Spaces:
Running on Zero
Running on Zero
Z User commited on
Commit ·
40b597c
1
Parent(s): c2cb3d9
LAL training worker 2
Browse files- README.md +6 -8
- app.py +46 -0
- requirements.txt +2 -0
README.md
CHANGED
|
@@ -1,13 +1,11 @@
|
|
| 1 |
---
|
| 2 |
-
title:
|
| 3 |
-
emoji:
|
| 4 |
-
colorFrom:
|
| 5 |
-
colorTo:
|
| 6 |
sdk: gradio
|
| 7 |
-
sdk_version:
|
| 8 |
-
python_version: '3.12'
|
| 9 |
app_file: app.py
|
| 10 |
pinned: false
|
| 11 |
---
|
| 12 |
-
|
| 13 |
-
Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
|
|
|
|
| 1 |
---
|
| 2 |
+
title: LAL Train 2
|
| 3 |
+
emoji: 🧠
|
| 4 |
+
colorFrom: blue
|
| 5 |
+
colorTo: green
|
| 6 |
sdk: gradio
|
| 7 |
+
sdk_version: "4.0.0"
|
|
|
|
| 8 |
app_file: app.py
|
| 9 |
pinned: false
|
| 10 |
---
|
| 11 |
+
LAL distributed training worker 2.
|
|
|
app.py
ADDED
|
@@ -0,0 +1,46 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import os, threading
|
| 2 |
+
import gradio as gr
|
| 3 |
+
from huggingface_hub import hf_hub_download
|
| 4 |
+
|
| 5 |
+
WORK_DIR = "/data/lalmodel"
|
| 6 |
+
os.makedirs(WORK_DIR, exist_ok=True)
|
| 7 |
+
|
| 8 |
+
def log(msg):
|
| 9 |
+
print(f"[SPACE] {msg}", flush=True)
|
| 10 |
+
|
| 11 |
+
def setup():
|
| 12 |
+
log("Setting up...")
|
| 13 |
+
if not os.path.exists(f"{WORK_DIR}/.git"):
|
| 14 |
+
os.system(f"git clone https://github.com/samaidev/lalmodel.git {WORK_DIR}")
|
| 15 |
+
if not os.path.exists(f"{WORK_DIR}/data/large_bpe_v3.bin"):
|
| 16 |
+
try:
|
| 17 |
+
p = hf_hub_download(repo_id="gasschina/lalmodel-code", filename="data/large_bpe_v3.bin", repo_type="model")
|
| 18 |
+
os.system(f"mkdir -p {WORK_DIR}/data && cp {p} {WORK_DIR}/data/")
|
| 19 |
+
except: pass
|
| 20 |
+
try:
|
| 21 |
+
p = hf_hub_download(repo_id="tchbcb/lalmodel-distributed", filename="merged_ckpt.ste", repo_type="model")
|
| 22 |
+
os.system(f"cp {p} {WORK_DIR}/model_dialogue.ste")
|
| 23 |
+
except: pass
|
| 24 |
+
if not os.path.exists(f"{WORK_DIR}/ste_train"):
|
| 25 |
+
os.system(f"cd {WORK_DIR} && make 2>&1 | tail -3")
|
| 26 |
+
os.system(f"cd {WORK_DIR} && OMP_NUM_THREADS=2 OMP_MAX_ACTIVE_LEVELS=1 nohup ./ste_train > train.log 2> train.err &")
|
| 27 |
+
log("Training started")
|
| 28 |
+
|
| 29 |
+
threading.Thread(target=setup, daemon=True).start()
|
| 30 |
+
|
| 31 |
+
def get_progress():
|
| 32 |
+
try:
|
| 33 |
+
with open(f"{WORK_DIR}/train.log") as f:
|
| 34 |
+
lines = f.readlines()[-10:]
|
| 35 |
+
return "```\n" + "".join(lines) + "\n```"
|
| 36 |
+
except:
|
| 37 |
+
return "Starting..."
|
| 38 |
+
|
| 39 |
+
with gr.Blocks() as demo:
|
| 40 |
+
gr.Markdown("# LAL Training Worker 2")
|
| 41 |
+
out = gr.Markdown(value="Starting...")
|
| 42 |
+
btn = gr.Button("Refresh")
|
| 43 |
+
btn.click(get_progress, outputs=out)
|
| 44 |
+
demo.load(get_progress, outputs=out)
|
| 45 |
+
|
| 46 |
+
demo.launch(server_name="0.0.0.0", server_port=7860)
|
requirements.txt
ADDED
|
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
|
|
|
| 1 |
+
gradio==4.0.0
|
| 2 |
+
huggingface_hub
|