ai-edge-torch-nightly 0.3.0.dev20250130__py3-none-any.whl → 0.3.0.dev20250201__py3-none-any.whl
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- ai_edge_torch/generative/examples/qwen_vl/__init__.py +14 -0
- ai_edge_torch/generative/examples/qwen_vl/decoder.py +91 -0
- ai_edge_torch/generative/examples/qwen_vl/verify_decoder.py +77 -0
- ai_edge_torch/version.py +1 -1
- {ai_edge_torch_nightly-0.3.0.dev20250130.dist-info → ai_edge_torch_nightly-0.3.0.dev20250201.dist-info}/METADATA +1 -1
- {ai_edge_torch_nightly-0.3.0.dev20250130.dist-info → ai_edge_torch_nightly-0.3.0.dev20250201.dist-info}/RECORD +9 -6
- {ai_edge_torch_nightly-0.3.0.dev20250130.dist-info → ai_edge_torch_nightly-0.3.0.dev20250201.dist-info}/LICENSE +0 -0
- {ai_edge_torch_nightly-0.3.0.dev20250130.dist-info → ai_edge_torch_nightly-0.3.0.dev20250201.dist-info}/WHEEL +0 -0
- {ai_edge_torch_nightly-0.3.0.dev20250130.dist-info → ai_edge_torch_nightly-0.3.0.dev20250201.dist-info}/top_level.txt +0 -0
@@ -0,0 +1,14 @@
|
|
1
|
+
# Copyright 2025 The AI Edge Torch Authors.
|
2
|
+
#
|
3
|
+
# Licensed under the Apache License, Version 2.0 (the "License");
|
4
|
+
# you may not use this file except in compliance with the License.
|
5
|
+
# You may obtain a copy of the License at
|
6
|
+
#
|
7
|
+
# http://www.apache.org/licenses/LICENSE-2.0
|
8
|
+
#
|
9
|
+
# Unless required by applicable law or agreed to in writing, software
|
10
|
+
# distributed under the License is distributed on an "AS IS" BASIS,
|
11
|
+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
12
|
+
# See the License for the specific language governing permissions and
|
13
|
+
# limitations under the License.
|
14
|
+
# ==============================================================================
|
@@ -0,0 +1,91 @@
|
|
1
|
+
# Copyright 2025 The AI Edge Torch Authors.
|
2
|
+
#
|
3
|
+
# Licensed under the Apache License, Version 2.0 (the "License");
|
4
|
+
# you may not use this file except in compliance with the License.
|
5
|
+
# You may obtain a copy of the License at
|
6
|
+
#
|
7
|
+
# http://www.apache.org/licenses/LICENSE-2.0
|
8
|
+
#
|
9
|
+
# Unless required by applicable law or agreed to in writing, software
|
10
|
+
# distributed under the License is distributed on an "AS IS" BASIS,
|
11
|
+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
12
|
+
# See the License for the specific language governing permissions and
|
13
|
+
# limitations under the License.
|
14
|
+
# ==============================================================================
|
15
|
+
|
16
|
+
"""Example of building decoder for Qwen 2.5 VL models."""
|
17
|
+
|
18
|
+
import ai_edge_torch.generative.layers.model_config as cfg
|
19
|
+
from ai_edge_torch.generative.utilities import model_builder
|
20
|
+
from torch import nn
|
21
|
+
|
22
|
+
TENSOR_NAMES = model_builder.TENSOR_NAMES
|
23
|
+
|
24
|
+
|
25
|
+
class Decoder(model_builder.DecoderOnlyModel):
|
26
|
+
"""A decoder for Qwen-VL model built from the Edge Generative API layers."""
|
27
|
+
pass
|
28
|
+
|
29
|
+
|
30
|
+
def get_decoder_config(kv_cache_max_len: int = 1024) -> cfg.ModelConfig:
|
31
|
+
"""Returns the model config for a Qwen 2.5 VL 3B model.
|
32
|
+
|
33
|
+
Args:
|
34
|
+
kv_cache_max_len (int): The maximum sequence length of the KV cache. Default
|
35
|
+
is 1024.
|
36
|
+
|
37
|
+
Returns:
|
38
|
+
The model config for a Qwen 2.5 VL 3B model.
|
39
|
+
"""
|
40
|
+
attn_config = cfg.AttentionConfig(
|
41
|
+
num_heads=16,
|
42
|
+
head_dim=128,
|
43
|
+
num_query_groups=2,
|
44
|
+
rotary_base=1000000,
|
45
|
+
rotary_percentage=1.0,
|
46
|
+
qkv_use_bias=True,
|
47
|
+
)
|
48
|
+
ff_config = cfg.FeedForwardConfig(
|
49
|
+
type=cfg.FeedForwardType.GATED,
|
50
|
+
activation=cfg.ActivationConfig(cfg.ActivationType.SILU),
|
51
|
+
intermediate_size=11008,
|
52
|
+
)
|
53
|
+
norm_config = cfg.NormalizationConfig(
|
54
|
+
type=cfg.NormalizationType.RMS_NORM,
|
55
|
+
epsilon=1e-06,
|
56
|
+
)
|
57
|
+
block_config = cfg.TransformerBlockConfig(
|
58
|
+
attn_config=attn_config,
|
59
|
+
ff_config=ff_config,
|
60
|
+
pre_attention_norm_config=norm_config,
|
61
|
+
post_attention_norm_config=norm_config,
|
62
|
+
)
|
63
|
+
config = cfg.ModelConfig(
|
64
|
+
vocab_size=151936,
|
65
|
+
num_layers=36,
|
66
|
+
max_seq_len=32768,
|
67
|
+
embedding_dim=2048,
|
68
|
+
kv_cache_max_len=kv_cache_max_len,
|
69
|
+
block_configs=block_config,
|
70
|
+
final_norm_config=norm_config,
|
71
|
+
enable_hlfb=True,
|
72
|
+
)
|
73
|
+
return config
|
74
|
+
|
75
|
+
|
76
|
+
def get_fake_decoder_config(**kwargs) -> cfg.ModelConfig:
|
77
|
+
config = get_decoder_config(**kwargs)
|
78
|
+
config.vocab_size = 128
|
79
|
+
config.num_layers = 2
|
80
|
+
# Decoder has only one block config.
|
81
|
+
config.block_config(0).ff_config.intermediate_size = 64
|
82
|
+
return config
|
83
|
+
|
84
|
+
|
85
|
+
def build_decoder(checkpoint_path: str, **kwargs) -> nn.Module:
|
86
|
+
return model_builder.build_decoder_only_model(
|
87
|
+
checkpoint_path=checkpoint_path,
|
88
|
+
config=get_decoder_config(**kwargs),
|
89
|
+
tensor_names=TENSOR_NAMES,
|
90
|
+
model_class=Decoder,
|
91
|
+
)
|
@@ -0,0 +1,77 @@
|
|
1
|
+
# Copyright 2024 The AI Edge Torch Authors.
|
2
|
+
#
|
3
|
+
# Licensed under the Apache License, Version 2.0 (the "License");
|
4
|
+
# you may not use this file except in compliance with the License.
|
5
|
+
# You may obtain a copy of the License at
|
6
|
+
#
|
7
|
+
# http://www.apache.org/licenses/LICENSE-2.0
|
8
|
+
#
|
9
|
+
# Unless required by applicable law or agreed to in writing, software
|
10
|
+
# distributed under the License is distributed on an "AS IS" BASIS,
|
11
|
+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
12
|
+
# See the License for the specific language governing permissions and
|
13
|
+
# limitations under the License.
|
14
|
+
# ==============================================================================
|
15
|
+
|
16
|
+
"""Verifies the reauthored decoder of Qwen 2.5 VL 3B models."""
|
17
|
+
|
18
|
+
import logging
|
19
|
+
import pathlib
|
20
|
+
|
21
|
+
from absl import app
|
22
|
+
from ai_edge_torch.generative.examples.qwen_vl import decoder
|
23
|
+
from ai_edge_torch.generative.utilities import verifier
|
24
|
+
import torch
|
25
|
+
import transformers
|
26
|
+
|
27
|
+
|
28
|
+
class DecoderWrapper(verifier.ModelWrapper):
|
29
|
+
"""Wraps the decoder of Qwen 2.5 VL models for verification."""
|
30
|
+
|
31
|
+
def __init__(self, model: torch.nn.Module, lm_head: torch.nn.Module):
|
32
|
+
super().__init__(model)
|
33
|
+
self.lm_head = lm_head
|
34
|
+
|
35
|
+
def forward(self, tokens: torch.Tensor) -> torch.Tensor:
|
36
|
+
output = self.model.forward(tokens)
|
37
|
+
return self.lm_head(output["last_hidden_state"])
|
38
|
+
|
39
|
+
|
40
|
+
def main(_):
|
41
|
+
checkpoint = "Qwen/Qwen2.5-VL-3B-Instruct"
|
42
|
+
logging.info("Loading the original model from: %s", checkpoint)
|
43
|
+
original_model = (
|
44
|
+
transformers.Qwen2_5_VLForConditionalGeneration.from_pretrained(
|
45
|
+
checkpoint
|
46
|
+
)
|
47
|
+
)
|
48
|
+
|
49
|
+
# Locate the cached dir.
|
50
|
+
cached_config_file = transformers.utils.cached_file(
|
51
|
+
checkpoint, transformers.utils.CONFIG_NAME
|
52
|
+
)
|
53
|
+
reauthored_checkpoint = pathlib.Path(cached_config_file).parent
|
54
|
+
logging.info("Building the reauthored model from: %s", reauthored_checkpoint)
|
55
|
+
reauthored_model = decoder.build_decoder(reauthored_checkpoint)
|
56
|
+
|
57
|
+
# Verify the reauthored model only with input IDs because the original decoder
|
58
|
+
# does not support generate() with prompts.
|
59
|
+
input_ids = [1, 2, 3, 4]
|
60
|
+
try:
|
61
|
+
verifier.verify_with_input_ids(
|
62
|
+
original_model=DecoderWrapper(
|
63
|
+
original_model.model,
|
64
|
+
original_model.lm_head,
|
65
|
+
),
|
66
|
+
reauthored_model=verifier.ReauthoredModelWrapper(reauthored_model),
|
67
|
+
input_ids=input_ids,
|
68
|
+
atol=1e-04,
|
69
|
+
)
|
70
|
+
except AssertionError as e:
|
71
|
+
logging.error("*** FAILED *** verify with input IDs: %s", e)
|
72
|
+
else:
|
73
|
+
logging.info("*** PASSED *** verify with input IDs: %s", input_ids)
|
74
|
+
|
75
|
+
|
76
|
+
if __name__ == "__main__":
|
77
|
+
app.run(main)
|
ai_edge_torch/version.py
CHANGED
@@ -1,6 +1,6 @@
|
|
1
1
|
Metadata-Version: 2.1
|
2
2
|
Name: ai-edge-torch-nightly
|
3
|
-
Version: 0.3.0.
|
3
|
+
Version: 0.3.0.dev20250201
|
4
4
|
Summary: Supporting PyTorch models with the Google AI Edge TFLite runtime.
|
5
5
|
Home-page: https://github.com/google-ai-edge/ai-edge-torch
|
6
6
|
Keywords: On-Device ML,AI,Google,TFLite,PyTorch,LLMs,GenAI
|
@@ -2,7 +2,7 @@ ai_edge_torch/__init__.py,sha256=8sPR_5uXJA4NEE0nIwNdSl-ADOJEoR8hAgYvBQDY70Y,120
|
|
2
2
|
ai_edge_torch/_config.py,sha256=AiqhbcheF7j_ozIGDLC89k1we95aVgFDa-tR6h7UI0s,2529
|
3
3
|
ai_edge_torch/conftest.py,sha256=r0GTrhMRhlmOGrrkvumHN8hkmyug6WvF60vWq8wRIBI,758
|
4
4
|
ai_edge_torch/model.py,sha256=N-pNpTxzhaFGhWhnSGd70lBzb9VlEhTOq5mddU7bvvI,5542
|
5
|
-
ai_edge_torch/version.py,sha256=
|
5
|
+
ai_edge_torch/version.py,sha256=9qgk7SSH80z-tSMSDRr1M3FiR7a69U40_TeqElVeor0,706
|
6
6
|
ai_edge_torch/_convert/__init__.py,sha256=hHLluseD2R0Hh4W6XZRIXY_dRQeYudjsrKGf6LZz65g,671
|
7
7
|
ai_edge_torch/_convert/conversion.py,sha256=mckvxznKLXdF2HuJg_IxQaT5Ty-iWl_iXElHEugH3VI,5452
|
8
8
|
ai_edge_torch/_convert/conversion_utils.py,sha256=Sr8qXVcTwc-ZnZmK7yxVrIOOp1S_vNrwzC0zUvLTI2o,2160
|
@@ -93,6 +93,9 @@ ai_edge_torch/generative/examples/qwen/__init__.py,sha256=hHLluseD2R0Hh4W6XZRIXY
|
|
93
93
|
ai_edge_torch/generative/examples/qwen/convert_to_tflite.py,sha256=tqvXVGNdDehdak9-5DDisACs9VlTwr8eFwcjQ_kZxgc,2776
|
94
94
|
ai_edge_torch/generative/examples/qwen/qwen.py,sha256=Zi_qiQ1JPokXZ95jgSEnQp3F-LKzFCvWvFLKhJjnASo,4199
|
95
95
|
ai_edge_torch/generative/examples/qwen/verify.py,sha256=9_AyEJTeUfvhhID64Rto2bflFPyXMFokdQLsseLUMiI,2775
|
96
|
+
ai_edge_torch/generative/examples/qwen_vl/__init__.py,sha256=JaAnrFoXTl3RJX97XspklkTyqOHVyAgRJsZtzNDd10c,671
|
97
|
+
ai_edge_torch/generative/examples/qwen_vl/decoder.py,sha256=rD_Ch5CzuXeatqv0C3z8vU-zou1z9QDUhoB6V4YTPIg,2829
|
98
|
+
ai_edge_torch/generative/examples/qwen_vl/verify_decoder.py,sha256=FEY_PifD9fQGnERzSOljFLraRIbUVF3XTnCv95A30Cs,2602
|
96
99
|
ai_edge_torch/generative/examples/smollm/__init__.py,sha256=hHLluseD2R0Hh4W6XZRIXY_dRQeYudjsrKGf6LZz65g,671
|
97
100
|
ai_edge_torch/generative/examples/smollm/convert_to_tflite.py,sha256=megskv1oiPhwHSnguoG7zV-esXp1Ns_FPeMLAYKhDb0,2522
|
98
101
|
ai_edge_torch/generative/examples/smollm/convert_v2_to_tflite.py,sha256=CjY1i0iCYxFSjhCpQZwxkmVxILgeo0zu1m0oBrHqyDU,2311
|
@@ -222,8 +225,8 @@ ai_edge_torch/quantize/quant_config.py,sha256=U0KisSW-uZkoMJcy-ZP9W57p3tsa594fr9
|
|
222
225
|
ai_edge_torch/testing/__init__.py,sha256=hHLluseD2R0Hh4W6XZRIXY_dRQeYudjsrKGf6LZz65g,671
|
223
226
|
ai_edge_torch/testing/model_coverage/__init__.py,sha256=5P8J6Zk5YYtDvTBucFvB9NGSRI7Gw_24WnrbhXgycEE,765
|
224
227
|
ai_edge_torch/testing/model_coverage/model_coverage.py,sha256=UPB448aMDUyC0HNYVqio2rcJPnDN0tBQMP08J6vPYew,4718
|
225
|
-
ai_edge_torch_nightly-0.3.0.
|
226
|
-
ai_edge_torch_nightly-0.3.0.
|
227
|
-
ai_edge_torch_nightly-0.3.0.
|
228
|
-
ai_edge_torch_nightly-0.3.0.
|
229
|
-
ai_edge_torch_nightly-0.3.0.
|
228
|
+
ai_edge_torch_nightly-0.3.0.dev20250201.dist-info/LICENSE,sha256=z8d0m5b2O9McPEK1xHG_dWgUBT6EfBDz6wA0F7xSPTA,11358
|
229
|
+
ai_edge_torch_nightly-0.3.0.dev20250201.dist-info/METADATA,sha256=oWpE5lBKL49aWAyb11Xi_eO55pPZZzW2M9gUH_lzVlg,1966
|
230
|
+
ai_edge_torch_nightly-0.3.0.dev20250201.dist-info/WHEEL,sha256=tZoeGjtWxWRfdplE7E3d45VPlLNQnvbKiYnx7gwAy8A,92
|
231
|
+
ai_edge_torch_nightly-0.3.0.dev20250201.dist-info/top_level.txt,sha256=5KXRaF2hwkApYxf7Y8y_tVb9aulGTlbOoNdbx1aKRkE,14
|
232
|
+
ai_edge_torch_nightly-0.3.0.dev20250201.dist-info/RECORD,,
|
File without changes
|
File without changes
|