[Core][Model] Terratorch backend integration (#23513)
Signed-off-by: Michele Gazzetti <michele.gazzetti1@ibm.com> Signed-off-by: Christian Pinto <christian.pinto@ibm.com> Co-authored-by: Christian Pinto <christian.pinto@ibm.com> Co-authored-by: Cyrus Leung <tlleungac@connect.ust.hk>
This commit is contained in:
co-authored by
Christian Pinto
Cyrus Leung
parent
e7fc70016f
commit
51d5e9be7d
@@ -6,10 +6,11 @@ from dataclasses import dataclass, field
|
||||
from typing import Any, Literal, Optional
|
||||
|
||||
import pytest
|
||||
import torch
|
||||
from packaging.version import Version
|
||||
from transformers import __version__ as TRANSFORMERS_VERSION
|
||||
|
||||
from vllm.config import TokenizerMode
|
||||
from vllm.config import ModelDType, TokenizerMode
|
||||
|
||||
|
||||
@dataclass(frozen=True)
|
||||
@@ -47,6 +48,23 @@ class _HfExamplesInfo:
|
||||
The reason for the minimum/maximum version requirement.
|
||||
"""
|
||||
|
||||
skip_tokenizer_init: bool = False
|
||||
"""
|
||||
If true, skip initialization of tokenizer and detokenizer.
|
||||
"""
|
||||
|
||||
dtype: ModelDType = "auto"
|
||||
"""
|
||||
The data type for the model weights and activations.
|
||||
"""
|
||||
|
||||
enforce_eager: bool = False
|
||||
"""
|
||||
Whether to enforce eager execution. If True, we will
|
||||
disable CUDA graph and always execute the model in eager mode.
|
||||
If False, we will use CUDA graph and eager execution in hybrid.
|
||||
"""
|
||||
|
||||
is_available_online: bool = True
|
||||
"""
|
||||
Set this to ``False`` if the name of this architecture no longer exists on
|
||||
@@ -76,6 +94,9 @@ class _HfExamplesInfo:
|
||||
If not specified, the default revision will be used.
|
||||
"""
|
||||
|
||||
max_num_seqs: Optional[int] = None
|
||||
"""Maximum number of sequences to be processed in a single iteration."""
|
||||
|
||||
def check_transformers_version(
|
||||
self,
|
||||
*,
|
||||
@@ -361,8 +382,21 @@ _EMBEDDING_EXAMPLE_MODELS = {
|
||||
"Phi3VForCausalLM": _HfExamplesInfo("TIGER-Lab/VLM2Vec-Full",
|
||||
trust_remote_code=True),
|
||||
"Qwen2VLForConditionalGeneration": _HfExamplesInfo("MrLight/dse-qwen2-2b-mrl-v1"), # noqa: E501
|
||||
"PrithviGeoSpatialMAE": _HfExamplesInfo("ibm-nasa-geospatial/Prithvi-EO-2.0-300M-TL-Sen1Floods11", # noqa: E501
|
||||
is_available_online=False), # noqa: E501
|
||||
"PrithviGeoSpatialMAE": _HfExamplesInfo("mgazz/Prithvi-EO-2.0-300M-TL-Sen1Floods11", # noqa: E501
|
||||
dtype=torch.float16,
|
||||
enforce_eager=True,
|
||||
skip_tokenizer_init=True,
|
||||
# This is to avoid the model
|
||||
# going OOM in CI
|
||||
max_num_seqs=32,
|
||||
),
|
||||
"Terratorch": _HfExamplesInfo("mgazz/Prithvi-EO-2.0-300M-TL-Sen1Floods11",
|
||||
dtype=torch.float16,
|
||||
enforce_eager=True,
|
||||
skip_tokenizer_init=True,
|
||||
# This is to avoid the model going OOM in CI
|
||||
max_num_seqs=32,
|
||||
),
|
||||
}
|
||||
|
||||
_SEQUENCE_CLASSIFICATION_EXAMPLE_MODELS = {
|
||||
|
||||
Reference in New Issue
Block a user