Update model name
Browse files
mteb_llama_embed_nemotron_reasoning_3b.py → mteb_llama_nv_embed_reasoning_3b.py
RENAMED
|
@@ -1,7 +1,7 @@
|
|
| 1 |
# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
|
| 2 |
# SPDX-License-Identifier: Apache-2.0.
|
| 3 |
"""
|
| 4 |
-
MTEB encoder and ModelMeta for nvidia/llama-embed-
|
| 5 |
"""
|
| 6 |
|
| 7 |
from mteb.models.model_meta import ModelMeta
|
|
@@ -30,8 +30,8 @@ BRIGHT_TASK_INSTRUCTIONS = {
|
|
| 30 |
|
| 31 |
BRIGHT_PASSAGE_PREFIX = "passage: "
|
| 32 |
|
| 33 |
-
class
|
| 34 |
-
"""
|
| 35 |
|
| 36 |
def __init__(self, model_name: str, revision: str, device: str | None = None, **kwargs) -> None:
|
| 37 |
super().__init__(model_name, revision=revision, device=device)
|
|
@@ -63,10 +63,10 @@ class LlamaEmbedNemotronReasoning(LlamaEmbedNemotron):
|
|
| 63 |
prefix = self.format_instruction(instruction, prompt_type)
|
| 64 |
return self._extract_embeddings(inputs, instruction=prefix, **kwargs)
|
| 65 |
|
| 66 |
-
|
| 67 |
-
loader=
|
| 68 |
loader_kwargs=dict(max_seq_length=8192),
|
| 69 |
-
name="nvidia/llama-embed-
|
| 70 |
model_type=["dense"],
|
| 71 |
languages=llama_embed_nemotron_evaluated_languages,
|
| 72 |
open_weights=True,
|
|
@@ -75,9 +75,9 @@ LLAMA_EMBED_NEMOTRON_REASONING_3B_META = ModelMeta(
|
|
| 75 |
n_parameters=3_212_749_824,
|
| 76 |
memory_usage_mb=6000,
|
| 77 |
embed_dim=3072,
|
| 78 |
-
license="https://huggingface.co/nvidia/llama-embed-
|
| 79 |
max_tokens=8192,
|
| 80 |
-
reference="https://huggingface.co/nvidia/llama-embed-
|
| 81 |
similarity_fn_name="cosine",
|
| 82 |
framework=["PyTorch", "Transformers"],
|
| 83 |
use_instructions=True,
|
|
|
|
| 1 |
# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
|
| 2 |
# SPDX-License-Identifier: Apache-2.0.
|
| 3 |
"""
|
| 4 |
+
MTEB encoder and ModelMeta for nvidia/llama-nv-embed-reasoning-3b.
|
| 5 |
"""
|
| 6 |
|
| 7 |
from mteb.models.model_meta import ModelMeta
|
|
|
|
| 30 |
|
| 31 |
BRIGHT_PASSAGE_PREFIX = "passage: "
|
| 32 |
|
| 33 |
+
class LlamaNvEmbedReasoning(LlamaEmbedNemotron):
|
| 34 |
+
"""LlamaNvEmbedReasoning for reasoning with BRIGHT benchmark prompts."""
|
| 35 |
|
| 36 |
def __init__(self, model_name: str, revision: str, device: str | None = None, **kwargs) -> None:
|
| 37 |
super().__init__(model_name, revision=revision, device=device)
|
|
|
|
| 63 |
prefix = self.format_instruction(instruction, prompt_type)
|
| 64 |
return self._extract_embeddings(inputs, instruction=prefix, **kwargs)
|
| 65 |
|
| 66 |
+
LLAMA_NV_EMBED_REASONING_3B_META = ModelMeta(
|
| 67 |
+
loader=LlamaNvEmbedReasoning,
|
| 68 |
loader_kwargs=dict(max_seq_length=8192),
|
| 69 |
+
name="nvidia/llama-nv-embed-reasoning-3b",
|
| 70 |
model_type=["dense"],
|
| 71 |
languages=llama_embed_nemotron_evaluated_languages,
|
| 72 |
open_weights=True,
|
|
|
|
| 75 |
n_parameters=3_212_749_824,
|
| 76 |
memory_usage_mb=6000,
|
| 77 |
embed_dim=3072,
|
| 78 |
+
license="https://huggingface.co/nvidia/llama-nv-embed-reasoning-3b/blob/main/LICENSE",
|
| 79 |
max_tokens=8192,
|
| 80 |
+
reference="https://huggingface.co/nvidia/llama-nv-embed-reasoning-3b",
|
| 81 |
similarity_fn_name="cosine",
|
| 82 |
framework=["PyTorch", "Transformers"],
|
| 83 |
use_instructions=True,
|