Spaces:
Runtime error
Runtime error
ffreemt
commited on
Commit
·
2fe65b3
1
Parent(s):
34fa40f
Update prompt_tempalte branch
Browse files
app.py
CHANGED
|
@@ -7,6 +7,7 @@ import random
|
|
| 7 |
import time
|
| 8 |
from dataclasses import asdict, dataclass, field
|
| 9 |
from pathlib import Path
|
|
|
|
| 10 |
|
| 11 |
# from types import SimpleNamespace
|
| 12 |
import gradio as gr
|
|
@@ -20,38 +21,11 @@ from examples_list import examples_list
|
|
| 20 |
|
| 21 |
url = "https://huggingface.co/TheBloke/llama-2-13B-Guanaco-QLoRA-GGML/blob/main/llama-2-13b-guanaco-qlora.ggmlv3.q4_K_S.bin" # 8.14G
|
| 22 |
|
| 23 |
-
# Prompt template: Guanaco
|
| 24 |
-
# {past_history}
|
| 25 |
-
prompt_template = """You are a helpful assistant. Let's think step by step.
|
| 26 |
-
### Human:
|
| 27 |
-
{question}
|
| 28 |
-
### Assistant:"""
|
| 29 |
-
human_prefix = "### Human"
|
| 30 |
-
ai_prefix = "### Assistant"
|
| 31 |
-
stop_list = [f"{human_prefix}:"]
|
| 32 |
-
|
| 33 |
-
prompt_template = """### System:
|
| 34 |
-
You are Stable Beluga, an AI that follows instructions extremely well. Help as much as you can.
|
| 35 |
-
|
| 36 |
-
### User: {question}
|
| 37 |
-
|
| 38 |
-
### Assistant:
|
| 39 |
-
"""
|
| 40 |
-
human_prefix = "### User"
|
| 41 |
-
ai_prefix = "### Assistant"
|
| 42 |
-
stop_list = [f"{human_prefix}:"]
|
| 43 |
-
|
| 44 |
-
_ = psutil.cpu_count(logical=False) - 1
|
| 45 |
-
cpu_count: int = int(_) if _ else 1
|
| 46 |
-
logger.debug(f"{cpu_count=}")
|
| 47 |
-
|
| 48 |
LLM = None
|
| 49 |
|
| 50 |
-
if "forindo" in platform.node():
|
| 51 |
# url = "https://huggingface.co/TheBloke/llama-2-70b-Guanaco-QLoRA-GGML/blob/main/llama-2-70b-guanaco-qlora.ggmlv3.q3_K_S.bin" # 29.7G
|
| 52 |
# model_loc = "/home/mu2018/github/langchain-llama-2-70b-guanaco-qlora-ggml/models/llama-2-70b-guanaco-qlora.ggmlv3.q3_K_S.bin"
|
| 53 |
-
model_loc = "models/stablebeluga2-70b.ggmlv3.q3_K_S.bin"
|
| 54 |
-
assert Path(model_loc).exists(), f"Make sure {model_loc=} exists."
|
| 55 |
_ = """
|
| 56 |
url = "https://huggingface.co/TheBloke/StableBeluga2-70B-GGML/blob/main/stablebeluga2-70b.ggmlv3.q3_K_S.bin"
|
| 57 |
try:
|
|
@@ -61,8 +35,11 @@ if "forindo" in platform.node():
|
|
| 61 |
logger.error(exc_)
|
| 62 |
raise SystemExit(1) from exc_
|
| 63 |
# """
|
|
|
|
|
|
|
| 64 |
else:
|
| 65 |
try:
|
|
|
|
| 66 |
model_loc, file_size = dl_hf_model(url)
|
| 67 |
logger.info(f"done load llm {model_loc=} {file_size=}G")
|
| 68 |
except Exception as exc_:
|
|
@@ -71,6 +48,35 @@ else:
|
|
| 71 |
|
| 72 |
# raise SystemExit(0)
|
| 73 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 74 |
logger.debug(f"{model_loc=}")
|
| 75 |
LLM = AutoModelForCausalLM.from_pretrained(
|
| 76 |
model_loc,
|
|
|
|
| 7 |
import time
|
| 8 |
from dataclasses import asdict, dataclass, field
|
| 9 |
from pathlib import Path
|
| 10 |
+
from textwrap import dedent
|
| 11 |
|
| 12 |
# from types import SimpleNamespace
|
| 13 |
import gradio as gr
|
|
|
|
| 21 |
|
| 22 |
url = "https://huggingface.co/TheBloke/llama-2-13B-Guanaco-QLoRA-GGML/blob/main/llama-2-13b-guanaco-qlora.ggmlv3.q4_K_S.bin" # 8.14G
|
| 23 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 24 |
LLM = None
|
| 25 |
|
| 26 |
+
if "forindo" in platform.node(): # deploy 70b model locally
|
| 27 |
# url = "https://huggingface.co/TheBloke/llama-2-70b-Guanaco-QLoRA-GGML/blob/main/llama-2-70b-guanaco-qlora.ggmlv3.q3_K_S.bin" # 29.7G
|
| 28 |
# model_loc = "/home/mu2018/github/langchain-llama-2-70b-guanaco-qlora-ggml/models/llama-2-70b-guanaco-qlora.ggmlv3.q3_K_S.bin"
|
|
|
|
|
|
|
| 29 |
_ = """
|
| 30 |
url = "https://huggingface.co/TheBloke/StableBeluga2-70B-GGML/blob/main/stablebeluga2-70b.ggmlv3.q3_K_S.bin"
|
| 31 |
try:
|
|
|
|
| 35 |
logger.error(exc_)
|
| 36 |
raise SystemExit(1) from exc_
|
| 37 |
# """
|
| 38 |
+
model_loc = "models/stablebeluga2-70b.ggmlv3.q3_K_S.bin"
|
| 39 |
+
assert Path(model_loc).exists(), f"Make sure {model_loc=} exists."
|
| 40 |
else:
|
| 41 |
try:
|
| 42 |
+
logger.debug(f" dl {url}")
|
| 43 |
model_loc, file_size = dl_hf_model(url)
|
| 44 |
logger.info(f"done load llm {model_loc=} {file_size=}G")
|
| 45 |
except Exception as exc_:
|
|
|
|
| 48 |
|
| 49 |
# raise SystemExit(0)
|
| 50 |
|
| 51 |
+
# Prompt template: Guanaco
|
| 52 |
+
# {past_history}
|
| 53 |
+
prompt_template = """You are a helpful assistant. Let's think step by step.
|
| 54 |
+
### Human:
|
| 55 |
+
{question}
|
| 56 |
+
### Assistant:"""
|
| 57 |
+
human_prefix = "### Human"
|
| 58 |
+
ai_prefix = "### Assistant"
|
| 59 |
+
stop_list = [f"{human_prefix}:"]
|
| 60 |
+
|
| 61 |
+
if "beluga" in model_loc.lower():
|
| 62 |
+
prompt_template = dedent("""
|
| 63 |
+
### System:
|
| 64 |
+
You are Stable Beluga, an AI that follows instructions extremely well. Help as much as you can.
|
| 65 |
+
Let's think step by step.
|
| 66 |
+
|
| 67 |
+
### User: {question}
|
| 68 |
+
|
| 69 |
+
### Assistant:
|
| 70 |
+
""").lstrip()
|
| 71 |
+
human_prefix = "### User"
|
| 72 |
+
ai_prefix = "### Assistant"
|
| 73 |
+
stop_list = [f"{human_prefix}:"]
|
| 74 |
+
|
| 75 |
+
_ = psutil.cpu_count(logical=False) - 1
|
| 76 |
+
cpu_count: int = int(_) if _ else 1
|
| 77 |
+
logger.debug(f"{cpu_count=}")
|
| 78 |
+
|
| 79 |
+
|
| 80 |
logger.debug(f"{model_loc=}")
|
| 81 |
LLM = AutoModelForCausalLM.from_pretrained(
|
| 82 |
model_loc,
|