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

7
tests/models/glm/test_modeling_glm.py
View File

@@ -88,7 +88,7 @@ class GlmIntegrationTest(unittest.TestCase):
]
model = AutoModelForCausalLM.from_pretrained(
self.model_id, low_cpu_mem_usage=True, torch_dtype=torch.float16, revision=self.revision
self.model_id, torch_dtype=torch.float16, revision=self.revision
).to(torch_device)
tokenizer = AutoTokenizer.from_pretrained(self.model_id, revision=self.revision)
@@ -106,7 +106,7 @@ class GlmIntegrationTest(unittest.TestCase):
]
model = AutoModelForCausalLM.from_pretrained(
self.model_id, low_cpu_mem_usage=True, torch_dtype=torch.bfloat16, revision=self.revision
self.model_id, torch_dtype=torch.bfloat16, revision=self.revision
).to(torch_device)
tokenizer = AutoTokenizer.from_pretrained(self.model_id, revision=self.revision)
@@ -125,7 +125,6 @@ class GlmIntegrationTest(unittest.TestCase):
model = AutoModelForCausalLM.from_pretrained(
self.model_id,
low_cpu_mem_usage=True,
torch_dtype=torch.bfloat16,
attn_implementation="eager",
revision=self.revision,
@@ -149,7 +148,6 @@ class GlmIntegrationTest(unittest.TestCase):
model = AutoModelForCausalLM.from_pretrained(
self.model_id,
low_cpu_mem_usage=True,
torch_dtype=torch.bfloat16,
attn_implementation="sdpa",
revision=self.revision,
@@ -174,7 +172,6 @@ class GlmIntegrationTest(unittest.TestCase):
model = AutoModelForCausalLM.from_pretrained(
self.model_id,
low_cpu_mem_usage=True,
torch_dtype=torch.bfloat16,
attn_implementation="flash_attention_2",
revision=self.revision,