titans-pytorch 0.1.28__tar.gz → 0.1.29__tar.gz
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.
- {titans_pytorch-0.1.28 → titans_pytorch-0.1.29}/PKG-INFO +1 -1
- {titans_pytorch-0.1.28 → titans_pytorch-0.1.29}/pyproject.toml +1 -1
- {titans_pytorch-0.1.28 → titans_pytorch-0.1.29}/train_mac.py +0 -2
- {titans_pytorch-0.1.28 → titans_pytorch-0.1.29}/.github/workflows/python-publish.yml +0 -0
- {titans_pytorch-0.1.28 → titans_pytorch-0.1.29}/.github/workflows/test.yaml +0 -0
- {titans_pytorch-0.1.28 → titans_pytorch-0.1.29}/.gitignore +0 -0
- {titans_pytorch-0.1.28 → titans_pytorch-0.1.29}/LICENSE +0 -0
- {titans_pytorch-0.1.28 → titans_pytorch-0.1.29}/README.md +0 -0
- {titans_pytorch-0.1.28 → titans_pytorch-0.1.29}/data/README.md +0 -0
- {titans_pytorch-0.1.28 → titans_pytorch-0.1.29}/data/enwik8.gz +0 -0
- {titans_pytorch-0.1.28 → titans_pytorch-0.1.29}/fig1.png +0 -0
- {titans_pytorch-0.1.28 → titans_pytorch-0.1.29}/fig2.png +0 -0
- {titans_pytorch-0.1.28 → titans_pytorch-0.1.29}/tests/test_titans.py +0 -0
- {titans_pytorch-0.1.28 → titans_pytorch-0.1.29}/titans_pytorch/__init__.py +0 -0
- {titans_pytorch-0.1.28 → titans_pytorch-0.1.29}/titans_pytorch/associative_scan.py +0 -0
- {titans_pytorch-0.1.28 → titans_pytorch-0.1.29}/titans_pytorch/mac_transformer.py +0 -0
- {titans_pytorch-0.1.28 → titans_pytorch-0.1.29}/titans_pytorch/titans.py +0 -0
@@ -37,7 +37,6 @@ NEURAL_MEM_SEGMENT_LEN = WINDOW_SIZE // 2 # set smaller for more granularity for
|
|
37
37
|
SLIDING_WINDOWS = True
|
38
38
|
STORE_ATTN_POOL_CHUNKS = True # whether to use attention pooling for chunk derived momentum, per-layer lr mod, decay
|
39
39
|
KV_RECON_LOSS_WEIGHT = 0.
|
40
|
-
LEARNED_MEM_MODEL_WEIGHTS = True
|
41
40
|
|
42
41
|
# experiment related
|
43
42
|
|
@@ -91,7 +90,6 @@ model = MemoryAsContextTransformer(
|
|
91
90
|
attn_pool_chunks = STORE_ATTN_POOL_CHUNKS,
|
92
91
|
momentum = NEURAL_MEM_MOMENTUM,
|
93
92
|
use_accelerated_scan = USE_ACCELERATED_SCAN,
|
94
|
-
learned_mem_model_weights = LEARNED_MEM_MODEL_WEIGHTS,
|
95
93
|
default_model_kwargs = dict(
|
96
94
|
depth = NEURAL_MEMORY_DEPTH,
|
97
95
|
)
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|