tocmo0nlord/axolotl
1
0
Fork 0
Code Issues Pull Requests Actions 3 Packages Projects Releases Wiki Activity

attempt to also run e2e tests that needs gpus (#1070)

Browse Source 
* attempt to also run e2e tests that needs gpus

* fix stray quote

* checkout specific github ref

* dockerfile for tests with proper checkout

ensure wandb is dissabled for docker pytests
clear wandb env after testing
clear wandb env after testing
make sure to provide a default val for pop
tryin skipping wandb validation tests
explicitly disable wandb in the e2e tests
explicitly report_to None to see if that fixes the docker e2e tests
split gpu from non-gpu unit tests
skip bf16 check in test for now
build docker w/o cache since it uses branch name ref
revert some changes now that caching is fixed
skip bf16 check if on gpu w support

* pytest skip for auto-gptq requirements

* skip mamba tests for now, split multipack and non packed lora llama tests

* split tests that use monkeypatches

* fix relative import for prev commit

* move other tests using monkeypatches to the correct run
This commit is contained in:
Wing Lian
2024-01-09 21:23:23 -05:00
committed by GitHub
parent 9be92d1448
commit 788649fe95
13 changed files with 214 additions and 105 deletions
Download Patch File Download Diff File Expand all files Collapse all files

99
tests/e2e/patched/test_model_patches.py Normal file
View File

@@ -0,0 +1,99 @@
"""
E2E smoke tests to check that the monkeypatches are in place for certain configurations
"""
import unittest
from axolotl.common.cli import TrainerCliArgs
from axolotl.utils.config import normalize_config
from axolotl.utils.dict import DictDefault
from axolotl.utils.models import load_model, load_tokenizer
from ..utils import with_temp_dir
class TestModelPatches(unittest.TestCase):
"""
TestCases for the multipack monkey patches
"""
@with_temp_dir
def test_mixtral_multipack(self, temp_dir):
cfg = DictDefault(
{
"base_model": "hf-internal-testing/Mixtral-tiny",
"tokenizer_config": "mistralai/Mixtral-8x7B-v0.1",
"flash_attention": True,
"sample_packing": True,
"sequence_len": 2048,
"val_set_size": 0.1,
"special_tokens": {},
"datasets": [
{
"path": "mhenrichsen/alpaca_2k_test",
"type": "alpaca",
},
],
"num_epochs": 2,
"micro_batch_size": 2,
"gradient_accumulation_steps": 1,
"output_dir": temp_dir,
"learning_rate": 0.00001,
"optimizer": "adamw_bnb_8bit",
"lr_scheduler": "cosine",
"max_steps": 20,
"save_steps": 10,
"eval_steps": 10,
}
)
normalize_config(cfg)
cli_args = TrainerCliArgs()
tokenizer = load_tokenizer(cfg)
model, _ = load_model(cfg, tokenizer, inference=cli_args.inference)
assert (
"axolotl.monkeypatch.mixtral.modeling_mixtral"
in model.model.layers[0].self_attn.__class__.__module__
)
assert (
"MixtralMultipackFlashAttention2"
in model.model.layers[0].self_attn.__class__.__name__
)
@with_temp_dir
def test_mistral_multipack(self, temp_dir):
cfg = DictDefault(
{
"base_model": "openaccess-ai-collective/tiny-mistral",
"flash_attention": True,
"sample_packing": True,
"sequence_len": 2048,
"val_set_size": 0.1,
"special_tokens": {},
"datasets": [
{
"path": "mhenrichsen/alpaca_2k_test",
"type": "alpaca",
},
],
"num_epochs": 2,
"micro_batch_size": 2,
"gradient_accumulation_steps": 1,
"output_dir": temp_dir,
"learning_rate": 0.00001,
"optimizer": "adamw_bnb_8bit",
"lr_scheduler": "cosine",
"max_steps": 20,
"save_steps": 10,
"eval_steps": 10,
}
)
normalize_config(cfg)
cli_args = TrainerCliArgs()
tokenizer = load_tokenizer(cfg)
model, _ = load_model(cfg, tokenizer, inference=cli_args.inference)
assert (
"axolotl.monkeypatch.mistral_attn_hijack_flash"
in model.model.layers[0].self_attn.forward.__module__
)