SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

Remove all traces of low_cpu_mem_usage (#38792)

Browse Source 
* remove it from all py files

* remove it from the doc

* remove it from examples

* style

* remove traces of _fast_init

* Update test_peft_integration.py

* CIs
This commit is contained in:
Cyril Vallez
2025-06-12 16:39:33 +02:00
committed by GitHub
parent 3542e0b844
commit 4b8ec667e9
76 changed files with 100 additions and 598 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
tests/models/gemma2/test_modeling_gemma2.py
View File

@@ -197,7 +197,7 @@ class Gemma2IntegrationTest(unittest.TestCase):
]
model = AutoModelForCausalLM.from_pretrained(
model_id, low_cpu_mem_usage=True, torch_dtype=torch.bfloat16, attn_implementation="eager"
model_id, torch_dtype=torch.bfloat16, attn_implementation="eager"
).to(torch_device)
tokenizer = AutoTokenizer.from_pretrained(model_id)
@@ -218,7 +218,7 @@ class Gemma2IntegrationTest(unittest.TestCase):
]
model = AutoModelForCausalLM.from_pretrained(
model_id, low_cpu_mem_usage=True, torch_dtype=torch.float16, attn_implementation="eager"
model_id, torch_dtype=torch.float16, attn_implementation="eager"
).to(torch_device)
tokenizer = AutoTokenizer.from_pretrained(model_id)
@@ -241,7 +241,7 @@ class Gemma2IntegrationTest(unittest.TestCase):
]
model = AutoModelForCausalLM.from_pretrained(
model_id, low_cpu_mem_usage=True, torch_dtype=torch.bfloat16, attn_implementation="flex_attention"
model_id, torch_dtype=torch.bfloat16, attn_implementation="flex_attention"
).to(torch_device)
tokenizer = AutoTokenizer.from_pretrained(model_id)
pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)
@@ -271,7 +271,7 @@ class Gemma2IntegrationTest(unittest.TestCase):
EXPECTED_BATCH_TEXT = EXPECTED_BATCH_TEXTS.get_expectation()
model = AutoModelForCausalLM.from_pretrained(
model_id, low_cpu_mem_usage=True, torch_dtype=torch.bfloat16, attn_implementation="flex_attention"
model_id, torch_dtype=torch.bfloat16, attn_implementation="flex_attention"
).to(torch_device)
tokenizer = AutoTokenizer.from_pretrained(model_id)
pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)
@@ -419,7 +419,7 @@ class Gemma2IntegrationTest(unittest.TestCase):
]
model = AutoModelForCausalLM.from_pretrained(
model_id, low_cpu_mem_usage=True, torch_dtype=torch.bfloat16, attn_implementation="flex_attention"
model_id, torch_dtype=torch.bfloat16, attn_implementation="flex_attention"
).to(torch_device)
assert model.config._attn_implementation == "flex_attention"
tokenizer = AutoTokenizer.from_pretrained(model_id)