Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

#9370: removed ndpcc work around and debug code in sdpa decode and re-enabled CI #13299

Merged
merged 2 commits into from
Oct 1, 2024
Merged

#9370: removed ndpcc work around and debug code in sdpa decode and re-enabled CI #13299

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
5d9af21
#9370: removed ndpcc work around and debug code in sdpa decode and re…
caixunshiren Sep 30, 2024
eff9714
#9370: reduced the number of test cases in sdpa decode to speed up ci…
caixunshiren Sep 30, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
41 changes: 18 additions & 23 deletions ...tt_eager/python_api_testing/unit_testing/misc/test_scaled_dot_product_attention_decode.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -433,7 +433,6 @@ def run_test_sdpa_decode_single_iter(


@skip_for_grayskull("Unsupported in GS since L1 runs OOM with most configs")
@pytest.mark.skip("Skipping due to potential nd pcc issue #9370")
@pytest.mark.parametrize(
"dtype, q_dtype",
[
Expand All @@ -452,16 +451,16 @@ def run_test_sdpa_decode_single_iter(
@pytest.mark.parametrize(
"b, nh, nkv, s, d, grid_size, single_iter, cur_pos_tensor",
(
[32, 8, 1, 32768, 128, (8, 6), True, True], # Llama2-70B
[16, 8, 1, 32768, 128, (8, 6), False, False], # Llama2-70B
# [32, 8, 1, 32768, 128, (8, 6), True, True], # Llama2-70B
# [16, 8, 1, 32768, 128, (8, 6), False, False], # Llama2-70B
[8, 8, 1, 32768, 128, (8, 6), True, False], # Llama2-70B
[4, 8, 1, 32768, 128, (8, 6), True, False], # Llama2-70B
# [4, 8, 1, 32768, 128, (8, 6), True, False], # Llama2-70B
[32, 8, 1, 32768, 128, (8, 8), True, True], # Mixtral8x7b
[32, 8, 1, 32768, 128, (8, 6), True, False], # Llama2-70B
[4, 32, 8, 32768, 128, (8, 8), True, False], # llama 3.1 8b
# [32, 8, 1, 32768, 128, (8, 6), True, False], # Llama2-70B
# [4, 32, 8, 32768, 128, (8, 8), True, False], # llama 3.1 8b
[4, 32, 8, 32768, 128, (8, 8), True, True], # llama 3.1 8b
[4, 32, 8, 32768, 128, (8, 8), False, False], # llama 3.1 8b
[4, 16, 4, 32768, 128, (8, 8), False, False], # llama 3.1 8b
# [4, 32, 8, 32768, 128, (8, 8), False, False], # llama 3.1 8b
# [4, 16, 4, 32768, 128, (8, 8), False, False], # llama 3.1 8b
),
)
def test_sdpa_decode(
Expand All @@ -482,7 +481,6 @@ def test_sdpa_decode(


@skip_for_grayskull("Unsupported in GS since L1 runs OOM with most configs")
@pytest.mark.skip("Skipping due to potential nd pcc issue #9370")
@pytest.mark.parametrize(
"dtype, q_dtype",
[
Expand Down Expand Up @@ -704,7 +702,6 @@ def to_contiguous_cache(paged_cache, batch, num_kv, max_num_blocks_per_seq, bloc

@skip_for_blackhole("Unsupported on BH, see #12349")
@skip_for_grayskull("Unsupported in GS since L1 runs OOM with most configs")
# @pytest.mark.skip("Skipping due to potential nd pcc issue #9370")
@pytest.mark.parametrize(
"kv_dtype, q_dtype",
[
Expand Down Expand Up @@ -756,7 +753,6 @@ def test_sdpa_decode_paged_attention(


@skip_for_grayskull("Unsupported in GS since L1 runs OOM with most configs")
@pytest.mark.skip("Skipping due to potential nd pcc issue #9370")
@pytest.mark.parametrize(
"dtype, q_dtype",
[
Expand Down Expand Up @@ -840,7 +836,6 @@ def test_sdpa_decode_perf(device, use_program_cache):


@skip_for_grayskull("Unsupported in GS since L1 runs OOM with most configs")
@pytest.mark.skip("Skipping due to potential nd pcc issue #9370")
@pytest.mark.parametrize(
"dtype",
[ttnn.bfloat8_b, ttnn.bfloat16],
Expand Down Expand Up @@ -962,8 +957,8 @@ def run_test_sdpa_decode_ndpcc(device, b, nh, nkv, s, d, dtype, grid_size, q_dty
)
dram_memcfg = ttnn.DRAM_MEMORY_CONFIG

K = fa_rand(nkv, b, s, d)
V = fa_rand(nkv, b, s, d)
K = fa_rand(b, nkv, s, d)
V = fa_rand(b, nkv, s, d)

tt_K = ttnn.as_tensor(K, device=device, dtype=dtype, layout=ttnn.TILE_LAYOUT, memory_config=dram_memcfg)
tt_V = ttnn.as_tensor(V, device=device, dtype=dtype, layout=ttnn.TILE_LAYOUT, memory_config=dram_memcfg)
Expand Down Expand Up @@ -998,8 +993,8 @@ def run_test_sdpa_decode_ndpcc(device, b, nh, nkv, s, d, dtype, grid_size, q_dty
attn_mask[:, :, :, start_idx + 1 :] = torch.finfo(torch.float32).min

Q_slice = Q[:, :, :nh, :].permute(1, 2, 0, 3) # b, nh, 1, d
K_slice = K[:, :, :padded_layer_len, :].permute(1, 0, 2, 3) # nh, b, S, d
V_slice = V[:, :, :padded_layer_len, :].permute(1, 0, 2, 3) # nh, b, S, d
K_slice = K[:, :, :padded_layer_len, :]
V_slice = V[:, :, :padded_layer_len, :]
attn_mask_slice = attn_mask[:, :, :nh, :].permute(1, 2, 0, 3) # b, nh, 1, S

expect = torch.nn.functional.scaled_dot_product_attention(
Expand All @@ -1009,7 +1004,7 @@ def run_test_sdpa_decode_ndpcc(device, b, nh, nkv, s, d, dtype, grid_size, q_dty

all_out_pass = True

for i in range(200):
for i in range(500):
tt_Q = ttnn.as_tensor(
Q[:, :, :nh],
device=device,
Expand Down Expand Up @@ -1049,9 +1044,9 @@ def run_test_sdpa_decode_ndpcc(device, b, nh, nkv, s, d, dtype, grid_size, q_dty
if not all_out_pass:
failed_start_pos.append(start_idx)

start_idx += 20 # if start_idx < 4096 else 3001
start_idx += 200 # if start_idx < 4096 else 3001

logger.info(f"ND Start Pos: {failed_start_pos}")
logger.info(f"PCC failed Start Pos: {failed_start_pos}")


@pytest.mark.timeout(600)
Expand All @@ -1060,13 +1055,13 @@ def run_test_sdpa_decode_ndpcc(device, b, nh, nkv, s, d, dtype, grid_size, q_dty
@pytest.mark.parametrize(
"dtype, q_dtype",
[
# [ttnn.bfloat16, ttnn.bfloat16],
# [ttnn.bfloat8_b, ttnn.bfloat8_b],
[ttnn.bfloat16, ttnn.bfloat16],
[ttnn.bfloat8_b, ttnn.bfloat8_b],
[ttnn.bfloat4_b, ttnn.bfloat4_b],
],
ids=[
# "bfp16_bfp16",
# "bfp8_bfp8",
"bfp16_bfp16",
"bfp8_bfp8",
"bfp4_bfp4",
],
)
Expand Down
2 changes: 0 additions & 2 deletions ...ager/python_api_testing/unit_testing/misc/test_scaled_dot_product_attention_decode_gqa.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -222,7 +222,6 @@ def run_test_sdpa_decode_single_iter(


@skip_for_grayskull("Unsupported in GS since L1 runs OOM with most configs")
@pytest.mark.skip("Skipping due to potential nd pcc issue #9370")
@pytest.mark.parametrize(
"dtype, q_dtype",
[
Expand Down Expand Up @@ -287,7 +286,6 @@ def test_sdpa_decode(


@skip_for_grayskull("Unsupported in GS since L1 runs OOM with most configs")
@pytest.mark.skip("Skipping due to potential nd pcc issue #9370")
@pytest.mark.parametrize(
"dtype",
[
Expand Down
Loading