AudioTextHTDemucs / src /models /stem_separation /CLAP_Text_Model_Fwd_Pass.txt
jacob1576's picture
Add application file and dependencies
7417a6a
** NOTE: get_text_features() method does a final projection to 512 dims and normalization after this forward pass
=========================================================================================================
Layer (type:depth-idx) Output Shape Param #
=========================================================================================================
ClapTextModel [1, 768] --
├─ClapTextEmbeddings: 1-1 [1, 5, 768] --
│ └─Embedding: 2-1 [1, 5, 768] (38,603,520)
│ └─Embedding: 2-2 [1, 5, 768] (768)
│ └─Embedding: 2-3 [1, 5, 768] (394,752)
│ └─LayerNorm: 2-4 [1, 5, 768] (1,536)
│ └─Dropout: 2-5 [1, 5, 768] --
├─ClapTextEncoder: 1-2 [1, 5, 768] --
│ └─ModuleList: 2-6 -- --
│ │ └─ClapTextLayer: 3-1 [1, 5, 768] (7,087,872)
│ │ └─ClapTextLayer: 3-2 [1, 5, 768] (7,087,872)
│ │ └─ClapTextLayer: 3-3 [1, 5, 768] (7,087,872)
│ │ └─ClapTextLayer: 3-4 [1, 5, 768] (7,087,872)
│ │ └─ClapTextLayer: 3-5 [1, 5, 768] (7,087,872)
│ │ └─ClapTextLayer: 3-6 [1, 5, 768] (7,087,872)
│ │ └─ClapTextLayer: 3-7 [1, 5, 768] (7,087,872)
│ │ └─ClapTextLayer: 3-8 [1, 5, 768] (7,087,872)
│ │ └─ClapTextLayer: 3-9 [1, 5, 768] (7,087,872)
│ │ └─ClapTextLayer: 3-10 [1, 5, 768] (7,087,872)
│ │ └─ClapTextLayer: 3-11 [1, 5, 768] (7,087,872)
│ │ └─ClapTextLayer: 3-12 [1, 5, 768] (7,087,872)
├─ClapTextPooler: 1-3 [1, 768] --
│ └─Linear: 2-7 [1, 768] (590,592)
│ └─Tanh: 2-8 [1, 768] --
=========================================================================================================
Total params: 124,645,632
Trainable params: 0
Non-trainable params: 124,645,632
Total mult-adds (Units.MEGABYTES): 124.65
=========================================================================================================
Input size (MB): 0.00
Forward/backward pass size (MB): 4.18
Params size (MB): 498.58
Estimated Total Size (MB): 502.77
=========================================================================================================