should be removed by consuming operation.\nSummary:\n-- No conversion rule found for operations: aten.adaptive_avg_pool2d.default, aten.conv2d.default, aten.linear.default\n-- Conversion is failed for: torch::None\n"
66+
message: "Issue-162009"
5567
torchvision/mobilenet_v3_small_BC_backend_CUDA_FX_TORCH:
5668
metric_value: 0.6664
69+
exception_xfail_reason:
70+
type: "OpConversionFailure"
71+
error_message: "Check 'is_conversion_successful' failed at src/frontends/pytorch/src/frontend.cpp:171:\nFrontEnd API failed with OpConversionFailure:\nModel wasn't fully converted. Failed operations detailed log:\n-- torch::None with a message:\nNone constant cannot be converted to OpenVINO opset and should be removed by consuming operation.\nSummary:\n-- No conversion rule found for operations: aten.adaptive_avg_pool2d.default, aten.conv2d.default, aten.linear.default\n-- Conversion is failed for: torch::None\n"
72+
message: "Issue-162009"
5773
torchvision/vit_b_16_backend_FP32:
5874
metric_value: 0.8107
5975
torchvision/vit_b_16_backend_OV:
6076
metric_value: 0.80948
6177
torchvision/vit_b_16_backend_FX_TORCH:
6278
metric_value: 0.80922
79+
exception_xfail_reason:
80+
type: "OpConversionFailure"
81+
error_message: "Check 'is_conversion_successful' failed at src/frontends/pytorch/src/frontend.cpp:171:\nFrontEnd API failed with OpConversionFailure:\nModel wasn't fully converted.\nSummary:\n-- No conversion rule found for operations: aten.conv2d.default, aten.layer_norm.default, aten.linear.default, aten.scaled_dot_product_attention.default\n"
82+
message: "Issue-162009"
6383
torchvision/vit_b_16_backend_CUDA_FX_TORCH:
6484
metric_value: 0.80922
85+
exception_xfail_reason:
86+
type: "OpConversionFailure"
87+
error_message: "Check 'is_conversion_successful' failed at src/frontends/pytorch/src/frontend.cpp:171:\nFrontEnd API failed with OpConversionFailure:\nModel wasn't fully converted.\nSummary:\n-- No conversion rule found for operations: aten.conv2d.default, aten.layer_norm.default, aten.linear.default, aten.scaled_dot_product_attention.default\n"
88+
message: "Issue-162009"
6589
torchvision/swin_v2_s_backend_FP32:
6690
metric_value: 0.83712
6791
torchvision/swin_v2_s_backend_OV:
6892
metric_value: 0.83638
6993
torchvision/swin_v2_s_backend_FX_TORCH:
7094
metric_value: 0.8360
95+
exception_xfail_reason:
96+
type: "OpConversionFailure"
97+
error_message: "Check 'is_conversion_successful' failed at src/frontends/pytorch/src/frontend.cpp:171:\nFrontEnd API failed with OpConversionFailure:\nModel wasn't fully converted.\nSummary:\n-- No conversion rule found for operations: aten.adaptive_avg_pool2d.default, aten.conv2d.default, aten.layer_norm.default, aten.linear.default, aten.matmul.default, aten.pad.default, aten.softmax.int, aten.where.ScalarSelf\n"
98+
message: "Issue-162009"
7199
torchvision/swin_v2_s_backend_CUDA_FX_TORCH:
72100
metric_value: 0.8360
101+
exception_xfail_reason:
102+
type: "OpConversionFailure"
103+
error_message: "Check 'is_conversion_successful' failed at src/frontends/pytorch/src/frontend.cpp:171:\nFrontEnd API failed with OpConversionFailure:\nModel wasn't fully converted.\nSummary:\n-- No conversion rule found for operations: aten.adaptive_avg_pool2d.default, aten.conv2d.default, aten.layer_norm.default, aten.linear.default, aten.matmul.default, aten.pad.default, aten.softmax.int, aten.where.ScalarSelf\n"
104+
message: "Issue-162009"
73105
timm/crossvit_9_240_backend_CUDA_TORCH:
74106
metric_value: 0.7275
75107
timm/crossvit_9_240_backend_FP32:

0 commit comments

Comments
 (0)