Skip to content

not necessary kl term #6

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
idning wants to merge 3 commits into
base: main
Choose a base branch
from
Open

not necessary kl term #6

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
86f353a
not necessary kl_divergence
idning Mar 13, 2025
66082ec
wandb to None
idning Mar 13, 2025
a5413cf
update
idning Mar 13, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
30 changes: 16 additions & 14 deletions train.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,23 +1,24 @@
from collections.abc import Callable
import json
from pathlib import Path
import random
import re
from collections.abc import Callable
from pathlib import Path
from typing import Any, Iterator, Optional
import wandb

import torch
import torch.optim as optim
import torch.nn.functional as F
import torch.optim as optim
import wandb
from loss import GRPOLoss
from replay_buffer import Experience, join_experience_batch, ReplayBuffer
from torch.nn.utils import clip_grad_norm_
from torch.utils.data import DataLoader
from transformers import (
AutoTokenizer,
PreTrainedTokenizer,
LlamaForCausalLM,
GenerationConfig,
LlamaForCausalLM,
PreTrainedTokenizer,
)
from loss import approx_kl_divergence, GRPOLoss
from replay_buffer import ReplayBuffer, Experience, join_experience_batch


def load_model(
Expand Down Expand Up @@ -193,7 +194,7 @@ def read_prompts(

def main():
seed = 42
wandb_project = None # "tiny_grpo"
wandb_project = "tiny-grpo" # "tiny_grpo"
device_index = 0
model_name = "meta-llama/Llama-3.2-1B-Instruct"
checkpoint_path = Path("./output")
Expand Down Expand Up @@ -291,11 +292,12 @@ def main():
sequence_ids=sequence_ids,
attention_mask=attention_mask,
)
kl = approx_kl_divergence(
log_probs=log_probs,
log_probs_ref=log_probs_ref,
action_mask=action_mask,
)
# kl = approx_kl_divergence(
# log_probs=log_probs,
# log_probs_ref=log_probs_ref,
# action_mask=action_mask,
# )
kl = 0

experience = Experience(
sequences=sequence_ids,
Expand Down