[CODE QUALITY]

.gitignore

@@ -11,6 +11,7 @@ data
 # Distribution / packaging
 .Python
 build/
+.ruff_cache
 .vscode
 develop-eggs/
 dist/

docs/corporate/zeta_cloud.md

@@ -58,3 +58,5 @@ The estimated timeline for shipping Zeta Cloud is as follows:
 | Marketplace for Pre-Trained Models | A platform for users to buy, sell, or license pre-trained models. | AI developers, companies looking for ready-to-use models. | Transaction fees, subscription for premium listings. |
 | Data Storage and Management | Integrated solutions for data storage, processing, and management. | All users of the platform. | Based on the amount of data stored/processed. |
 | API Access for Third-Party Integrations | Providing API access for integration with other tools and services. | Developers, businesses needing integrations. | Monthly/Annual subscription or pay-per-use. |
+
+

playground/models/flamingo.py

@@ -2,7 +2,6 @@
 import torch.nn.functional as F
 from einops import rearrange
 from torch import einsum, nn
-from zeta.nn.modules.simple_feedforward import SimpleFeedForward
 from zeta.nn.attention.cross_attn_images import MultiModalCrossAttention
 import zeta.nn as znn
 

playground/models/simple_transformer.py

@@ -3,7 +3,6 @@
 from zeta.nn.modules.feedforward import FeedForward
 from zeta.nn.attention.shaped_attention import ShapedAttention
 from zeta.nn.modules.residual import Residual
-from zeta.nn.attention import FlashAttention
 
 
 class SimpleTransformerBlock(nn.Module):

tests/nn/attentions/test_cross_attn.py

@@ -1,6 +1,4 @@
-import pytest
 import torch
-from torch import nn
 from zeta.nn.attention.cross_attention import CrossAttention
 
 # Create an instance of CrossAttention for testing

tests/nn/attentions/test_local_attn_mha.py

@@ -1,6 +1,5 @@
 import pytest
 import torch
-import torch.nn as nn
 from torch.autograd import gradcheck
 from zeta.nn.attention.local_attention_mha import LocalMHA
 

tests/nn/attentions/test_mgqa.py

@@ -1,7 +1,6 @@
 import pytest
 import torch
 from zeta.nn.attention.mgqa import MGQA, CacheView
-from zeta.utils.main import exists
 
 
 # Create an instance of MGQA for testing

tests/nn/attentions/test_shaped_attn.py

@@ -1,7 +1,4 @@
-import pytest
 import torch
-import torch.nn as nn
-import torch.nn.functional as F
 from zeta.nn.attention.shaped_attention import ShapedAttention
 
 

tests/nn/attentions/test_sparse_attn.py

@@ -65,10 +65,6 @@ def test_sparse_attention_forward():
  n_batch = 4
  n_ctx = 1024
  n_embd = 256
- heads = 4
- attn_mode = "all"
- local_attn_ctx = 32
- blocksize = 32
 
  q = torch.randn(n_batch, n_ctx, n_embd)
  k = torch.randn(n_batch, n_ctx, n_embd)

tests/nn/attentions/test_xc_attention.py

@@ -42,7 +42,7 @@ def test_xc_attention_forward_with_invalid_inputs(xc_attention_model):
  with pytest.raises(Exception):
  x = torch.randn(1, 256, 16, 16)
  cond = torch.randn(1, 128) # Mismatched conditioning dimension
- output = xc_attention_model(x, cond)
+ xc_attention_model(x, cond)
 
 
 # Test case to check if XCAttention handles different head configurations correctly
@@ -81,10 +81,10 @@ def test_xc_attention_with_different_cond_dims():
 # Test case to check if XCAttention handles negative input dimensions correctly
 def test_xc_attention_negative_input_dim():
  with pytest.raises(ValueError):
- model = XCAttention(dim=-256, cond_dim=64, heads=8)
+ XCAttention(dim=-256, cond_dim=64, heads=8)
 
 
 # Test case to check if XCAttention handles negative conditioning dimensions correctly
 def test_xc_attention_negative_cond_dim():
  with pytest.raises(ValueError):
- model = XCAttention(dim=256, cond_dim=-64, heads=8)
+ XCAttention(dim=256, cond_dim=-64, heads=8)

tests/nn/biases/test_alibi.py

@@ -152,9 +152,9 @@ def tensors_equal(tensor1, tensor2):
 
 # Test for the existence of a helper function exists
 def test_exists_function():
- assert exists(None) == False
- assert exists(0) == True
- assert exists("Hello") == True
+ assert exists(None) is False
+ assert exists(0) is True
+ assert exists("Hello") is True
 
 
 # Test for the pad_at_dim helper function
@@ -170,8 +170,8 @@ def test_tensors_equal_function():
  tensor2 = torch.tensor([1.0, 2.0, 3.0])
  tensor3 = torch.tensor([1.0, 2.0, 3.1])
 
- assert tensors_equal(tensor1, tensor2) == True
- assert tensors_equal(tensor1, tensor3) == False
+ assert tensors_equal(tensor1, tensor2) is True
+ assert tensors_equal(tensor1, tensor3) is False
 
 
 # Additional tests for tensor manipulation functions
@@ -193,8 +193,8 @@ def test_einops_rearrange_function():
 
 # Test for the nn.Module class inheritance
 def test_nn_module_inheritance():
- assert issubclass(AlibiPositionalBias, nn.Module) == True
- assert issubclass(LearnedAlibiPositionalBias, nn.Module) == True
+ assert issubclass(AlibiPositionalBias, nn.Module) is True
+ assert issubclass(LearnedAlibiPositionalBias, nn.Module) is True
 
 
 # Helper function to create random data

tests/nn/biases/test_relative_position_bias.py

@@ -1,6 +1,5 @@
 import pytest
 import torch
-import torch.nn as nn
 from zeta.nn.biases.relative_position_bias import RelativePositionBias
 
 
@@ -238,13 +237,13 @@ def test_different_bidirectional_bias_values():
 # Test case for initializing with negative max distance
 def test_negative_max_distance_init():
  with pytest.raises(ValueError):
- bias = RelativePositionBias(max_distance=-128)
+ RelativePositionBias(max_distance=-128)
 
 
 # Test case for initializing with negative num buckets
 def test_negative_num_buckets_init():
  with pytest.raises(ValueError):
- bias = RelativePositionBias(num_buckets=-32)
+ RelativePositionBias(num_buckets=-32)
 
 
 # Test case for initializing with a large max distance
@@ -280,4 +279,4 @@ def test_large_num_buckets():
 # Test case for bidirectional bias with negative max distance
 def test_bidirectional_bias_negative_max_distance():
  with pytest.raises(ValueError):
- bias = RelativePositionBias(bidirectional=True, max_distance=-128)
+ RelativePositionBias(bidirectional=True, max_distance=-128)

tests/nn/embeddings/test_QFTSPEmbeddings.py

@@ -69,18 +69,18 @@ def test_qftspembeddings_forward_negative_dim():
  vocab_size = 10000
  dim = -512
  with pytest.raises(ValueError):
- model = QFTSPEmbeddings(vocab_size, dim)
+ QFTSPEmbeddings(vocab_size, dim)
 
 
 def test_qftspembeddings_forward_negative_vocab_size():
  vocab_size = -10000
  dim = 512
  with pytest.raises(ValueError):
- model = QFTSPEmbeddings(vocab_size, dim)
+ QFTSPEmbeddings(vocab_size, dim)
 
 
 def test_qftspembeddings_forward_zero_vocab_size():
  vocab_size = 0
  dim = 512
  with pytest.raises(ValueError):
- model = QFTSPEmbeddings(vocab_size, dim)
+ QFTSPEmbeddings(vocab_size, dim)

tests/nn/embeddings/test_patch_embedding.py

@@ -1,4 +1,3 @@
-import pytest
 import torch
 from torch import nn
 from einops.layers.torch import Rearrange

tests/nn/embeddings/test_rope.py

@@ -1,6 +1,4 @@
-import pytest
 import torch
-from torch import nn
 
 from zeta.nn.embeddings.rope import (
  RotaryEmbedding,

tests/nn/embeddings/test_sine_positional_embs.py

@@ -1,6 +1,5 @@
 import pytest
 import torch
-from torch import nn
 from zeta.nn.embeddings.sine_positional import SinePositionalEmbedding
 
 
@@ -76,11 +75,11 @@ def test_extend_pe():
 def test_negative_dimension():
  dim_model = -512
  with pytest.raises(ValueError):
- module = SinePositionalEmbedding(dim_model)
+ SinePositionalEmbedding(dim_model)
 
 
 # Test case for initializing with alpha=True and dropout > 0
 def test_alpha_and_dropout():
  dim_model = 512
  with pytest.raises(ValueError):
- module = SinePositionalEmbedding(dim_model, alpha=True, dropout=0.2)
+ SinePositionalEmbedding(dim_model, alpha=True, dropout=0.2)

tests/nn/embeddings/test_truncated_rotary_emb.py

@@ -1,6 +1,4 @@
 import pytest
-import torch
-from torch import nn
 from zeta.nn.embeddings.truncated_rope import TruncatedRotaryEmbedding
 
 
@@ -50,7 +48,7 @@ def test_negative_dimension():
  b = 1.0
  rho = 0.0
  with pytest.raises(ValueError):
- module = TruncatedRotaryEmbedding(dim, a, b, rho)
+ TruncatedRotaryEmbedding(dim, a, b, rho)
 
 
 # Test case for initializing with a > b
@@ -60,7 +58,7 @@ def test_a_greater_than_b():
  b = 0.5
  rho = 0.0
  with pytest.raises(ValueError):
- module = TruncatedRotaryEmbedding(dim, a, b, rho)
+ TruncatedRotaryEmbedding(dim, a, b, rho)
 
 
 # Test case for initializing with rho > b
@@ -70,4 +68,4 @@ def test_rho_greater_than_b():
  b = 1.0
  rho = 1.5
  with pytest.raises(ValueError):
- module = TruncatedRotaryEmbedding(dim, a, b, rho)
+ TruncatedRotaryEmbedding(dim, a, b, rho)

tests/nn/embeddings/test_vision_embeddings.py

@@ -98,25 +98,25 @@ def test_forward_custom():
 # Test case for initializing with incorrect image size
 def test_incorrect_img_size_init():
  with pytest.raises(AssertionError):
- module = VisionEmbedding(img_size=256)
+ VisionEmbedding(img_size=256)
 
 
 # Test case for initializing with incorrect patch size
 def test_incorrect_patch_size_init():
  with pytest.raises(AssertionError):
- module = VisionEmbedding(patch_size=64)
+ VisionEmbedding(patch_size=64)
 
 
 # Test case for initializing with negative in_chans
 def test_negative_in_chans_init():
  with pytest.raises(ValueError):
- module = VisionEmbedding(in_chans=-3)
+ VisionEmbedding(in_chans=-3)
 
 
 # Test case for initializing with negative embed_dim
 def test_negative_embed_dim_init():
  with pytest.raises(ValueError):
- module = VisionEmbedding(embed_dim=-768)
+ VisionEmbedding(embed_dim=-768)
 
 
 # Test case for initializing with invalid masked_position

tests/nn/embeddings/test_vision_lang_embeddings.py

@@ -49,15 +49,15 @@ def test_incorrect_text_embedding_init():
  text_embed = nn.Linear(10, 10)
  vision_embed = nn.Embedding(10, 10)
  with pytest.raises(AssertionError):
- module = VisionLanguageEmbedding(text_embed, vision_embed)
+ VisionLanguageEmbedding(text_embed, vision_embed)
 
 
 # Test case for initializing with incorrect vision embedding
 def test_incorrect_vision_embedding_init():
  text_embed = nn.Embedding(10, 10)
  vision_embed = nn.Linear(10, 10)
  with pytest.raises(AssertionError):
- module = VisionLanguageEmbedding(text_embed, vision_embed)
+ VisionLanguageEmbedding(text_embed, vision_embed)
 
 
 # Test case for forward pass with text input being None

tests/nn/modules/test_cross_attn_images.py

@@ -1,6 +1,5 @@
 import torch
 import torch.nn as nn
-import numpy as np
 import pytest
 from torch.autograd import gradcheck
 from zeta.nn.attention.cross_attn_images import MultiModalCrossAttention

tests/nn/modules/test_custom_mlp.py

@@ -1,7 +1,6 @@
 import pytest
 import torch
 import torch.nn as nn
-import torch.nn.functional as F
 from zeta.nn.modules.flexible_mlp import CustomMLP
 
 

tests/nn/modules/test_hebbian.py

@@ -1,6 +1,5 @@
 import pytest
 import torch
-import torch.nn as nn
 
 from zeta.nn.modules.hebbian import (
  BasicHebbianGRUModel,

tests/nn/modules/test_image_projector.py

@@ -90,7 +90,7 @@ def test_patch_projector_performance(sample_input_tensor):
  # Measure the time taken for 100 forward passes
  start_time = time.time()
  for _ in range(100):
- output_tensor = patch_projector(input_tensor)
+ patch_projector(input_tensor)
  end_time = time.time()
 
  elapsed_time = end_time - start_time
@@ -211,7 +211,7 @@ def test_patch_projector_performance_various_input_sizes(
  # Measure the time taken for 100 forward passes
  start_time = time.time()
  for _ in range(100):
- output_tensor = patch_projector(input_tensor)
+ patch_projector(input_tensor)
  end_time = time.time()
 
  elapsed_time = end_time - start_time
@@ -249,15 +249,15 @@ def test_patch_projector_output_shape_consistency(sample_input_tensor):
 # Test case for edge case: invalid max_patch_size
 def test_patch_projector_invalid_max_patch_size():
  with pytest.raises(ValueError):
- patch_projector = ImagePatchCreatorProjector(
+ ImagePatchCreatorProjector(
  max_patch_size=0, embedding_dim=768
  )
 
 
 # Test case for edge case: invalid embedding_dim
 def test_patch_projector_invalid_embedding_dim():
  with pytest.raises(ValueError):
- patch_projector = ImagePatchCreatorProjector(
+ ImagePatchCreatorProjector(
  max_patch_size=16, embedding_dim=0
  )
 
@@ -270,7 +270,7 @@ def test_patch_projector_invalid_input_shape():
  input_tensor = torch.randn(1, 3, 32, 32) # Smaller image
 
  with pytest.raises(ValueError):
- output_tensor = patch_projector(input_tensor)
+ patch_projector(input_tensor)
 
 
 # Test case for dynamic patch size calculation

tests/nn/modules/test_log_ff.py

@@ -1,6 +1,6 @@
 import torch
 import pytest
-from zeta.nn.modules.log_ff import LogFF, compute_entropy_safe
+from zeta.nn.modules.log_ff import LogFF
 
 
 # Test fixture for a sample input tensor

tests/nn/modules/test_test_conv_lang.py

@@ -78,7 +78,7 @@ def test_with_mocked_convolution_layer():
  block = ConvolutionLanguageBlock(128, 256, 3, 1)
  block.conv_layers[0] = mock_convolution
  x = torch.randn(1, 128, 1024)
- output = block(x)
+ block(x)
  assert mock_convolution.called