CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-31 08:51:55 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'master' into compilade/refactor-kv-cache

Browse Source
This commit is contained in:
Francis Couture-Harpin
2024-11-25 10:40:20 -05:00
parent 8006f3b3c8 cce5a90075
commit 691698e152
375 changed files with 86333 additions and 53885 deletions
Download Patch File Download Diff File Expand all files Collapse all files

32
gguf-py/gguf/constants.py
View File

@@ -64,15 +64,27 @@ class Keys:
BASE_MODEL_AUTHOR = "general.base_model.{id}.author"
BASE_MODEL_VERSION = "general.base_model.{id}.version"
BASE_MODEL_ORGANIZATION = "general.base_model.{id}.organization"
BASE_MODEL_DESCRIPTION = "general.base_model.{id}.description"
BASE_MODEL_URL = "general.base_model.{id}.url" # Model Website/Paper
BASE_MODEL_DOI = "general.base_model.{id}.doi"
BASE_MODEL_UUID = "general.base_model.{id}.uuid"
BASE_MODEL_REPO_URL = "general.base_model.{id}.repo_url" # Model Source Repository (git/svn/etc...)
# Dataset Source
DATASET_COUNT = "general.dataset.count"
DATASET_NAME = "general.dataset.{id}.name"
DATASET_AUTHOR = "general.dataset.{id}.author"
DATASET_VERSION = "general.dataset.{id}.version"
DATASET_ORGANIZATION = "general.dataset.{id}.organization"
DATASET_DESCRIPTION = "general.dataset.{id}.description"
DATASET_URL = "general.dataset.{id}.url" # Model Website/Paper
DATASET_DOI = "general.dataset.{id}.doi"
DATASET_UUID = "general.dataset.{id}.uuid"
DATASET_REPO_URL = "general.dataset.{id}.repo_url" # Model Source Repository (git/svn/etc...)
# Array based KV stores
TAGS = "general.tags"
LANGUAGES = "general.languages"
DATASETS = "general.datasets"
class LLM:
VOCAB_SIZE = "{arch}.vocab_size"
@@ -232,6 +244,7 @@ class MODEL_ARCH(IntEnum):
COMMAND_R = auto()
DBRX = auto()
OLMO = auto()
OLMO_1124 = auto()
OLMOE = auto()
OPENELM = auto()
ARCTIC = auto()
@@ -397,6 +410,7 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
MODEL_ARCH.COMMAND_R: "command-r",
MODEL_ARCH.DBRX: "dbrx",
MODEL_ARCH.OLMO: "olmo",
MODEL_ARCH.OLMO_1124: "olmo_1124",
MODEL_ARCH.OLMOE: "olmoe",
MODEL_ARCH.OPENELM: "openelm",
MODEL_ARCH.ARCTIC: "arctic",
@@ -1093,6 +1107,22 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
MODEL_TENSOR.FFN_DOWN,
MODEL_TENSOR.FFN_UP,
],
MODEL_ARCH.OLMO_1124: [
MODEL_TENSOR.TOKEN_EMBD,
MODEL_TENSOR.OUTPUT_NORM,
MODEL_TENSOR.OUTPUT,
MODEL_TENSOR.ATTN_Q,
MODEL_TENSOR.ATTN_K,
MODEL_TENSOR.ATTN_V,
MODEL_TENSOR.ATTN_OUT,
MODEL_TENSOR.ATTN_POST_NORM,
MODEL_TENSOR.ATTN_Q_NORM,
MODEL_TENSOR.ATTN_K_NORM,
MODEL_TENSOR.FFN_POST_NORM,
MODEL_TENSOR.FFN_GATE,
MODEL_TENSOR.FFN_DOWN,
MODEL_TENSOR.FFN_UP,
],
MODEL_ARCH.OLMOE: [
MODEL_TENSOR.TOKEN_EMBD,
MODEL_TENSOR.OUTPUT_NORM,

36
gguf-py/gguf/gguf_writer.py
View File

@@ -568,6 +568,9 @@ class GGUFWriter:
def add_base_model_organization(self, source_id: int, organization: str) -> None:
self.add_string(Keys.General.BASE_MODEL_ORGANIZATION.format(id=source_id), organization)
def add_base_model_description(self, source_id: int, description: str) -> None:
self.add_string(Keys.General.BASE_MODEL_DESCRIPTION.format(id=source_id), description)
def add_base_model_url(self, source_id: int, url: str) -> None:
self.add_string(Keys.General.BASE_MODEL_URL.format(id=source_id), url)
@@ -580,15 +583,42 @@ class GGUFWriter:
def add_base_model_repo_url(self, source_id: int, repo_url: str) -> None:
self.add_string(Keys.General.BASE_MODEL_REPO_URL.format(id=source_id), repo_url)
def add_dataset_count(self, source_count: int) -> None:
self.add_uint32(Keys.General.DATASET_COUNT, source_count)
def add_dataset_name(self, source_id: int, name: str) -> None:
self.add_string(Keys.General.DATASET_NAME.format(id=source_id), name)
def add_dataset_author(self, source_id: int, author: str) -> None:
self.add_string(Keys.General.DATASET_AUTHOR.format(id=source_id), author)
def add_dataset_version(self, source_id: int, version: str) -> None:
self.add_string(Keys.General.DATASET_VERSION.format(id=source_id), version)
def add_dataset_organization(self, source_id: int, organization: str) -> None:
self.add_string(Keys.General.DATASET_ORGANIZATION.format(id=source_id), organization)
def add_dataset_description(self, source_id: int, description: str) -> None:
self.add_string(Keys.General.DATASET_DESCRIPTION.format(id=source_id), description)
def add_dataset_url(self, source_id: int, url: str) -> None:
self.add_string(Keys.General.DATASET_URL.format(id=source_id), url)
def add_dataset_doi(self, source_id: int, doi: str) -> None:
self.add_string(Keys.General.DATASET_DOI.format(id=source_id), doi)
def add_dataset_uuid(self, source_id: int, uuid: str) -> None:
self.add_string(Keys.General.DATASET_UUID.format(id=source_id), uuid)
def add_dataset_repo_url(self, source_id: int, repo_url: str) -> None:
self.add_string(Keys.General.DATASET_REPO_URL.format(id=source_id), repo_url)
def add_tags(self, tags: Sequence[str]) -> None:
self.add_array(Keys.General.TAGS, tags)
def add_languages(self, languages: Sequence[str]) -> None:
self.add_array(Keys.General.LANGUAGES, languages)
def add_datasets(self, datasets: Sequence[str]) -> None:
self.add_array(Keys.General.DATASETS, datasets)
def add_tensor_data_layout(self, layout: str) -> None:
self.add_string(Keys.LLM.TENSOR_DATA_LAYOUT.format(arch=self.arch), layout)

150
gguf-py/gguf/metadata.py
View File

@@ -41,7 +41,7 @@ class Metadata:
base_models: Optional[list[dict]] = None
tags: Optional[list[str]] = None
languages: Optional[list[str]] = None
datasets: Optional[list[str]] = None
datasets: Optional[list[dict]] = None
@staticmethod
def load(metadata_override_path: Optional[Path] = None, model_path: Optional[Path] = None, model_name: Optional[str] = None, total_params: int = 0) -> Metadata:
@@ -91,9 +91,11 @@ class Metadata:
# Base Models is received here as an array of models
metadata.base_models = metadata_override.get("general.base_models", metadata.base_models)
# Datasets is received here as an array of datasets
metadata.datasets = metadata_override.get("general.datasets", metadata.datasets)
metadata.tags = metadata_override.get(Keys.General.TAGS, metadata.tags)
metadata.languages = metadata_override.get(Keys.General.LANGUAGES, metadata.languages)
metadata.datasets = metadata_override.get(Keys.General.DATASETS, metadata.datasets)
# Direct Metadata Override (via direct cli argument)
if model_name is not None:
@@ -346,12 +348,12 @@ class Metadata:
use_model_card_metadata("author", "model_creator")
use_model_card_metadata("basename", "model_type")
if "base_model" in model_card:
if "base_model" in model_card or "base_models" in model_card or "base_model_sources" in model_card:
# This represents the parent models that this is based on
# Example: stabilityai/stable-diffusion-xl-base-1.0. Can also be a list (for merges)
# Example of merges: https://huggingface.co/EmbeddedLLM/Mistral-7B-Merge-14-v0.1/blob/main/README.md
metadata_base_models = []
base_model_value = model_card.get("base_model", None)
base_model_value = model_card.get("base_model", model_card.get("base_models", model_card.get("base_model_sources", None)))
if base_model_value is not None:
if isinstance(base_model_value, str):
@@ -364,18 +366,106 @@ class Metadata:
for model_id in metadata_base_models:
# NOTE: model size of base model is assumed to be similar to the size of the current model
model_full_name_component, org_component, basename, finetune, version, size_label = Metadata.get_model_id_components(model_id, total_params)
base_model = {}
if model_full_name_component is not None:
base_model["name"] = Metadata.id_to_title(model_full_name_component)
if org_component is not None:
base_model["organization"] = Metadata.id_to_title(org_component)
if version is not None:
base_model["version"] = version
if org_component is not None and model_full_name_component is not None:
base_model["repo_url"] = f"https://huggingface.co/{org_component}/{model_full_name_component}"
if isinstance(model_id, str):
if model_id.startswith("http://") or model_id.startswith("https://") or model_id.startswith("ssh://"):
base_model["repo_url"] = model_id
# Check if Hugging Face ID is present in URL
if "huggingface.co" in model_id:
match = re.match(r"https?://huggingface.co/([^/]+/[^/]+)$", model_id)
if match:
model_id_component = match.group(1)
model_full_name_component, org_component, basename, finetune, version, size_label = Metadata.get_model_id_components(model_id_component, total_params)
# Populate model dictionary with extracted components
if model_full_name_component is not None:
base_model["name"] = Metadata.id_to_title(model_full_name_component)
if org_component is not None:
base_model["organization"] = Metadata.id_to_title(org_component)
if version is not None:
base_model["version"] = version
else:
# Likely a Hugging Face ID
model_full_name_component, org_component, basename, finetune, version, size_label = Metadata.get_model_id_components(model_id, total_params)
# Populate model dictionary with extracted components
if model_full_name_component is not None:
base_model["name"] = Metadata.id_to_title(model_full_name_component)
if org_component is not None:
base_model["organization"] = Metadata.id_to_title(org_component)
if version is not None:
base_model["version"] = version
if org_component is not None and model_full_name_component is not None:
base_model["repo_url"] = f"https://huggingface.co/{org_component}/{model_full_name_component}"
elif isinstance(model_id, dict):
base_model = model_id
else:
logger.error(f"base model entry '{str(model_id)}' not in a known format")
metadata.base_models.append(base_model)
if "datasets" in model_card or "dataset" in model_card or "dataset_sources" in model_card:
# This represents the datasets that this was trained from
metadata_datasets = []
dataset_value = model_card.get("datasets", model_card.get("dataset", model_card.get("dataset_sources", None)))
if dataset_value is not None:
if isinstance(dataset_value, str):
metadata_datasets.append(dataset_value)
elif isinstance(dataset_value, list):
metadata_datasets.extend(dataset_value)
if metadata.datasets is None:
metadata.datasets = []
for dataset_id in metadata_datasets:
# NOTE: model size of base model is assumed to be similar to the size of the current model
dataset = {}
if isinstance(dataset_id, str):
if dataset_id.startswith(("http://", "https://", "ssh://")):
dataset["repo_url"] = dataset_id
# Check if Hugging Face ID is present in URL
if "huggingface.co" in dataset_id:
match = re.match(r"https?://huggingface.co/([^/]+/[^/]+)$", dataset_id)
if match:
dataset_id_component = match.group(1)
dataset_name_component, org_component, basename, finetune, version, size_label = Metadata.get_model_id_components(dataset_id_component, total_params)
# Populate dataset dictionary with extracted components
if dataset_name_component is not None:
dataset["name"] = Metadata.id_to_title(dataset_name_component)
if org_component is not None:
dataset["organization"] = Metadata.id_to_title(org_component)
if version is not None:
dataset["version"] = version
else:
# Likely a Hugging Face ID
dataset_name_component, org_component, basename, finetune, version, size_label = Metadata.get_model_id_components(dataset_id, total_params)
# Populate dataset dictionary with extracted components
if dataset_name_component is not None:
dataset["name"] = Metadata.id_to_title(dataset_name_component)
if org_component is not None:
dataset["organization"] = Metadata.id_to_title(org_component)
if version is not None:
dataset["version"] = version
if org_component is not None and dataset_name_component is not None:
dataset["repo_url"] = f"https://huggingface.co/{org_component}/{dataset_name_component}"
elif isinstance(dataset_id, dict):
dataset = dataset_id
else:
logger.error(f"dataset entry '{str(dataset_id)}' not in a known format")
metadata.datasets.append(dataset)
use_model_card_metadata("license", "license")
use_model_card_metadata("license_name", "license_name")
use_model_card_metadata("license_link", "license_link")
@@ -386,9 +476,6 @@ class Metadata:
use_array_model_card_metadata("languages", "languages")
use_array_model_card_metadata("languages", "language")
use_array_model_card_metadata("datasets", "datasets")
use_array_model_card_metadata("datasets", "dataset")
# Hugging Face Parameter Heuristics
####################################
@@ -458,7 +545,10 @@ class Metadata:
gguf_writer.add_size_label(self.size_label)
if self.license is not None:
gguf_writer.add_license(self.license)
if isinstance(self.license, list):
gguf_writer.add_license(",".join(self.license))
else:
gguf_writer.add_license(self.license)
if self.license_name is not None:
gguf_writer.add_license_name(self.license_name)
if self.license_link is not None:
@@ -493,6 +583,8 @@ class Metadata:
gguf_writer.add_base_model_version(key, base_model_entry["version"])
if "organization" in base_model_entry:
gguf_writer.add_base_model_organization(key, base_model_entry["organization"])
if "description" in base_model_entry:
gguf_writer.add_base_model_description(key, base_model_entry["description"])
if "url" in base_model_entry:
gguf_writer.add_base_model_url(key, base_model_entry["url"])
if "doi" in base_model_entry:
@@ -502,9 +594,29 @@ class Metadata:
if "repo_url" in base_model_entry:
gguf_writer.add_base_model_repo_url(key, base_model_entry["repo_url"])
if self.datasets is not None:
gguf_writer.add_dataset_count(len(self.datasets))
for key, dataset_entry in enumerate(self.datasets):
if "name" in dataset_entry:
gguf_writer.add_dataset_name(key, dataset_entry["name"])
if "author" in dataset_entry:
gguf_writer.add_dataset_author(key, dataset_entry["author"])
if "version" in dataset_entry:
gguf_writer.add_dataset_version(key, dataset_entry["version"])
if "organization" in dataset_entry:
gguf_writer.add_dataset_organization(key, dataset_entry["organization"])
if "description" in dataset_entry:
gguf_writer.add_dataset_description(key, dataset_entry["description"])
if "url" in dataset_entry:
gguf_writer.add_dataset_url(key, dataset_entry["url"])
if "doi" in dataset_entry:
gguf_writer.add_dataset_doi(key, dataset_entry["doi"])
if "uuid" in dataset_entry:
gguf_writer.add_dataset_uuid(key, dataset_entry["uuid"])
if "repo_url" in dataset_entry:
gguf_writer.add_dataset_repo_url(key, dataset_entry["repo_url"])
if self.tags is not None:
gguf_writer.add_tags(self.tags)
if self.languages is not None:
gguf_writer.add_languages(self.languages)
if self.datasets is not None:
gguf_writer.add_datasets(self.datasets)

28
gguf-py/gguf/tensor_mapping.py
View File

@@ -13,7 +13,7 @@ class TensorNameMap:
"transformer.wte", # gpt2 gpt-j mpt refact qwen dbrx jais exaone
"transformer.word_embeddings", # falcon
"word_embeddings", # bloom
"model.embed_tokens", # llama-hf nemotron olmoe
"model.embed_tokens", # llama-hf nemotron olmoe olmo_1124
"tok_embeddings", # llama-pth
"embeddings.word_embeddings", # bert nomic-bert
"language_model.embedding.word_embeddings", # persimmon
@@ -54,7 +54,7 @@ class TensorNameMap:
# Output
MODEL_TENSOR.OUTPUT: (
"embed_out", # gptneox
"lm_head", # gpt2 mpt falcon llama-hf baichuan qwen mamba dbrx jais nemotron exaone olmoe
"lm_head", # gpt2 mpt falcon llama-hf baichuan qwen mamba dbrx jais nemotron exaone olmoe olmo_1124
"output", # llama-pth bloom internlm2
"word_embeddings_for_head", # persimmon
"lm_head.linear", # phi2
@@ -66,7 +66,7 @@ class TensorNameMap:
MODEL_TENSOR.OUTPUT_NORM: (
"gpt_neox.final_layer_norm", # gptneox
"transformer.ln_f", # gpt2 gpt-j falcon jais exaone
"model.norm", # llama-hf baichuan internlm2 olmoe
"model.norm", # llama-hf baichuan internlm2 olmoe olmo_1124
"norm", # llama-pth
"transformer.norm_f", # mpt dbrx
"ln_f", # refact bloom qwen gpt2
@@ -145,7 +145,7 @@ class TensorNameMap:
# Attention query
MODEL_TENSOR.ATTN_Q: (
"model.layers.{bid}.self_attn.q_proj", # llama-hf nemotron olmoe
"model.layers.{bid}.self_attn.q_proj", # llama-hf nemotron olmoe olmo_1124
"layers.{bid}.attention.wq", # llama-pth
"encoder.layer.{bid}.attention.self.query", # bert
"transformer.h.{bid}.attn.q_proj", # gpt-j
@@ -157,7 +157,7 @@ class TensorNameMap:
# Attention key
MODEL_TENSOR.ATTN_K: (
"model.layers.{bid}.self_attn.k_proj", # llama-hf nemotron olmoe
"model.layers.{bid}.self_attn.k_proj", # llama-hf nemotron olmoe olmo_1124
"layers.{bid}.attention.wk", # llama-pth
"encoder.layer.{bid}.attention.self.key", # bert
"transformer.h.{bid}.attn.k_proj", # gpt-j
@@ -170,7 +170,7 @@ class TensorNameMap:
# Attention value
MODEL_TENSOR.ATTN_V: (
"model.layers.{bid}.self_attn.v_proj", # llama-hf nemotron olmoe
"model.layers.{bid}.self_attn.v_proj", # llama-hf nemotron olmoe olmo_1124
"layers.{bid}.attention.wv", # llama-pth
"encoder.layer.{bid}.attention.self.value", # bert
"transformer.h.{bid}.attn.v_proj", # gpt-j
@@ -188,7 +188,7 @@ class TensorNameMap:
"transformer.blocks.{bid}.attn.out_proj", # mpt
"transformer.h.{bid}.self_attention.dense", # falcon
"h.{bid}.self_attention.dense", # bloom
"model.layers.{bid}.self_attn.o_proj", # llama-hf nemotron olmoe
"model.layers.{bid}.self_attn.o_proj", # llama-hf nemotron olmoe olmo_1124
"layers.{bid}.attention.wo", # llama-pth
"encoder.layer.{bid}.attention.output.dense", # bert
"transformer.h.{bid}.attn.out_proj", # gpt-j
@@ -215,7 +215,7 @@ class TensorNameMap:
),
MODEL_TENSOR.ATTN_POST_NORM: (
"model.layers.{bid}.post_attention_layernorm", # gemma2
"model.layers.{bid}.post_attention_layernorm", # gemma2 olmo_1124
),
# Rotary embeddings
@@ -252,7 +252,7 @@ class TensorNameMap:
# Post feed-forward norm
MODEL_TENSOR.FFN_POST_NORM: (
"model.layers.{bid}.post_feedforward_layernorm", # gemma2
"model.layers.{bid}.post_feedforward_layernorm", # gemma2 olmo_1124
),
MODEL_TENSOR.FFN_GATE_INP: (
@@ -276,7 +276,7 @@ class TensorNameMap:
"transformer.blocks.{bid}.ffn.up_proj", # mpt
"transformer.h.{bid}.mlp.dense_h_to_4h", # falcon
"h.{bid}.mlp.dense_h_to_4h", # bloom
"model.layers.{bid}.mlp.up_proj", # llama-hf refact nemotron
"model.layers.{bid}.mlp.up_proj", # llama-hf refact nemotron olmo_1124
"layers.{bid}.feed_forward.w3", # llama-pth
"encoder.layer.{bid}.intermediate.dense", # bert
"transformer.h.{bid}.mlp.fc_in", # gpt-j
@@ -318,7 +318,7 @@ class TensorNameMap:
# Feed-forward gate
MODEL_TENSOR.FFN_GATE: (
"model.layers.{bid}.mlp.gate_proj", # llama-hf refact
"model.layers.{bid}.mlp.gate_proj", # llama-hf refact olmo_1124
"layers.{bid}.feed_forward.w1", # llama-pth
"transformer.h.{bid}.mlp.w2", # qwen
"transformer.h.{bid}.mlp.c_fc2", # jais
@@ -351,7 +351,7 @@ class TensorNameMap:
"transformer.blocks.{bid}.ffn.down_proj", # mpt
"transformer.h.{bid}.mlp.dense_4h_to_h", # falcon
"h.{bid}.mlp.dense_4h_to_h", # bloom
"model.layers.{bid}.mlp.down_proj", # llama-hf nemotron
"model.layers.{bid}.mlp.down_proj", # llama-hf nemotron olmo_1124
"layers.{bid}.feed_forward.w2", # llama-pth
"encoder.layer.{bid}.output.dense", # bert
"transformer.h.{bid}.mlp.fc_out", # gpt-j
@@ -389,7 +389,7 @@ class TensorNameMap:
MODEL_TENSOR.ATTN_Q_NORM: (
"language_model.encoder.layers.{bid}.self_attention.q_layernorm",
"model.layers.{bid}.self_attn.q_layernorm", # persimmon
"model.layers.{bid}.self_attn.q_norm", # cohere olmoe chameleon
"model.layers.{bid}.self_attn.q_norm", # cohere olmoe chameleon olmo_1124
"transformer.blocks.{bid}.attn.q_ln", # sea-lion
"encoder.layer.{bid}.attention.self.layer_norm_q", # jina-bert-v2
"transformer.layers.{bid}.attn.q_norm", # openelm
@@ -398,7 +398,7 @@ class TensorNameMap:
MODEL_TENSOR.ATTN_K_NORM: (
"language_model.encoder.layers.{bid}.self_attention.k_layernorm",
"model.layers.{bid}.self_attn.k_layernorm", # persimmon
"model.layers.{bid}.self_attn.k_norm", # cohere olmoe chameleon
"model.layers.{bid}.self_attn.k_norm", # cohere olmoe chameleon olmo_1124
"transformer.blocks.{bid}.attn.k_ln", # sea-lion
"encoder.layer.{bid}.attention.self.layer_norm_k", # jina-bert-v2
"transformer.layers.{bid}.attn.k_norm", # openelm

37
gguf-py/tests/test_metadata.py
View File

@@ -182,8 +182,43 @@ class TestMetadataMethod(unittest.TestCase):
expect.base_models=[{'name': 'Mistral 7B Merge 14 v0', 'organization': 'EmbeddedLLM', 'version': '14-v0', 'repo_url': 'https://huggingface.co/EmbeddedLLM/Mistral-7B-Merge-14-v0'}, {'name': 'Trinity v1', 'organization': 'Janai Hq', 'version': 'v1', 'repo_url': 'https://huggingface.co/janai-hq/trinity-v1'}]
expect.tags=['Llama-3', 'instruct', 'finetune', 'chatml', 'DPO', 'RLHF', 'gpt4', 'synthetic data', 'distillation', 'function calling', 'json mode', 'axolotl']
expect.languages=['en']
expect.datasets=['teknium/OpenHermes-2.5']
expect.datasets=[{'name': 'OpenHermes 2.5', 'organization': 'Teknium', 'version': '2.5', 'repo_url': 'https://huggingface.co/teknium/OpenHermes-2.5'}]
self.assertEqual(got, expect)
# Base Model spec is inferred from model id
model_card = {'base_models': 'teknium/OpenHermes-2.5'}
expect = gguf.Metadata(base_models=[{'name': 'OpenHermes 2.5', 'organization': 'Teknium', 'version': '2.5', 'repo_url': 'https://huggingface.co/teknium/OpenHermes-2.5'}])
got = gguf.Metadata.apply_metadata_heuristic(gguf.Metadata(), model_card, None, None)
self.assertEqual(got, expect)
# Base Model spec is only url
model_card = {'base_models': ['https://huggingface.co/teknium/OpenHermes-2.5']}
expect = gguf.Metadata(base_models=[{'name': 'OpenHermes 2.5', 'organization': 'Teknium', 'version': '2.5', 'repo_url': 'https://huggingface.co/teknium/OpenHermes-2.5'}])
got = gguf.Metadata.apply_metadata_heuristic(gguf.Metadata(), model_card, None, None)
self.assertEqual(got, expect)
# Base Model spec is given directly
model_card = {'base_models': [{'name': 'OpenHermes 2.5', 'organization': 'Teknium', 'version': '2.5', 'repo_url': 'https://huggingface.co/teknium/OpenHermes-2.5'}]}
expect = gguf.Metadata(base_models=[{'name': 'OpenHermes 2.5', 'organization': 'Teknium', 'version': '2.5', 'repo_url': 'https://huggingface.co/teknium/OpenHermes-2.5'}])
got = gguf.Metadata.apply_metadata_heuristic(gguf.Metadata(), model_card, None, None)
self.assertEqual(got, expect)
# Dataset spec is inferred from model id
model_card = {'datasets': 'teknium/OpenHermes-2.5'}
expect = gguf.Metadata(datasets=[{'name': 'OpenHermes 2.5', 'organization': 'Teknium', 'version': '2.5', 'repo_url': 'https://huggingface.co/teknium/OpenHermes-2.5'}])
got = gguf.Metadata.apply_metadata_heuristic(gguf.Metadata(), model_card, None, None)
self.assertEqual(got, expect)
# Dataset spec is only url
model_card = {'datasets': ['https://huggingface.co/teknium/OpenHermes-2.5']}
expect = gguf.Metadata(datasets=[{'name': 'OpenHermes 2.5', 'organization': 'Teknium', 'version': '2.5', 'repo_url': 'https://huggingface.co/teknium/OpenHermes-2.5'}])
got = gguf.Metadata.apply_metadata_heuristic(gguf.Metadata(), model_card, None, None)
self.assertEqual(got, expect)
# Dataset spec is given directly
model_card = {'datasets': [{'name': 'OpenHermes 2.5', 'organization': 'Teknium', 'version': '2.5', 'repo_url': 'https://huggingface.co/teknium/OpenHermes-2.5'}]}
expect = gguf.Metadata(datasets=[{'name': 'OpenHermes 2.5', 'organization': 'Teknium', 'version': '2.5', 'repo_url': 'https://huggingface.co/teknium/OpenHermes-2.5'}])
got = gguf.Metadata.apply_metadata_heuristic(gguf.Metadata(), model_card, None, None)
self.assertEqual(got, expect)
def test_apply_metadata_heuristic_from_hf_parameters(self):