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

optim e2e tests to run a bit faster #2069

Merged
merged 3 commits into from
Nov 18, 2024
Merged

optim e2e tests to run a bit faster #2069

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
7e72a43
optim e2e tests to run a bit faster
winglian Nov 16, 2024
f5a25a3
run prequant w/o lora_modules_to_save
winglian Nov 16, 2024
30b7701
use smollm2
winglian Nov 18, 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
19 changes: 19 additions & 0 deletions tests/e2e/conftest.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,25 @@
import tempfile

import pytest
from huggingface_hub import snapshot_download


@pytest.fixture(scope="session", autouse=True)
def download_smollm2_135m_model():
# download the model
snapshot_download("HuggingFaceTB/SmolLM2-135M")


@pytest.fixture(scope="session", autouse=True)
def download_tatsu_lab_alpaca_dataset():
# download the model
snapshot_download("tatsu-lab/alpaca", repo_type="dataset")


@pytest.fixture(scope="session", autouse=True)
def download_mhenrichsen_alpaca_2k_dataset():
# download the model
snapshot_download("mhenrichsen/alpaca_2k_test", repo_type="dataset")


@pytest.fixture
Expand Down
38 changes: 17 additions & 21 deletions tests/e2e/multigpu/test_llama.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,7 +25,7 @@
@pytest.fixture(scope="session", autouse=True)
def download_model():
# download the model
snapshot_download("TinyLlama/TinyLlama_v1.1")
snapshot_download("HuggingFaceTB/SmolLM2-135M")


class TestMultiGPULlama:
Expand All @@ -37,7 +37,7 @@ def test_lora_ddp(self, temp_dir):
# pylint: disable=duplicate-code
cfg = DictDefault(
{
"base_model": "HuggingFaceTB/SmolLM-135M",
"base_model": "HuggingFaceTB/SmolLM2-135M",
"sequence_len": 2048,
"adapter": "lora",
"lora_r": 8,
Expand Down Expand Up @@ -93,7 +93,7 @@ def test_lora_ddp_packed(self, temp_dir, gradient_accumulation_steps):
# pylint: disable=duplicate-code
cfg = DictDefault(
{
"base_model": "HuggingFaceTB/SmolLM-135M",
"base_model": "HuggingFaceTB/SmolLM2-135M",
"sequence_len": 2048,
"sample_packing": True,
"eval_sample_packing": False,
Expand Down Expand Up @@ -149,8 +149,7 @@ def test_dpo_lora_ddp(self, temp_dir):
# pylint: disable=duplicate-code
cfg = DictDefault(
{
"base_model": "TinyLlama/TinyLlama_v1.1",
"tokenizer_type": "LlamaTokenizer",
"base_model": "HuggingFaceTB/SmolLM2-135M",
"sequence_len": 2048,
"sample_packing": False,
"eval_sample_packing": False,
Expand All @@ -163,12 +162,10 @@ def test_dpo_lora_ddp(self, temp_dir):
"lora_target_linear": True,
"val_set_size": 0.05,
"special_tokens": {
"unk_token": "<unk>",
"bos_token": "<s>",
"eos_token": "</s>",
"pad_token": "<|endoftext|>",
},
"rl": "dpo",
"chat_template": "llama3",
"chat_template": "chatml",
"datasets": [
{
"path": "fozziethebeat/alpaca_messages_2k_dpo_test",
Expand Down Expand Up @@ -221,7 +218,7 @@ def test_dpo_qlora_ddp(self, temp_dir):
# pylint: disable=duplicate-code
cfg = DictDefault(
{
"base_model": "HuggingFaceTB/SmolLM-135M",
"base_model": "HuggingFaceTB/SmolLM2-135M",
"sequence_len": 2048,
"sample_packing": False,
"eval_sample_packing": False,
Expand Down Expand Up @@ -294,7 +291,7 @@ def test_fsdp(self, temp_dir, gradient_accumulation_steps):
# pylint: disable=duplicate-code
cfg = DictDefault(
{
"base_model": "HuggingFaceTB/SmolLM-135M",
"base_model": "HuggingFaceTB/SmolLM2-135M",
"sequence_len": 2048,
"val_set_size": 0.01,
"special_tokens": {
Expand Down Expand Up @@ -359,7 +356,7 @@ def test_fsdp_packed(self, temp_dir, fsdp_state_dict_type):
# pylint: disable=duplicate-code
cfg = DictDefault(
{
"base_model": "HuggingFaceTB/SmolLM-135M",
"base_model": "HuggingFaceTB/SmolLM2-135M",
"sample_packing": True,
"pad_to_sequence_len": True,
"sequence_len": 2048,
Expand Down Expand Up @@ -422,26 +419,25 @@ def test_fsdp_qlora_prequant_packed(self, temp_dir):
# pylint: disable=duplicate-code
cfg = DictDefault(
{
"base_model": "axolotl-ai-co/TinyLlama_v1.1-bnb-nf4-bf16",
"tokenizer_type": "AutoTokenizer",
"base_model": "axolotl-ai-co/SmolLM2-135M-bnb-nf4-bf16",
"adapter": "qlora",
"mean_resizing_embeddings": True,
"load_in_4bit": True,
"lora_r": 8,
"lora_alpha": 16,
"lora_dropout": 0.05,
"lora_target_linear": True,
"lora_modules_to_save": [
"embed_tokens",
"lm_head",
],
# "lora_modules_to_save": [
# "embed_tokens",
# "lm_head",
# ],
"sample_packing": True,
"eval_sample_packing": False,
"pad_to_sequence_len": True,
"sequence_len": 2048,
"val_set_size": 0.05,
"special_tokens": {
"pad_token": "</s>",
"pad_token": "<|endoftext|>",
},
"datasets": [
{
Expand Down Expand Up @@ -503,7 +499,7 @@ def test_ds_zero3_packed(self, temp_dir, gradient_accumulation_steps):
# pylint: disable=duplicate-code
cfg = DictDefault(
{
"base_model": "HuggingFaceTB/SmolLM-135M",
"base_model": "HuggingFaceTB/SmolLM2-135M",
"sample_packing": True,
"pad_to_sequence_len": True,
"sequence_len": 2048,
Expand Down Expand Up @@ -553,7 +549,7 @@ def test_ds_zero3_qlora_packed(self, temp_dir):
# pylint: disable=duplicate-code
cfg = DictDefault(
{
"base_model": "HuggingFaceTB/SmolLM-135M",
"base_model": "HuggingFaceTB/SmolLM2-135M",
"load_in_4bit": True,
"adapter": "qlora",
"lora_r": 8,
Expand Down
2 changes: 2 additions & 0 deletions tests/e2e/patched/test_fa_xentropy.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -66,6 +66,8 @@ def test_lora_packing_fa_cross_entropy(self, temp_dir):
},
],
"num_epochs": 1,
"max_steps": 10,
"save_steps": 10,
"micro_batch_size": 8,
"gradient_accumulation_steps": 1,
"output_dir": temp_dir,
Expand Down
3 changes: 3 additions & 0 deletions tests/e2e/patched/test_lora_llama_multipack.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -56,6 +56,8 @@ def test_lora_packing(self, temp_dir):
},
],
"num_epochs": 2,
"max_steps": 20,
"save_steps": 10,
"micro_batch_size": 8,
"gradient_accumulation_steps": 1,
"output_dir": temp_dir,
Expand Down Expand Up @@ -109,6 +111,7 @@ def test_lora_gptq_packed(self, temp_dir):
},
],
"num_epochs": 2,
"max_steps": 20,
"save_steps": 0.5,
"micro_batch_size": 8,
"gradient_accumulation_steps": 1,
Expand Down
2 changes: 1 addition & 1 deletion tests/e2e/test_optimizers.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -113,7 +113,7 @@ def test_adopt_adamw(self, temp_dir):
def test_fft_schedule_free_adamw(self, temp_dir):
cfg = DictDefault(
{
"base_model": "HuggingFaceTB/SmolLM-135M",
"base_model": "HuggingFaceTB/SmolLM2-135M",
"sequence_len": 1024,
"val_set_size": 0.1,
"special_tokens": {
Expand Down
2 changes: 1 addition & 1 deletion tests/e2e/test_packing_loss.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -31,7 +31,7 @@ def test_loss_packed(self, temp_dir):
# pylint: disable=duplicate-code
cfg = DictDefault(
{
"base_model": "HuggingFaceTB/SmolLM-135M",
"base_model": "HuggingFaceTB/SmolLM2-135M",
"sequence_len": 1024,
"sample_packing": True,
"flash_attention": True,
Expand Down
Loading