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

[Tests] add min_new_tokens to prevent flaky length checks (#37175)

Browse Source
This commit is contained in:
Joao Gante
2025-04-02 15:24:00 +01:00
committed by GitHub
parent cbfa14823b
commit e90d55ebcc
1 changed files with 7 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
tests/generation/test_utils.py
View File

@@ -281,6 +281,7 @@ class GenerationTesterMixin:
do_sample=False,
num_beams=1,
max_new_tokens=self.max_new_tokens,
min_new_tokens=self.max_new_tokens,
output_attentions=output_attentions,
output_hidden_states=output_hidden_states,
output_scores=output_scores,
@@ -311,6 +312,7 @@ class GenerationTesterMixin:
do_sample=True,
num_beams=1,
max_new_tokens=self.max_new_tokens,
min_new_tokens=self.max_new_tokens,
num_return_sequences=num_return_sequences,
output_scores=output_scores,
output_logits=output_logits,
@@ -340,6 +342,7 @@ class GenerationTesterMixin:
output_generate = model.generate(
do_sample=False,
max_new_tokens=self.max_new_tokens,
min_new_tokens=self.max_new_tokens,
output_scores=output_scores,
output_logits=output_logits,
output_attentions=output_attentions,
@@ -370,6 +373,7 @@ class GenerationTesterMixin:
output_generate = model.generate(
do_sample=True,
max_new_tokens=self.max_new_tokens,
min_new_tokens=self.max_new_tokens,
output_scores=output_scores,
output_logits=output_logits,
output_attentions=output_attentions,
@@ -399,6 +403,7 @@ class GenerationTesterMixin:
output_generate = model.generate(
do_sample=False,
max_new_tokens=self.max_new_tokens,
min_new_tokens=self.max_new_tokens,
output_scores=output_scores,
output_logits=output_logits,
output_attentions=output_attentions,
@@ -429,6 +434,7 @@ class GenerationTesterMixin:
output_generate = model.generate(
do_sample=False,
max_new_tokens=self.max_new_tokens,
min_new_tokens=self.max_new_tokens,
output_scores=output_scores,
output_logits=output_logits,
output_attentions=output_attentions,
@@ -464,6 +470,7 @@ class GenerationTesterMixin:
do_sample=False,
num_beams=1,
max_new_tokens=self.max_new_tokens,
min_new_tokens=self.max_new_tokens,
output_attentions=output_attentions,
output_hidden_states=output_hidden_states,
output_scores=output_scores,