Spaces:
Running
on
CPU Upgrade
Running
on
CPU Upgrade
Alina Lozovskaya
commited on
Commit
•
9469eae
1
Parent(s):
0e60add
Improve model size calculation
Browse files
backend/app/utils/model_validation.py
CHANGED
@@ -5,10 +5,12 @@ import re
|
|
5 |
from typing import Tuple, Optional, Dict, Any
|
6 |
import aiohttp
|
7 |
from huggingface_hub import HfApi, ModelCard, hf_hub_download
|
|
|
8 |
from transformers import AutoConfig, AutoTokenizer
|
9 |
from app.config.base import HF_TOKEN, API
|
10 |
from app.utils.logging import LogFormatter
|
11 |
|
|
|
12 |
logger = logging.getLogger(__name__)
|
13 |
|
14 |
class ModelValidator:
|
@@ -54,78 +56,78 @@ class ModelValidator:
|
|
54 |
logger.error(LogFormatter.error(error_msg, e))
|
55 |
return False, str(e), None
|
56 |
|
57 |
-
async def get_safetensors_metadata(self, model_id: str,
|
58 |
"""Get metadata from a safetensors file"""
|
59 |
try:
|
60 |
-
|
61 |
-
|
62 |
-
|
63 |
-
|
64 |
-
|
65 |
-
|
66 |
-
|
67 |
-
|
68 |
-
|
69 |
-
|
70 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
71 |
except Exception as e:
|
72 |
-
logger.
|
73 |
return None
|
74 |
-
|
75 |
async def get_model_size(
|
76 |
self,
|
77 |
model_info: Any,
|
78 |
precision: str,
|
79 |
-
base_model: str
|
|
|
80 |
) -> Tuple[Optional[float], Optional[str]]:
|
81 |
"""Get model size in billions of parameters"""
|
82 |
try:
|
83 |
logger.info(LogFormatter.info(f"Checking model size for {model_info.modelId}"))
|
84 |
-
|
85 |
# Check if model is adapter
|
86 |
is_adapter = any(s.rfilename == "adapter_config.json" for s in model_info.siblings if hasattr(s, 'rfilename'))
|
87 |
-
|
88 |
# Try to get size from safetensors first
|
89 |
model_size = None
|
90 |
-
|
91 |
if is_adapter and base_model:
|
92 |
# For adapters, we need both adapter and base model sizes
|
93 |
-
adapter_meta = await self.get_safetensors_metadata(model_info.id,
|
94 |
-
base_meta = await self.get_safetensors_metadata(base_model)
|
95 |
-
|
96 |
if adapter_meta and base_meta:
|
97 |
-
adapter_size = sum(
|
98 |
-
base_size = sum(
|
99 |
model_size = (adapter_size + base_size) / (2 * 1e9) # Convert to billions, assuming float16
|
100 |
else:
|
101 |
# For regular models, just get the model size
|
102 |
-
meta = await self.get_safetensors_metadata(model_info.id)
|
103 |
if meta:
|
104 |
-
total_params = sum(
|
105 |
model_size = total_params / (2 * 1e9) # Convert to billions, assuming float16
|
106 |
-
|
107 |
if model_size is None:
|
108 |
-
#
|
109 |
-
|
110 |
-
|
111 |
-
|
112 |
-
if size_match:
|
113 |
-
size_str = size_match.group(1)
|
114 |
-
model_size = float(size_str)
|
115 |
-
else:
|
116 |
-
return None, "Could not determine model size from safetensors or model name"
|
117 |
-
|
118 |
# Adjust size for GPTQ models
|
119 |
size_factor = 8 if (precision == "GPTQ" or "gptq" in model_info.id.lower()) else 1
|
120 |
model_size = round(size_factor * model_size, 3)
|
121 |
-
|
122 |
logger.info(LogFormatter.success(f"Model size: {model_size}B parameters"))
|
123 |
return model_size, None
|
124 |
-
|
125 |
except Exception as e:
|
126 |
-
|
127 |
-
logger.error(LogFormatter.error(error_msg, e))
|
128 |
return None, str(e)
|
|
|
129 |
|
130 |
async def check_chat_template(
|
131 |
self,
|
|
|
5 |
from typing import Tuple, Optional, Dict, Any
|
6 |
import aiohttp
|
7 |
from huggingface_hub import HfApi, ModelCard, hf_hub_download
|
8 |
+
from huggingface_hub import hf_api
|
9 |
from transformers import AutoConfig, AutoTokenizer
|
10 |
from app.config.base import HF_TOKEN, API
|
11 |
from app.utils.logging import LogFormatter
|
12 |
|
13 |
+
|
14 |
logger = logging.getLogger(__name__)
|
15 |
|
16 |
class ModelValidator:
|
|
|
56 |
logger.error(LogFormatter.error(error_msg, e))
|
57 |
return False, str(e), None
|
58 |
|
59 |
+
async def get_safetensors_metadata(self, model_id: str, is_adapter: bool = False, revision: str = "main") -> Optional[Dict]:
|
60 |
"""Get metadata from a safetensors file"""
|
61 |
try:
|
62 |
+
if is_adapter:
|
63 |
+
metadata = await asyncio.to_thread(
|
64 |
+
hf_api.parse_safetensors_file_metadata,
|
65 |
+
model_id,
|
66 |
+
"adapter_model.safetensors",
|
67 |
+
token=self.token,
|
68 |
+
revision=revision,
|
69 |
+
)
|
70 |
+
else:
|
71 |
+
metadata = await asyncio.to_thread(
|
72 |
+
hf_api.get_safetensors_metadata,
|
73 |
+
repo_id=model_id,
|
74 |
+
token=self.token,
|
75 |
+
revision=revision,
|
76 |
+
)
|
77 |
+
return metadata
|
78 |
+
|
79 |
except Exception as e:
|
80 |
+
logger.error(f"Failed to get safetensors metadata: {str(e)}")
|
81 |
return None
|
82 |
+
|
83 |
async def get_model_size(
|
84 |
self,
|
85 |
model_info: Any,
|
86 |
precision: str,
|
87 |
+
base_model: str,
|
88 |
+
revision: str
|
89 |
) -> Tuple[Optional[float], Optional[str]]:
|
90 |
"""Get model size in billions of parameters"""
|
91 |
try:
|
92 |
logger.info(LogFormatter.info(f"Checking model size for {model_info.modelId}"))
|
93 |
+
|
94 |
# Check if model is adapter
|
95 |
is_adapter = any(s.rfilename == "adapter_config.json" for s in model_info.siblings if hasattr(s, 'rfilename'))
|
96 |
+
|
97 |
# Try to get size from safetensors first
|
98 |
model_size = None
|
99 |
+
|
100 |
if is_adapter and base_model:
|
101 |
# For adapters, we need both adapter and base model sizes
|
102 |
+
adapter_meta = await self.get_safetensors_metadata(model_info.id, is_adapter=True, revision=revision)
|
103 |
+
base_meta = await self.get_safetensors_metadata(base_model, revision="main")
|
104 |
+
|
105 |
if adapter_meta and base_meta:
|
106 |
+
adapter_size = sum(adapter_meta.parameter_count.values())
|
107 |
+
base_size = sum(base_meta.parameter_count.values())
|
108 |
model_size = (adapter_size + base_size) / (2 * 1e9) # Convert to billions, assuming float16
|
109 |
else:
|
110 |
# For regular models, just get the model size
|
111 |
+
meta = await self.get_safetensors_metadata(model_info.id, revision=revision)
|
112 |
if meta:
|
113 |
+
total_params = sum(meta.parameter_count.values())
|
114 |
model_size = total_params / (2 * 1e9) # Convert to billions, assuming float16
|
115 |
+
|
116 |
if model_size is None:
|
117 |
+
# If model size could not be determined, return an error
|
118 |
+
return None, "Model size could not be determined"
|
119 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
120 |
# Adjust size for GPTQ models
|
121 |
size_factor = 8 if (precision == "GPTQ" or "gptq" in model_info.id.lower()) else 1
|
122 |
model_size = round(size_factor * model_size, 3)
|
123 |
+
|
124 |
logger.info(LogFormatter.success(f"Model size: {model_size}B parameters"))
|
125 |
return model_size, None
|
126 |
+
|
127 |
except Exception as e:
|
128 |
+
logger.error(LogFormatter.error(f"Error while determining model size: {e}"))
|
|
|
129 |
return None, str(e)
|
130 |
+
|
131 |
|
132 |
async def check_chat_template(
|
133 |
self,
|