Skip to content

Commit

Permalink
progress commit
Browse files Browse the repository at this point in the history
  • Loading branch information
NishantBaheti committed Jan 1, 2025
1 parent 1ee339e commit 618e32b
Showing 1 changed file with 27 additions and 22 deletions.
49 changes: 27 additions & 22 deletions notebooks/transformers/explore.ipynb
Original file line number Diff line number Diff line change
Expand Up @@ -2598,33 +2598,38 @@
"\n",
"```sh\n",
"\n",
" input\n",
" |\n",
" |\n",
" |\\\n",
" | \\\n",
" | \\ \n",
" | BN\n",
" | |\n",
" | ReLU\n",
" | |\n",
" | Weight\n",
" | |\n",
" | BN\n",
" | |\n",
" | ReLU\n",
" | /\n",
" | weight\n",
" | /\n",
" ADD\n",
" |\n",
" |\n",
" Output\n",
" input\n",
" |\n",
" o Fork\n",
" |\\\n",
" | | \\ A\n",
" | | \\ |\n",
" | | BN |\n",
"Forward | | | | Backward\n",
" | | ReLU |\n",
" | | | |\n",
" V | Weight\n",
" | |\n",
" | BN\n",
" | |\n",
" | ReLU\n",
" | /\n",
" | weight\n",
" | /\n",
" addition\n",
" |\n",
" |\n",
" Output\n",
"```\n",
"\n",
"Because of skip connection the gradients can travel faster to initial layers and initial layers can learn as fast as final layers, This helps when we are building very Deep Neural Networks"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": []
},
{
"cell_type": "code",
"execution_count": null,
Expand Down

0 comments on commit 618e32b

Please sign in to comment.