biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[V1] V1 Enablement Oracle (#13726)

Browse Source 
Signed-off-by: rshaw@neuralmagic.com <rshaw@neuralmagic.com>
Co-authored-by: rshaw@neuralmagic.com <rshaw@neuralmagic.com>
Co-authored-by: Nicolò Lucchesi <nlucches@redhat.com>
Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com>
Co-authored-by: Michael Goin <michael@neuralmagic.com>
This commit is contained in:
Robert Shaw
2025-03-15 01:02:20 -04:00
committed by GitHub
parent 8c0d15d5c5
commit d4d93db2c5
96 changed files with 1537 additions and 512 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
tests/models/decoder_only/language/test_gguf.py
View File

@@ -110,16 +110,6 @@ def test_models(
example_prompts = tokenizer.apply_chat_template(
messages, tokenize=False, add_generation_prompt=True)
# Run unquantized model.
with vllm_runner(
model_name=model.original_model,
enforce_eager=True, # faster tests
dtype=dtype,
max_model_len=MAX_MODEL_LEN,
tensor_parallel_size=tp_size) as original_model:
original_outputs = original_model.generate_greedy_logprobs(
example_prompts[:-1], max_tokens, num_logprobs)
# Run gguf model.
with vllm_runner(model_name=model.gguf_model,
enforce_eager=True,
@@ -130,6 +120,16 @@ def test_models(
gguf_outputs = gguf_model.generate_greedy_logprobs(
example_prompts[:-1], max_tokens, num_logprobs)
# Run unquantized model.
with vllm_runner(
model_name=model.original_model,
enforce_eager=True, # faster tests
dtype=dtype,
max_model_len=MAX_MODEL_LEN,
tensor_parallel_size=tp_size) as original_model:
original_outputs = original_model.generate_greedy_logprobs(
example_prompts[:-1], max_tokens, num_logprobs)
check_logprobs_close(
outputs_0_lst=original_outputs,
outputs_1_lst=gguf_outputs,