Instructions to use agentic-ptb/grok-record with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Grok
How to use agentic-ptb/grok-record with Grok:
# No code snippets available yet for this library. # To use this model, check the repository files and the library's documentation. # Want to help? PRs adding snippets are welcome at: # https://github.com/huggingface/huggingface.js
- Notebooks
- Google Colab
- Kaggle
File size: 6,468 Bytes
9b2f1cf | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 | #!/usr/bin/env python3
"""SFT rows: chain commands, don't repeat ls, stop after the artifact exists."""
from __future__ import annotations
import json
import random
from pathlib import Path
from datasets import Dataset, load_dataset
SYS = """You are an expert coding assistant operating inside pi.
Be concise. Prefer one bash call that chains steps with && or newlines.
Never repeat the same ls/find/pwd. After the required output exists, say one short sentence and STOP. Do not celebrate via more commands."""
TOOLS = json.dumps(
[
{
"type": "function",
"function": {
"name": "bash",
"description": "Execute a bash command",
"parameters": {
"type": "object",
"properties": {"command": {"type": "string"}},
"required": ["command"],
},
},
},
{
"type": "function",
"function": {
"name": "write",
"description": "Write a file",
"parameters": {
"type": "object",
"properties": {"path": {"type": "string"}, "content": {"type": "string"}},
"required": ["path", "content"],
},
},
},
{
"type": "function",
"function": {
"name": "read",
"description": "Read a file",
"parameters": {
"type": "object",
"properties": {"path": {"type": "string"}},
"required": ["path"],
},
},
},
{
"type": "function",
"function": {
"name": "edit",
"description": "Edit a file",
"parameters": {
"type": "object",
"properties": {"path": {"type": "string"}, "edits": {"type": "array"}},
"required": ["path", "edits"],
},
},
},
]
)
def tc(i, name, args):
return {"id": f"c{i}", "type": "function", "function": {"name": name, "arguments": json.dumps(args)}}
def asst(text, calls=None):
m = {"role": "assistant", "content": text}
if calls:
m["tool_calls"] = calls
return m
def tool(i, content):
return {"role": "tool", "tool_call_id": f"c{i}", "content": content}
def pack(msgs, src):
return {"messages": [{"role": "system", "content": SYS}] + msgs, "tools": TOOLS, "source": src}
def main():
rng = random.Random(11)
rows = []
for i in range(400):
rows.append(
pack(
[
{"role": "user", "content": f"Write READY to /output/status{i}.txt then stop."},
asst(
"Writing the file.",
[tc(0, "bash", {"command": f"mkdir -p /output && printf 'READY\\n' > /output/status{i}.txt && cat /output/status{i}.txt"})],
),
tool(0, "READY\n"),
asst("Wrote /output/status.txt."),
],
"stop:write",
)
)
for i in range(250):
rows.append(
pack(
[
{"role": "user", "content": "Explore once, then compile src/main.c to /output/app and stop."},
asst(
"Listing and compiling in one command.",
[tc(0, "bash", {"command": "ls -la && gcc -O2 src/main.c -o /output/app && ls -la /output/app"})],
),
tool(0, "-rwxr-xr-x 1 root root 16384 app\n"),
asst("Built /output/app."),
],
"stop:build",
)
)
for i in range(250):
rows.append(
pack(
[
{"role": "user", "content": "Fix the off-by-one in add.py and run the test. Stop when it passes."},
asst("Reading, patching, testing.", [tc(0, "bash", {"command": "cat add.py && python3 - <<'PY'\nfrom pathlib import Path\np=Path('add.py')\np.write_text(p.read_text().replace('a + b + 1','a + b'))\nPY\npython3 -c 'from add import add; assert add(2,2)==4'"})]),
tool(0, ""),
asst("Fixed add.py; test passed."),
],
"stop:fix",
)
)
# saw listing already -> do not list again
for i in range(200):
rows.append(
pack(
[
{"role": "user", "content": "There is a bug in util.py. Find it and patch it."},
asst("Listing once.", [tc(0, "bash", {"command": "ls && cat util.py"})]),
tool(0, "README.md\nutil.py\ndef inc(x):\n return x\n"),
asst(
"Patching without listing again.",
[
tc(
1,
"edit",
{"path": "util.py", "edits": [{"oldText": " return x\n", "newText": " return x + 1\n"}]},
)
],
),
tool(1, "Edited util.py"),
asst("inc now returns x+1."),
],
"stop:no-relist",
)
)
# mix prior finish data
prev = Path("/var/lib/agentptb-cache/tb1/tmp/sft-data/pi-sft-finish/train.parquet")
if prev.exists():
ds = load_dataset("parquet", data_files=str(prev), split="train")
idx = list(range(len(ds)))
rng.shuffle(idx)
for j in idx[:4000]:
rec = ds[j]
rows.append({"messages": rec["messages"], "tools": rec["tools"], "source": rec["source"]})
rng.shuffle(rows)
out = Path("/var/lib/agentptb-cache/tb1/tmp/sft-data/pi-sft-stop")
out.mkdir(parents=True, exist_ok=True)
Dataset.from_dict(
{
"messages": [r["messages"] for r in rows],
"tools": [r["tools"] if isinstance(r["tools"], str) else json.dumps(r["tools"]) for r in rows],
"source": [r["source"] for r in rows],
}
).to_parquet(str(out / "train.parquet"))
print("wrote", len(rows), out)
if __name__ == "__main__":
main()
|