#13685: Update backward embedding document (#14666)

### Ticket
#13685 

### Problem description
- ttnn.embedding_bw doesn't support output and input gradient tensors
having different dtypes.
- ttnn.embedding_bw is not an eltwise op. So, moving corresponding tests
outside eltwise folder.

### What's changed
- Updated document to add a note mentioning input and output gradients
must have the same datatype.
- Moved sweep test from
`tests/sweep_framework/sweeps/eltwise/binary_backward/embedding_bw/embedding_bw.py`
to `tests/sweep_framework/sweeps/embedding_bw/embedding_bw.py`
- Move unit test from from
`tests/ttnn/unit_tests/operations/eltwise/backward/test_backward_embedding.py`
to `tests/ttnn/unit_tests/operations/test_backward_embedding.py`


### Updated doc
<img width="1139" alt="Screenshot 2024-11-04 at 17 49 31"
src="https://github.com/user-attachments/assets/467fec4f-cb30-4ec9-ac26-9e59827ed45b">


### Checklist
- [x] [Post commit CI
passes](https://github.com/tenstorrent/tt-metal/actions/runs/11680989159)
- PASSED
- [x] [Run
Sweeps](https://github.com/tenstorrent/tt-metal/actions/runs/11678538301)
- PASSED
- [x] [Night fast dispatch
tests](https://github.com/tenstorrent/tt-metal/actions/runs/11678522765)
(same as main)

.github/workflows/ttnn-run-sweeps.yaml

@@ -235,7 +235,6 @@ on:
           - eltwise.binary_backward.ldexp_bw
           - eltwise.binary_backward.logaddexp_bw
           - eltwise.binary_backward.logaddexp2_bw
-          - eltwise.binary_backward.embedding_bw.embedding_bw
           - eltwise.binary_backward.addalpha_bw.addalpha_bw
           - eltwise.binary_backward.subalpha_bw.subalpha_bw
           - eltwise.binary_backward.xlogy_bw.xlogy_bw
@@ -266,6 +265,7 @@ on:
           - eltwise.ternary_backward.addcmul_bw
           - eltwise.ternary_backward.addcdiv_bw
           - embedding.embedding
+          - embedding_bw.embedding_bw
           - reduction.backward.prod_bw.prod_bw
           - reduction.topk.topk
           - reduction.argmax.argmax

tests/sweep_framework/sweeps/eltwise/binary_backward/embedding_bw/embedding_bw.py → tests/sweep_framework/sweeps/embedding_bw/embedding_bw.py

@@ -6,19 +6,13 @@
 from functools import partial
 
 import torch
-import random
 import ttnn
 from tests.sweep_framework.sweep_utils.utils import gen_shapes
 from tests.tt_eager.python_api_testing.sweep_tests.generation_funcs import gen_func_with_cast_tt
 
 from tests.ttnn.utils_for_testing import check_with_pcc, start_measuring_time, stop_measuring_time
 from models.utility_functions import torch_random
 
-# Override the default timeout in seconds for hang detection.
-TIMEOUT = 30
-
-random.seed(0)
-
 
 # Parameters provided to the test vector generator are defined here.
 # They are defined as dict-type suites that contain the arguments to the run function as keys, and lists of possible inputs as values.
@@ -91,8 +85,7 @@ def run(
     *,
     device,
 ) -> list:
-    data_seed = random.randint(0, 20000000)
-    torch.manual_seed(data_seed)
+    torch.manual_seed(0)
 
     batch_size, seq_length, embeddings_dim, num_embeddings = embedding_args
 

ttnn/cpp/ttnn/operations/embedding_backward/embedding_backward_pybind.cpp

@@ -35,6 +35,8 @@ void py_bind_embedding_backward(py::module& module) {
         Returns:
             ttnn.Tensor: the output tensor.
 
+        Note:
+            The input and the output gradient tensors must have the same datatype.
 
         Example:
             >>> device_id = 0