Update app.py
Browse files
app.py
CHANGED
|
@@ -7,6 +7,7 @@ import gc
|
|
| 7 |
import ctypes
|
| 8 |
import multiprocessing as mp
|
| 9 |
from pathlib import Path
|
|
|
|
| 10 |
import pyarrow.parquet as pq
|
| 11 |
from tokenizers import Tokenizer
|
| 12 |
|
|
@@ -72,8 +73,9 @@ def tokenize_chunk(texts):
|
|
| 72 |
def process_shard(name, raw_path, pool):
|
| 73 |
print(f" [{WORKER_ID}] Processing: {name}")
|
| 74 |
|
| 75 |
-
out_name = name.replace(".parquet", ".
|
| 76 |
out_path = Path(OUT_DIR) / out_name
|
|
|
|
| 77 |
total_tokens = 0
|
| 78 |
|
| 79 |
try:
|
|
@@ -83,7 +85,7 @@ def process_shard(name, raw_path, pool):
|
|
| 83 |
return False, f"read_failed: {e}"
|
| 84 |
|
| 85 |
try:
|
| 86 |
-
with open(
|
| 87 |
for batch in pf.iter_batches(batch_size=5_000, columns=["text"]):
|
| 88 |
texts = batch.column("text").to_pylist()
|
| 89 |
mid = len(texts) // 2
|
|
@@ -91,18 +93,22 @@ def process_shard(name, raw_path, pool):
|
|
| 91 |
try:
|
| 92 |
results = pool.map(tokenize_chunk, [texts[:mid], texts[mid:]])
|
| 93 |
except Exception as e:
|
|
|
|
| 94 |
return False, f"tokenize_failed: {e}"
|
| 95 |
|
| 96 |
for ids in results[0] + results[1]:
|
| 97 |
-
|
|
|
|
| 98 |
total_tokens += len(ids)
|
| 99 |
|
| 100 |
del texts, results
|
| 101 |
gc.collect()
|
| 102 |
|
| 103 |
except Exception as e:
|
|
|
|
| 104 |
return False, f"write_failed: {e}"
|
| 105 |
|
|
|
|
| 106 |
print(f" ✓ [{WORKER_ID}] {out_name} | {total_tokens:,} tokens")
|
| 107 |
return True, None
|
| 108 |
|
|
@@ -141,7 +147,7 @@ def worker_loop():
|
|
| 141 |
|
| 142 |
total = len(state["shards"]) + len(state.get("queue", []))
|
| 143 |
done = sum(1 for v in state["shards"].values() if v["status"] == "done")
|
| 144 |
-
if
|
| 145 |
print(f" [{WORKER_ID}] All done. Sleeping.")
|
| 146 |
time.sleep(300)
|
| 147 |
continue
|
|
|
|
| 7 |
import ctypes
|
| 8 |
import multiprocessing as mp
|
| 9 |
from pathlib import Path
|
| 10 |
+
import numpy as np
|
| 11 |
import pyarrow.parquet as pq
|
| 12 |
from tokenizers import Tokenizer
|
| 13 |
|
|
|
|
| 73 |
def process_shard(name, raw_path, pool):
|
| 74 |
print(f" [{WORKER_ID}] Processing: {name}")
|
| 75 |
|
| 76 |
+
out_name = name.replace(".parquet", ".bin")
|
| 77 |
out_path = Path(OUT_DIR) / out_name
|
| 78 |
+
tmp_path = Path(OUT_DIR) / f"{out_name}.tmp"
|
| 79 |
total_tokens = 0
|
| 80 |
|
| 81 |
try:
|
|
|
|
| 85 |
return False, f"read_failed: {e}"
|
| 86 |
|
| 87 |
try:
|
| 88 |
+
with open(tmp_path, "wb") as f:
|
| 89 |
for batch in pf.iter_batches(batch_size=5_000, columns=["text"]):
|
| 90 |
texts = batch.column("text").to_pylist()
|
| 91 |
mid = len(texts) // 2
|
|
|
|
| 93 |
try:
|
| 94 |
results = pool.map(tokenize_chunk, [texts[:mid], texts[mid:]])
|
| 95 |
except Exception as e:
|
| 96 |
+
tmp_path.unlink(missing_ok=True)
|
| 97 |
return False, f"tokenize_failed: {e}"
|
| 98 |
|
| 99 |
for ids in results[0] + results[1]:
|
| 100 |
+
arr = np.array(ids, dtype=np.uint16)
|
| 101 |
+
arr.tofile(f)
|
| 102 |
total_tokens += len(ids)
|
| 103 |
|
| 104 |
del texts, results
|
| 105 |
gc.collect()
|
| 106 |
|
| 107 |
except Exception as e:
|
| 108 |
+
tmp_path.unlink(missing_ok=True)
|
| 109 |
return False, f"write_failed: {e}"
|
| 110 |
|
| 111 |
+
tmp_path.rename(out_path) # ← atomic, only visible when complete
|
| 112 |
print(f" ✓ [{WORKER_ID}] {out_name} | {total_tokens:,} tokens")
|
| 113 |
return True, None
|
| 114 |
|
|
|
|
| 147 |
|
| 148 |
total = len(state["shards"]) + len(state.get("queue", []))
|
| 149 |
done = sum(1 for v in state["shards"].values() if v["status"] == "done")
|
| 150 |
+
if total > 0 and done == total:
|
| 151 |
print(f" [{WORKER_ID}] All done. Sleeping.")
|
| 152 |
time.sleep(300)
|
| 153 |
continue
|