Spaces:
Sleeping
Sleeping
ffreemt
commited on
Commit
·
adb2c74
1
Parent(s):
6721fd3
Update 30b-Lazarus.ggmlv3.q3_K_S.bin
Browse files
README.md
CHANGED
@@ -1,5 +1,5 @@
|
|
1 |
---
|
2 |
-
title: TheBloke/
|
3 |
emoji: 🚀
|
4 |
colorFrom: green
|
5 |
colorTo: green
|
|
|
1 |
---
|
2 |
+
title: TheBloke/30B-Lazarus-GGML
|
3 |
emoji: 🚀
|
4 |
colorFrom: green
|
5 |
colorTo: green
|
app.py
CHANGED
@@ -1,6 +1,5 @@
|
|
1 |
"""Run codes."""
|
2 |
# pylint: disable=line-too-long, broad-exception-caught, invalid-name, missing-function-docstring, too-many-instance-attributes, missing-class-docstring
|
3 |
-
# ruff: noqa: E501
|
4 |
import os
|
5 |
import time
|
6 |
from dataclasses import asdict, dataclass
|
@@ -14,7 +13,7 @@ from about_time import about_time
|
|
14 |
|
15 |
# from ctransformers import AutoConfig, AutoModelForCausalLM
|
16 |
from ctransformers import AutoModelForCausalLM
|
17 |
-
from huggingface_hub import hf_hub_download
|
18 |
from loguru import logger
|
19 |
|
20 |
filename_list = [
|
@@ -39,7 +38,9 @@ URL = "https://huggingface.co/TheBloke/Wizard-Vicuna-7B-Uncensored-GGML/raw/main
|
|
39 |
URL = "https://huggingface.co/TheBloke/30B-Lazarus-GGML/blob/main/30b-Lazarus.ggmlv3.q4_0.bin"
|
40 |
URL = "https://huggingface.co/TheBloke/30B-Lazarus-GGML/blob/main/30b-Lazarus.ggmlv3.q4_1.bin"
|
41 |
URL = "https://huggingface.co/TheBloke/30B-Lazarus-GGML/resolve/main/30b-Lazarus.ggmlv3.q4_K_M.bin"
|
42 |
-
URL = "https://huggingface.co/TheBloke/30B-Lazarus-GGML/resolve/main/30b-Lazarus.ggmlv3.q4_K_S.bin"
|
|
|
|
|
43 |
|
44 |
MODEL_FILENAME = Path(URL).name
|
45 |
|
@@ -48,7 +49,7 @@ MODEL_FILENAME = Path(URL).name
|
|
48 |
|
49 |
REPO_ID = "/".join(
|
50 |
urlparse(URL).path.strip("/").split("/")[:2]
|
51 |
-
)
|
52 |
# TheBloke/30B-Lazarus-GGML
|
53 |
# # TheBloke/Wizard-Vicuna-7B-Uncensored-GGML
|
54 |
|
@@ -291,7 +292,15 @@ llm = AutoModelForCausalLM.from_pretrained(
|
|
291 |
# """
|
292 |
|
293 |
logger.info(f"start dl, {REPO_ID=}, {MODEL_FILENAME=}, {DESTINATION_FOLDER=}")
|
294 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
295 |
logger.info("done dl")
|
296 |
|
297 |
logger.debug(f"{os.cpu_count()=} {psutil.cpu_count(logical=False)=}")
|
@@ -302,12 +311,17 @@ logger.debug(f"{cpu_count=}")
|
|
302 |
|
303 |
logger.info("load llm")
|
304 |
|
|
|
|
|
|
|
|
|
305 |
_ = Path("models", MODEL_FILENAME).absolute().as_posix()
|
306 |
logger.debug(f"model_file: {_}, exists: {Path(_).exists()}")
|
307 |
LLM = AutoModelForCausalLM.from_pretrained(
|
308 |
# "TheBloke/WizardCoder-15B-1.0-GGML",
|
309 |
-
REPO_ID, # DESTINATION_FOLDER, # model_path_or_repo_id: str required
|
310 |
-
model_file=_,
|
|
|
311 |
model_type="llama", # "starcoder", AutoConfig.from_pretrained(REPO_ID)
|
312 |
threads=cpu_count,
|
313 |
)
|
|
|
1 |
"""Run codes."""
|
2 |
# pylint: disable=line-too-long, broad-exception-caught, invalid-name, missing-function-docstring, too-many-instance-attributes, missing-class-docstring
|
|
|
3 |
import os
|
4 |
import time
|
5 |
from dataclasses import asdict, dataclass
|
|
|
13 |
|
14 |
# from ctransformers import AutoConfig, AutoModelForCausalLM
|
15 |
from ctransformers import AutoModelForCausalLM
|
16 |
+
from huggingface_hub import hf_hub_download, snapshot_download
|
17 |
from loguru import logger
|
18 |
|
19 |
filename_list = [
|
|
|
38 |
URL = "https://huggingface.co/TheBloke/30B-Lazarus-GGML/blob/main/30b-Lazarus.ggmlv3.q4_0.bin"
|
39 |
URL = "https://huggingface.co/TheBloke/30B-Lazarus-GGML/blob/main/30b-Lazarus.ggmlv3.q4_1.bin"
|
40 |
URL = "https://huggingface.co/TheBloke/30B-Lazarus-GGML/resolve/main/30b-Lazarus.ggmlv3.q4_K_M.bin"
|
41 |
+
URL = "https://huggingface.co/TheBloke/30B-Lazarus-GGML/resolve/main/30b-Lazarus.ggmlv3.q4_K_S.bin" # 18GB
|
42 |
+
|
43 |
+
URL = "https://huggingface.co/TheBloke/30B-Lazarus-GGML/blob/main/30b-Lazarus.ggmlv3.q3_K_S.bin" # 14GB
|
44 |
|
45 |
MODEL_FILENAME = Path(URL).name
|
46 |
|
|
|
49 |
|
50 |
REPO_ID = "/".join(
|
51 |
urlparse(URL).path.strip("/").split("/")[:2]
|
52 |
+
)
|
53 |
# TheBloke/30B-Lazarus-GGML
|
54 |
# # TheBloke/Wizard-Vicuna-7B-Uncensored-GGML
|
55 |
|
|
|
292 |
# """
|
293 |
|
294 |
logger.info(f"start dl, {REPO_ID=}, {MODEL_FILENAME=}, {DESTINATION_FOLDER=}")
|
295 |
+
|
296 |
+
# download_quant(DESTINATION_FOLDER, REPO_ID, MODEL_FILENAME)
|
297 |
+
snapshot_download(
|
298 |
+
repo_id=REPO_ID, # TheBloke/30B-Lazarus-GGML
|
299 |
+
allow_patterns=MODEL_FILENAME, # 30b-Lazarus.ggmlv3.q4_K_S.bin 18.3G
|
300 |
+
# revision="ggmlv3",
|
301 |
+
local_dir="models",
|
302 |
+
)
|
303 |
+
|
304 |
logger.info("done dl")
|
305 |
|
306 |
logger.debug(f"{os.cpu_count()=} {psutil.cpu_count(logical=False)=}")
|
|
|
311 |
|
312 |
logger.info("load llm")
|
313 |
|
314 |
+
# from ctransformers import AutoConfig
|
315 |
+
# AutoConfig(REPO_ID)
|
316 |
+
# AutoConfig(config='TheBloke/30B-Lazarus-GGML', model_type=None)
|
317 |
+
|
318 |
_ = Path("models", MODEL_FILENAME).absolute().as_posix()
|
319 |
logger.debug(f"model_file: {_}, exists: {Path(_).exists()}")
|
320 |
LLM = AutoModelForCausalLM.from_pretrained(
|
321 |
# "TheBloke/WizardCoder-15B-1.0-GGML",
|
322 |
+
# REPO_ID, # DESTINATION_FOLDER, # model_path_or_repo_id: str required
|
323 |
+
# model_file=_,
|
324 |
+
_,
|
325 |
model_type="llama", # "starcoder", AutoConfig.from_pretrained(REPO_ID)
|
326 |
threads=cpu_count,
|
327 |
)
|