tocmo0nlord/axolotl
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

bump transformers==4.52.4 (#2800) [skip ci]

Browse Source 
* bump transformers==4.52.4

* don't use hf offline for qwen tokenizer

* increase timeout

* don't use methodtype

* increase timeout

* better assertion logging

* upgrade deepspeed version too
This commit is contained in:
Wing Lian
2025-06-18 15:46:14 -04:00
committed by GitHub
parent 06a648263b
commit a85efffbef
18 changed files with 34 additions and 35 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
tests/e2e/multigpu/patched/test_sp.py
View File

@@ -91,7 +91,10 @@ class TestSequenceParallelism:
)
check_tensorboard(
temp_dir + "/runs", "train/train_loss", threshold, "Train Loss is too high"
temp_dir + "/runs",
"train/train_loss",
threshold,
"Train Loss (%s) is too high",
)
@pytest.mark.parametrize(

2
tests/e2e/multigpu/solo/test_flex.py
View File

@@ -85,5 +85,5 @@ class TestPackedFlex:
)
check_tensorboard(
temp_dir + "/runs", "train/train_loss", 2.0, "Train Loss is too high"
temp_dir + "/runs", "train/train_loss", 2.0, "Train Loss (%s) is too high"
)

2
tests/e2e/multigpu/test_gemma3.py
View File

@@ -91,5 +91,5 @@ class TestMultiGPUGemma3:
)
check_tensorboard(
temp_dir + "/runs", "train/train_loss", 1.8, "Train Loss is too high"
temp_dir + "/runs", "train/train_loss", 1.8, "Train Loss (%s) is too high"
)

24
tests/e2e/multigpu/test_llama.py
View File

@@ -89,7 +89,7 @@ class TestMultiGPULlama:
)
check_tensorboard(
temp_dir + "/runs", "train/train_loss", 2.3, "Train Loss is too high"
temp_dir + "/runs", "train/train_loss", 2.3, "Train Loss (%s) is too high"
)
@pytest.mark.parametrize(
@@ -154,7 +154,7 @@ class TestMultiGPULlama:
)
check_tensorboard(
temp_dir + "/runs", "train/train_loss", 2.3, "Train Loss is too high"
temp_dir + "/runs", "train/train_loss", 2.3, "Train Loss (%s) is too high"
)
def test_dpo_lora_ddp(self, temp_dir):
@@ -232,7 +232,7 @@ class TestMultiGPULlama:
temp_dir + "/runs",
"train/train_loss",
loss_threshold,
"Train Loss is too high",
"Train Loss (%s) is too high",
)
def test_dpo_qlora_ddp(self, temp_dir):
@@ -310,7 +310,7 @@ class TestMultiGPULlama:
temp_dir + "/runs",
"train/train_loss",
loss_threshold,
"Train Loss is too high",
"Train Loss (%s) is too high",
)
@pytest.mark.parametrize(
@@ -380,7 +380,7 @@ class TestMultiGPULlama:
)
check_tensorboard(
temp_dir + "/runs", "train/train_loss", 2.3, "Train Loss is too high"
temp_dir + "/runs", "train/train_loss", 2.3, "Train Loss (%s) is too high"
)
@pytest.mark.parametrize(
@@ -452,7 +452,7 @@ class TestMultiGPULlama:
)
check_tensorboard(
temp_dir + "/runs", "train/train_loss", 2.3, "Train Loss is too high"
temp_dir + "/runs", "train/train_loss", 2.3, "Train Loss (%s) is too high"
)
@require_torch_2_6_0
@@ -533,7 +533,7 @@ class TestMultiGPULlama:
)
check_tensorboard(
temp_dir + "/runs", "train/train_loss", 2.1, "Train Loss is too high"
temp_dir + "/runs", "train/train_loss", 2.1, "Train Loss (%s) is too high"
)
def test_fsdp_qlora_prequant_packed(self, temp_dir):
@@ -613,7 +613,7 @@ class TestMultiGPULlama:
)
check_tensorboard(
temp_dir + "/runs", "train/train_loss", 2.3, "Train Loss is too high"
temp_dir + "/runs", "train/train_loss", 2.3, "Train Loss (%s) is too high"
)
@pytest.mark.parametrize(
@@ -697,7 +697,7 @@ class TestMultiGPULlama:
)
check_tensorboard(
temp_dir + "/runs", "train/train_loss", 2.3, "Train Loss is too high"
temp_dir + "/runs", "train/train_loss", 2.4, "Train Loss (%s) is too high"
)
@pytest.mark.parametrize(
@@ -771,7 +771,7 @@ class TestMultiGPULlama:
)
check_tensorboard(
temp_dir + "/runs", "train/train_loss", 2.3, "Train Loss is too high"
temp_dir + "/runs", "train/train_loss", 2.3, "Train Loss (%s) is too high"
)
@pytest.mark.parametrize(
@@ -845,7 +845,7 @@ class TestMultiGPULlama:
)
check_tensorboard(
temp_dir + "/runs", "train/train_loss", 2.3, "Train Loss is too high"
temp_dir + "/runs", "train/train_loss", 2.3, "Train Loss (%s) is too high"
)
@pytest.mark.skip(
@@ -912,5 +912,5 @@ class TestMultiGPULlama:
)
check_tensorboard(
temp_dir + "/runs", "train/train_loss", 4.0, "Train Loss is too high"
temp_dir + "/runs", "train/train_loss", 4.0, "Train Loss (%s) is too high"
)

4
tests/e2e/multigpu/test_ray.py
View File

@@ -75,7 +75,7 @@ class TestMultiGPURay:
)
check_tensorboard(
temp_dir + "/runs", "train/train_loss", 2.3, "Train Loss is too high"
temp_dir + "/runs", "train/train_loss", 2.3, "Train Loss (%s) is too high"
)
@require_torch_lt_2_6_0
@@ -133,5 +133,5 @@ class TestMultiGPURay:
)
check_tensorboard(
temp_dir + "/runs", "train/train_loss", 2.3, "Train Loss is too high"
temp_dir + "/runs", "train/train_loss", 2.3, "Train Loss (%s) is too high"
)