from torch import Tensor
import torch
import torch.nn as nn
from torch.nn import functional as F
from trl.trainer.grpo_trainer import (
Any,
AutoModelForCausalLM,
AutoModelForSequenceClassification,
AutoTokenizer,
Dataset,
GRPOConfig,
GRPOTrainer,
GenerationConfig,
IterableDataset,
LLM,
Optional,
PeftConfig,
PreTrainedModel,
PreTrainedTokenizerBase,
RepeatRandomSampler,
RewardFunc,
Sampler,
SamplingParams,
SyncRefModelCallback,
Trainer,
TrainerCallback,
Union,
apply_chat_template,
broadcast_object_list,
create_reference_model,
defaultdict,
gather,
gather_object,
generate_model_card,
get_comet_experiment_url,
is_conversational,
is_deepspeed_zero3_enabled,
is_peft_model,
is_wandb_available,
maybe_apply_chat_template,
nn,
os,
pad,
patch,
prepare_deepspeed,
set_seed,
textwrap,
torch,
transformers,
unwrap_model_for_generation,
version,
wandb,
warnings,
os,
torch,
transformers,
Any,
LLM,
Union,
apply_chat_template,
broadcast_object_list,
gather,
gather_object,
is_conversational,
maybe_apply_chat_template,
nn,
os,
pad,
torch,
unwrap_model_for_generation,
wandb,
GRPOTrainer,
Trainer,
gather,
os,
torch,
)
import os
from typing import *
from dataclasses import dataclass, field
from packaging.version import Version
import torch
import numpy as np
from contextlib import nullcontext
from torch.nn import functional as F
torch_compile_options = {
"epilogue_fusion": True,
"max_autotune": False,
"shape_padding": True,
"trace.enabled": False,
"triton.cudagraphs": False,
}
@torch.compile(
dynamic=True,
fullgraph=True,
options=torch_compile_options,
)
def selective_log_softmax(logits, index):
logits = logits.to(torch.float32)
selected_logits = torch.gather(logits, dim=-1, index=index.unsqueeze(-1)).squeeze(
-1
)
# loop to reduce peak mem consumption
# logsumexp_values = torch.stack([torch.logsumexp(lg, dim=-1) for lg in logits])
logsumexp_values = torch.logsumexp(logits, dim=-1)
per_token_logps = (
selected_logits - logsumexp_values
) # log_softmax(x_i) = x_i - logsumexp(x)
return per_token_logps
def grpo_compute_loss(old_logits, new_logits, input_ids, mask, beta, advantages):
# All Unsloth Zoo code licensed under LGPLv3
old_logits = old_logits.to(torch.float32)
new_logits = new_logits.to(torch.float32)
input_ids = input_ids.unsqueeze(-1)
# x_i - logsumexp(x_i)
old_x = torch.gather(old_logits, dim=-1, index=input_ids).squeeze(-1)
new_x = torch.gather(new_logits, dim=-1, index=input_ids).squeeze(-1)
old = old_x - torch.logsumexp(old_logits, dim=-1)
new = new_x - torch.logsumexp(new_logits, dim=-1)
# Reverse KL
kl_i = torch.exp(old - new) - (old - new) - 1.0
# Full correct reverse KL divergence?? Missing term maybe?
# kl_i = torch.exp(new) * kl_i
# Below is forward KL (normal KL)
# kl_i = torch.exp(old) * (old - new)
# Must detach - otherwise gradients are not propagated correctly!
# exp(x - x) == 1
loss_i = torch.exp(new - new.detach()) * advantages.unsqueeze(1)
loss_i = -(loss_i - beta * kl_i)
mask = mask.to(torch.float32)
n_mask_per_reward = mask.sum(1)
# See https://github.com/huggingface/trl/pull/2881
# loss_per_reward = (loss_i * mask).sum(1) / n_mask_per_reward
# loss = loss_per_reward.mean()
loss = (loss_i * mask).sum() / mask.sum()
# Get metrics as well which are folded
with torch.inference_mode():
completion_length = n_mask_per_reward.mean()
mean_kl_per_reward = (kl_i * mask).sum(1) / n_mask_per_reward
mean_kl = mean_kl_per_reward.mean()
pass
return loss, completion_length, mean_kl
class UnslothEfficientGRPO(torch.autograd.Function):
# All Unsloth Zoo code licensed under LGPLv3
@staticmethod
def forward(
ctx,
_new_hidden_states,
_old_hidden_states,
lm_head,
_input_ids,
_mask,
_advantages,
beta,
scaler=None,
n_chunks=1,
):
def compute_loss(
new_hidden_states, old_hidden_states, input_ids, mask, advantages, scaling
):
new_logits = torch.matmul(new_hidden_states, lm_head.t())
new_logits = new_logits[
:, :-1, :
] # exclude the last logit: it corresponds to the next token pred
old_logits = torch.matmul(old_hidden_states, lm_head.t())
old_logits = old_logits[
:, :-1, :
] # exclude the last logit: it corresponds to the next token pred
loss, completion_length, mean_kl = grpo_compute_loss(
old_logits,
new_logits,
input_ids,
mask,
beta,
advantages,
)
# Scale loss if needed for mixed precision training
scaled_loss = loss * scaling
# Must add .loss.detach otherwise autograd uses 2x VRAM
return scaled_loss, (
loss.detach(),
completion_length,
mean_kl,
)
pass
device = _new_hidden_states.device
grad_inputs = torch.empty_like(_new_hidden_states)
accumulated_loss = torch.zeros(1, device=device)
accumulated_completion_length = torch.zeros(1, device=device)
accumulated_mean_kl = torch.zeros(1, device=device)
def accumulate_chunk(
new_hidden_states_j,
old_hidden_states_j,
input_ids_j,
mask_j,
advantages_j,
scaling,
):
(
(chunk_grad_input,),
(
chunk_loss,
(
unscaled_loss,
chunk_completion_length,
chunk_mean_kl,
),
),
) = torch.func.grad_and_value(
compute_loss,
argnums=(0,),
has_aux=True,
)(
new_hidden_states_j,
old_hidden_states_j,
input_ids_j,
mask_j,
advantages_j,
scaling,
)
accumulated_loss.add_(unscaled_loss)
accumulated_completion_length.add_(chunk_completion_length)
accumulated_mean_kl.add_(chunk_mean_kl)
return chunk_grad_input
pass
accumulate_chunk = torch.compile(
accumulate_chunk,
fullgraph=True,
options=torch_compile_options,
)
grad_inputs_chunks = torch.chunk(grad_inputs, chunks=n_chunks, dim=0)
new_hidden_states = torch.chunk(_new_hidden_states, chunks=n_chunks, dim=0)
old_hidden_states = torch.chunk(_old_hidden_states, chunks=n_chunks, dim=0)
input_ids = torch.chunk(_input_ids, chunks=n_chunks, dim=0)
mask = torch.chunk(_mask, chunks=n_chunks, dim=0)
advantages = torch.chunk(_advantages, chunks=n_chunks, dim=0)
# Get mixed precision scaling if seen
scaling = scaler.get_scale() if scaler is not None else 1.0
# Force torch.compile to use dynamic shapes for seqlen dim
mark_dynamic = lambda x: torch._dynamo.mark_dynamic(x, 1)
for (
grad_inputs_j,
new_hidden_states_j,
old_hidden_states_j,
input_ids_j,
mask_j,
advantages_j,
) in zip(
grad_inputs_chunks,
new_hidden_states,
old_hidden_states,
input_ids,
mask,
advantages,
):
mark_dynamic(new_hidden_states_j)
mark_dynamic(old_hidden_states_j)
mark_dynamic(input_ids_j)
mark_dynamic(mask_j)
grad_inputs_j.copy_(
accumulate_chunk(
new_hidden_states_j,
old_hidden_states_j,
input_ids_j,
mask_j,
advantages_j,
scaling,
)
)
pass
grad_inputs.div_(n_chunks)
accumulated_loss.div_(n_chunks)
accumulated_completion_length.div_(n_chunks)
accumulated_mean_kl.div_(n_chunks)
ctx.save_for_backward(grad_inputs)
return (
accumulated_loss,
accumulated_completion_length,
accumulated_mean_kl,
)
pass
@staticmethod
def backward(ctx, grad_output, dcompletion_length, dmean_kl):
(grad_input,) = ctx.saved_tensors
return (
grad_input,
None,
None,
None,
None,
None,
None,
None,
None,
)
pass
def grpo_accumulated_loss(
trainer,
input_ids,
logits_to_keep,
completion_mask,
advantages,
n_chunks=-1,
):
# All Unsloth Zoo code licensed under LGPLv3
bsz, qlen = input_ids.shape
# Find closest multiple
factors = [i for i in range(1, bsz + 1) if bsz % i == 0]
if n_chunks == -1:
n_chunks = bsz
n_chunks = factors[min(np.searchsorted(factors, n_chunks), len(factors) - 1)]
mixed_dtype = (
torch.float16
if os.environ.get("ACCELERATE_MIXED_PRECISION", "fp16") == "fp16"
else torch.bfloat16
)
os.environ["UNSLOTH_RETURN_HIDDEN_STATES"] = "1"
completion_input_ids = input_ids[:, -logits_to_keep:]
lm_head = trainer.model.get_output_embeddings().weight
with torch.amp.autocast(device_type="cuda", dtype=mixed_dtype):
with (
torch.inference_mode(),
trainer.accelerator.unwrap_model(
trainer.model, keep_fp32_wrapper=False
).disable_adapter(),
):
old_hidden_states = trainer.model(
input_ids=input_ids, logits_to_keep=logits_to_keep + 1
).logits
pass
new_hidden_states = trainer.model(
input_ids=input_ids, logits_to_keep=logits_to_keep + 1
).logits
loss, completion_length, mean_kl = UnslothEfficientGRPO.apply(
new_hidden_states,
old_hidden_states,
lm_head,
completion_input_ids,
completion_mask,
advantages,
trainer.beta,
trainer.accelerator.scaler,
n_chunks,
)
return loss, completion_length, mean_kl
# Old non efficient code path
new_logits = torch.matmul(new_hidden_states, lm_head.t())
new_logits = new_logits[
:, :-1, :
] # exclude the last logit: it corresponds to the next token pred
old_logits = torch.matmul(old_hidden_states, lm_head.t())
old_logits = old_logits[
:, :-1, :
] # exclude the last logit: it corresponds to the next token pred
loss, completion_length, mean_kl = grpo_compute_loss(
old_logits,
new_logits,
completion_input_ids,
completion_mask,
trainer.beta,
advantages,
)
return loss, completion_length, mean_kl
pass
def vLLMSamplingParams(**kwargs):
from vllm import SamplingParams
sampling_params = SamplingParams(**kwargs)
sampling_params._set_kwargs = kwargs
return sampling_params
@dataclass
class UnslothGRPOConfig(GRPOConfig):
"""
Configuration class for the [`GRPOTrainer`].
Only the parameters specific to GRPO training are listed here. For details on other parameters, refer to the
[`~transformers.TrainingArguments`] documentation.
Using [`~transformers.HfArgumentParser`] we can turn this class into
[argparse](https://docs.python.org/3/library/argparse#module-argparse) arguments that can be specified on the
command line.
Parameters:
> Parameters that control the model and reference model
model_init_kwargs (`dict[str, Any]` or `None`, *optional*, defaults to `None`):
Keyword arguments for [`~transformers.AutoModelForCausalLM.from_pretrained`], used when the `model`
argument of the [`GRPOTrainer`] is provided as a string.
> Parameters that control the data preprocessing
remove_unused_columns (`bool`, *optional*, defaults to `False`):
Whether to only keep the column `"prompt"` in the dataset. If you use a custom reward function that
requires any column other than `"prompts"` and `"completions"`, you should keep this to `False`.
max_prompt_length (`int` or `None`, *optional*, defaults to `512`):
Maximum length of the prompt. If the prompt is longer than this value, it will be truncated left.
num_generations (`int` or `None`, *optional*, defaults to `8`):
Number of generations per prompt to sample. The global batch size (num_processes * per_device_batch_size)
must be divisible by this value.
temperature (`float`, *optional*, defaults to `0.9`):
Temperature for sampling. The higher the temperature, the more random the completions.
max_completion_length (`int` or `None`, *optional*, defaults to `256`):
Maximum length of the generated completion.
ds3_gather_for_generation (`bool`, *optional*, defaults to `True`):
This setting applies to DeepSpeed ZeRO-3. If enabled, the policy model weights are gathered for generation,
improving generation speed. However, disabling this option allows training models that exceed the VRAM
capacity of a single GPU, albeit at the cost of slower generation. Disabling this option is not compatible
with vLLM generation.
> Parameters that control generation acceleration powered by vLLM
use_vllm (`bool`, *optional*, defaults to `False`):
Whether to use vLLM for generating completions. If set to `True`, ensure that a GPU is kept unused for
training, as vLLM will require one for generation. vLLM must be installed (`pip install vllm`).
vllm_device (`str`, *optional*, defaults to `"auto"`):
Device where vLLM generation will run, e.g. `"cuda:1"`. If set to `"auto"` (default), the system will
automatically select the next available GPU after the last one used for training. This assumes that
training has not already occupied all available GPUs. If only one device is available, the device will be
shared between both training and vLLM.
vllm_gpu_memory_utilization (`float`, *optional*, defaults to `0.9`):
Ratio (between 0 and 1) of GPU memory to reserve for the model weights, activations, and KV cache on the
device dedicated to generation powered by vLLM. Higher values will increase the KV cache size and thus
improve the model's throughput. However, if the value is too high, it may cause out-of-memory (OOM) errors
during initialization.
vllm_dtype (`str`, *optional*, defaults to `"auto"`):
Data type to use for vLLM generation. If set to `"auto"`, the data type will be automatically determined
based on the model configuration. Find the supported values in the vLLM documentation.
vllm_max_model_len (`int` or `None`, *optional*, defaults to `None`):
If set, the `max_model_len` to use for vLLM. This could be useful when running with reduced
`vllm_gpu_memory_utilization`, leading to a reduced KV cache size. If not set, vLLM will use the model
context size, which might be much larger than the KV cache, leading to inefficiencies.
> Parameters that control the training
learning_rate (`float`, *optional*, defaults to `1e-6`):
Initial learning rate for [`AdamW`] optimizer. The default value replaces that of
[`~transformers.TrainingArguments`].
beta (`float`, *optional*, defaults to `0.04`):
KL coefficient.
reward_weights (`list[float]` or `None`, *optional*, defaults to `None`):
Weights for each reward function. Must match the number of reward functions. If `None`, all rewards are
weighted equally with weight `1.0`.
sync_ref_model (`bool`, *optional*, defaults to `False`):
Whether to synchronize the reference model with the active model every `ref_model_sync_steps` steps, using
the `ref_model_mixup_alpha` parameter. This synchronization originates from the
[TR-DPO](https://huggingface.co/papers/2404.09656) paper.
ref_model_mixup_alpha (`float`, *optional*, defaults to `0.9`):
α parameter from the [TR-DPO](https://huggingface.co/papers/2404.09656) paper, which controls the mix
between the current policy and the previous reference policy during updates. The reference policy is
updated according to the equation: `π_ref = α * π_θ + (1 - α) * π_ref_prev`. To use this parameter, you
must set `sync_ref_model=True`.
ref_model_sync_steps (`int`, *optional*, defaults to `64`):
τ parameter from the [TR-DPO](https://huggingface.co/papers/2404.09656) paper, which determines how
frequently the current policy is synchronized with the reference policy. To use this parameter, you must
set `sync_ref_model=True`.
> Parameters that control the logging
log_completions (`bool`, *optional*, defaults to `False`):
Whether to log the completions during training.
"""
vllm_sampling_params: Optional[Any] = field(
default=None,
metadata={"help": "vLLM SamplingParams"},
)
unsloth_num_chunks: Optional[int] = field(
default=-1,
metadata={"help": "Chunk size to reduce memory usage. -1 is most efficient."},
)
def __init__(
self,
output_dir=None,
overwrite_output_dir=None,
do_train=False,
do_eval=False,
do_predict=False,
eval_strategy="no",
prediction_loss_only=False,
per_device_train_batch_size=4,
per_device_eval_batch_size=4,
per_gpu_train_batch_size=None,
per_gpu_eval_batch_size=None,
gradient_accumulation_steps=2,
eval_accumulation_steps=2,
eval_delay=0,
torch_empty_cache_steps=250,
learning_rate=5e-05,
weight_decay=0.01,
adam_beta1=0.9,
adam_beta2=0.999,
adam_epsilon=1e-08,
max_grad_norm=1.0,
num_train_epochs=3.0,
max_steps=-1,
lr_scheduler_type="linear",
warmup_ratio=0.1,
warmup_steps=0,
log_level="passive",
log_level_replica="warning",
log_on_each_node=True,
logging_dir=None,
logging_strategy="steps",
logging_first_step=False,
logging_steps=1,
logging_nan_inf_filter=False,
save_strategy="steps",
save_steps=500,
save_total_limit=None,
save_safetensors=True,
save_on_each_node=False,
save_only_model=False,
restore_callback_states_from_checkpoint=False,
no_cuda=False,
use_cpu=False,
use_mps_device=False,
seed=3407,
data_seed=3407,
jit_mode_eval=False,
use_ipex=False,
bf16=False,
fp16=False,
fp16_opt_level="O1",
half_precision_backend="auto",
bf16_full_eval=False,
fp16_full_eval=False,
tf32=None,
local_rank=-1,
ddp_backend=None,
tpu_num_cores=None,
tpu_metrics_debug=False,
debug="",
dataloader_drop_last=False,
eval_steps=None,
dataloader_num_workers=0,
dataloader_prefetch_factor=None,
past_index=-1,
run_name=None,
disable_tqdm=None,
remove_unused_columns=False,
label_names=None,
load_best_model_at_end=False,
metric_for_best_model=None,
greater_is_better=None,
ignore_data_skip=False,
fsdp="",
fsdp_min_num_params=0,
fsdp_config=None,
fsdp_transformer_layer_cls_to_wrap=None,
accelerator_config=None,
deepspeed=None,
label_smoothing_factor=0.0,
optim="adamw_8bit",
optim_args=None,
adafactor=False,
group_by_length=False,
length_column_name="length",
report_to=None,
ddp_find_unused_parameters=None,
ddp_bucket_cap_mb=None,
ddp_broadcast_buffers=None,
dataloader_pin_memory=True,
dataloader_persistent_workers=False,
skip_memory_metrics=True,
use_legacy_prediction_loop=False,
push_to_hub=False,
resume_from_checkpoint=None,
hub_model_id=None,
hub_strategy="every_save",
hub_token=None,
hub_private_repo=None,
hub_always_push=False,
gradient_checkpointing=False,
gradient_checkpointing_kwargs=None,
include_inputs_for_metrics=False,
eval_do_concat_batches=True,
fp16_backend="auto",
evaluation_strategy=None,
push_to_hub_model_id=None,
push_to_hub_organization=None,
push_to_hub_token=None,
mp_parameters="",
auto_find_batch_size=False,
full_determinism=False,
torchdynamo=None,
ray_scope="last",
ddp_timeout=1800,
torch_compile=False,
torch_compile_backend=None,
torch_compile_mode=None,
dispatch_batches=None,
split_batches=None,
include_tokens_per_second=False,
include_num_input_tokens_seen=False,
neftune_noise_alpha=None,
optim_target_modules=None,
batch_eval_metrics=False,
eval_on_start=False,
use_liger_kernel=False,
eval_use_gather_object=False,
average_tokens_across_devices=False,
model_init_kwargs=None,
max_prompt_length=512,
num_generations=8,
temperature=0.9,
max_completion_length=256,
ds3_gather_for_generation=True,
use_vllm=False,
use_agentic_generate=False,
vllm_device="auto",
vllm_gpu_memory_utilization=0.9,
vllm_dtype="auto",
vllm_max_model_len=None,
beta=0.04,
reward_weights=None,
sync_ref_model=False,
ref_model_mixup_alpha=0.9,
ref_model_sync_steps=64,
log_completions=False,
vllm_sampling_params=None,
unsloth_num_chunks=-1,
**kwargs,
):
if learning_rate < 1e-7:
raise FloatingPointError(
f"Unsloth: Your learning rate of `{learning_rate}` is too small and less than 1e-7! Consider increasing it, otherwise gradient updates will be close to 0!"
)
if learning_rate > 1:
raise OverflowError(
f"Unsloth: Your learning rate of `{learning_rate}` is way too larger > 1! Consider decreasing it to 1e-1, otherwise gradient updates will explode!"
)
if output_dir is None and save_strategy == "steps" and save_steps == 500:
output_dir = "unsloth_training_checkpoints"
save_strategy = "no"
div = per_device_train_batch_size // num_generations
if div * num_generations != per_device_train_batch_size:
print(
"Unsloth: We now expect `per_device_train_batch_size` to be a multiple of `num_generations`.\nWe will change the batch size of "
+ str(per_device_train_batch_size)
+ " to the `num_generations` of "
+ str(num_generations)
)
per_device_train_batch_size = num_generations
super().__init__(
output_dir=output_dir,
overwrite_output_dir=overwrite_output_dir,
do_train=do_train,
do_eval=do_eval,
do_predict=do_predict,
eval_strategy=eval_strategy,
prediction_loss_only=prediction_loss_only,
per_device_train_batch_size=per_device_train_batch_size,
per_device_eval_batch_size=per_device_eval_batch_size,
per_gpu_train_batch_size=per_gpu_train_batch_size,
per_gpu_eval_batch_size=per_gpu_eval_batch_size,
gradient_accumulation_steps=gradient_accumulation_steps,
eval_accumulation_steps=eval_accumulation_steps,
eval_delay=eval_delay,
torch_empty_cache_steps=torch_empty_cache_steps,
learning_rate=learning_rate,
weight_decay=weight_decay,
adam_beta1=adam_beta1,
adam_beta2=adam_beta2,
adam_epsilon=adam_epsilon,
max_grad_norm=max_grad_norm,
num_train_epochs=num_train_epochs,
max_steps=max_steps,
lr_scheduler_type=lr_scheduler_type,
warmup_ratio=warmup_ratio,
warmup_steps=warmup_steps,
log_level=log_level,
log_level_replica=log_level_replica,
log_on_each_node=log_on_each_node,
logging_dir=logging_dir,
logging_strategy=logging_strategy,
logging_first_step=logging_first_step,
logging_steps=logging_steps,
logging_nan_inf_filter=logging_nan_inf_filter,
save_strategy=save_strategy,
save_steps=save_steps,
save_total_limit=save_total_limit,
save_safetensors=save_safetensors,
save_on_each_node=save_on_each_node,
save_only_model=save_only_model,
restore_callback_states_from_checkpoint=restore_callback_states_from_checkpoint,
no_cuda=no_cuda,
use_cpu=use_cpu,
use_mps_device=use_mps_device,
seed=seed,
data_seed=data_seed,
jit_mode_eval=jit_mode_eval,
use_ipex=use_ipex,
bf16=bf16,
fp16=fp16,
fp16_opt_level=fp16_opt_level,
half_precision_backend=half_precision_backend,
bf16_full_eval=bf16_full_eval,
fp16_full_eval=fp16_full_eval,
tf32=tf32,
local_rank=local_rank,
ddp_backend=ddp_backend,
tpu_num_cores=tpu_num_cores,
tpu_metrics_debug=tpu_metrics_debug,
debug=debug,
dataloader_drop_last=dataloader_drop_last,
eval_steps=eval_steps,
dataloader_num_workers=dataloader_num_workers,
dataloader_prefetch_factor=dataloader_prefetch_factor,
past_index=past_index,
run_name=run_name,
disable_tqdm=disable_tqdm,
remove_unused_columns=remove_unused_columns,
label_names=label_names,
load_best_model_at_end=load_best_model_at_end,
metric_for_best_model=metric_for_best_model,
greater_is_better=greater_is_better,
ignore_data_skip=ignore_data_skip,
fsdp=fsdp,
fsdp_min_num_params=fsdp_min_num_params,
fsdp_config=fsdp_config,
fsdp_transformer_layer_cls_to_wrap=fsdp_transformer_layer_cls_to_wrap,
accelerator_config=accelerator_config,
deepspeed=deepspeed,
label_smoothing_factor=label_smoothing_factor,
optim=optim,
optim_args=optim_args,
adafactor=adafactor,
group_by_length=group_by_length,
length_column_name=length_column_name,
report_to=report_to,
ddp_find_unused_parameters=ddp_find_unused_parameters,
ddp_bucket_cap_mb=ddp_bucket_cap_mb,
ddp_broadcast_buffers=ddp_broadcast_buffers,
dataloader_pin_memory=dataloader_pin_memory,
dataloader_persistent_workers=dataloader_persistent_workers,
skip_memory_metrics=skip_memory_metrics,
use_legacy_prediction_loop=use_legacy_prediction_loop,
push_to_hub=push_to_hub,
resume_from_checkpoint=resume_from_checkpoint,
hub_model_id=hub_model_id,
hub_strategy=hub_strategy,
hub_token=hub_token,
hub_private_repo=hub_private_repo,
hub_always_push=hub_always_push,
gradient_checkpointing=gradient_checkpointing,
gradient_checkpointing_kwargs=gradient_checkpointing_kwargs,
include_inputs_for_metrics=include_inputs_for_metrics,
eval_do_concat_batches=eval_do_concat_batches,
fp16_backend=fp16_backend,
evaluation_strategy=evaluation_strategy,
push_to_hub_model_id=push_to_hub_model_id,
push_to_hub_organization=push_to_hub_organization,
push_to_hub_token=push_to_hub_token,
mp_parameters=mp_parameters,
auto_find_batch_size=auto_find_batch_size,
full_determinism=full_determinism,
torchdynamo=torchdynamo,
ray_scope=ray_scope,
ddp_timeout=ddp_timeout,
torch_compile=torch_compile,
torch_compile_backend=torch_compile_backend,
torch_compile_mode=torch_compile_mode,
dispatch_batches=dispatch_batches,
split_batches=split_batches,
include_tokens_per_second=include_tokens_per_second,
include_num_input_tokens_seen=include_num_input_tokens_seen,
neftune_noise_alpha=neftune_noise_alpha,
optim_target_modules=optim_target_modules,
batch_eval_metrics=batch_eval_metrics,
eval_on_start=eval_on_start,
use_liger_kernel=use_liger_kernel,
eval_use_gather_object=eval_use_gather_object,
average_tokens_across_devices=average_tokens_across_devices,
model_init_kwargs=model_init_kwargs,
max_prompt_length=max_prompt_length,
num_generations=num_generations,
temperature=temperature,
max_completion_length=max_completion_length,
ds3_gather_for_generation=ds3_gather_for_generation,
use_vllm=use_vllm,
vllm_device=vllm_device,
vllm_gpu_memory_utilization=vllm_gpu_memory_utilization,
vllm_dtype=vllm_dtype,
vllm_max_model_len=vllm_max_model_len,
beta=beta,
reward_weights=reward_weights,
sync_ref_model=sync_ref_model,
ref_model_mixup_alpha=ref_model_mixup_alpha,
ref_model_sync_steps=ref_model_sync_steps,
log_completions=log_completions,
**kwargs,
)
self.vllm_sampling_params = vllm_sampling_params
self.unsloth_num_chunks = unsloth_num_chunks
self.use_agentic_generate = use_agentic_generate
pass
class _UnslothGRPOTrainer(Trainer):
""""""
_tag_names = ["trl", "grpo"]
def __init__(
self,
model: Union[str, PreTrainedModel],
reward_funcs: Union[RewardFunc, list[RewardFunc]],
args: GRPOConfig = None,
train_dataset: Optional[Union[Dataset, IterableDataset]] = None,
eval_dataset: Optional[
Union[Dataset, IterableDataset, dict[str, Union[Dataset, IterableDataset]]]
] = None,
processing_class: Optional[PreTrainedTokenizerBase] = None,
reward_processing_classes: Optional[
Union[PreTrainedTokenizerBase, list[PreTrainedTokenizerBase]]
] = None,
callbacks: Optional[list[TrainerCallback]] = None,
optimizers: tuple[
Optional[torch.optim.Optimizer], Optional[torch.optim.lr_scheduler.LambdaLR]
] = (None, None),
peft_config: Optional["PeftConfig"] = None,
):
if (
hasattr(model, "vllm_engine")
and hasattr(args, "use_vllm")
and (getattr(args, "use_vllm", False) == False)
):
args.use_vllm = True
# Args
if args is None:
model_name = model if isinstance(model, str) else model.config._name_or_path
model_name = model_name.split("/")[-1]
args = GRPOConfig(f"{model_name}-GRPO")
# Models
# Trained model
model_init_kwargs = args.model_init_kwargs or {}
if isinstance(model, str):
model_id = model
torch_dtype = model_init_kwargs.get("torch_dtype")
if (
isinstance(torch_dtype, torch.dtype)
or torch_dtype == "auto"
or torch_dtype is None
):
pass # torch_dtype is already a torch.dtype or "auto" or None
elif isinstance(torch_dtype, str): # it's a str, but not "auto"
torch_dtype = getattr(torch, torch_dtype)
model_init_kwargs["torch_dtype"] = torch_dtype
else:
raise ValueError(
"Invalid `torch_dtype` passed to `GRPOConfig`. Expected either 'auto' or a string representing "
f"a `torch.dtype` (e.g., 'float32'), but got {torch_dtype}."
)
# Disable caching if gradient checkpointing is enabled (not supported)
model_init_kwargs["use_cache"] = (
False
if args.gradient_checkpointing
else model_init_kwargs.get("use_cache")
)
model = AutoModelForCausalLM.from_pretrained(model, **model_init_kwargs)
else:
model_id = model.config._name_or_path
if args.model_init_kwargs is not None:
raise ValueError(
"You passed `model_init_kwargs` to the `GRPOConfig`, but your model is already instantiated. "
"This argument can only be used when the `model` argument is a string."
)
if False:
model = model
# Reference model
if is_deepspeed_zero3_enabled():
self.ref_model = AutoModelForCausalLM.from_pretrained(
model_id, **model_init_kwargs
)
elif not is_peft_model(model):
# If PEFT configuration is not provided, create a reference model based on the initial model.
self.ref_model = create_reference_model(model)
else:
# If PEFT is used, the reference model is not needed since the adapter can be disabled
# to revert to the initial model.
self.ref_model = None
# Processing class
if processing_class is None:
processing_class = AutoTokenizer.from_pretrained(
model.config._name_or_path, padding_side="left"
)
# Reward functions
if not isinstance(reward_funcs, list):
reward_funcs = [reward_funcs]
for i, reward_func in enumerate(reward_funcs):
if isinstance(reward_func, str):
reward_funcs[i] = AutoModelForSequenceClassification.from_pretrained(
reward_func, num_labels=1, **model_init_kwargs
)
self.reward_funcs = reward_funcs
# Reward weights
if args.reward_weights is not None:
if len(args.reward_weights) != len(reward_funcs):
raise ValueError(
f"Number of reward weights ({len(args.reward_weights)}) must match number of reward "
f"functions ({len(reward_funcs)})"
)
self.reward_weights = torch.tensor(args.reward_weights, dtype=torch.float32)
else:
self.reward_weights = torch.ones(len(reward_funcs), dtype=torch.float32)
# Reward processing class
if reward_processing_classes is None:
reward_processing_classes = [None] * len(reward_funcs)
elif not isinstance(reward_processing_classes, list):
reward_processing_classes = [reward_processing_classes]
else:
if len(reward_processing_classes) != len(reward_funcs):
raise ValueError(
"The number of reward processing classes must match the number of reward functions."
)
for i, (reward_processing_class, reward_func) in enumerate(
zip(reward_processing_classes, reward_funcs)
):
if isinstance(reward_func, PreTrainedModel):
if reward_processing_class is None:
reward_processing_class = AutoTokenizer.from_pretrained(
reward_func.config._name_or_path
)
if reward_processing_class.pad_token_id is None:
reward_processing_class.pad_token = (
reward_processing_class.eos_token
)
# The reward model computes the reward for the latest non-padded token in the input sequence.
# So it's important to set the pad token ID to the padding token ID of the processing class.
reward_func.config.pad_token_id = reward_processing_class.pad_token_id
reward_processing_classes[i] = reward_processing_class
self.reward_processing_classes = reward_processing_classes
# Data collator
def data_collator(features): # No data collation is needed in GRPO
return features
# Training arguments
self.max_prompt_length = args.max_prompt_length
self.max_completion_length = (
args.max_completion_length
) # = |o_i| in the GRPO paper
self.num_generations = args.num_generations # = G in the GRPO paper
self.use_vllm = args.use_vllm
self.use_agentic_generate = args.use_agentic_generate
self.beta = args.beta
# The trainer estimates the number of FLOPs (floating-point operations) using the number of elements in the
# input tensor associated with the key "input_ids". However, in GRPO, the sampled data does not include the
# "input_ids" key. Instead, the available keys is "prompt". As a result, the trainer issues the warning:
# "Could not estimate the number of tokens of the input, floating-point operations will not be computed." To
# suppress this warning, we set the "estimate_tokens" key in the model's "warnings_issued" dictionary to True.
# This acts as a flag to indicate that the warning has already been issued.
model.warnings_issued["estimate_tokens"] = True
# Initialize the metrics
self._metrics = defaultdict(list)
self.log_completions = args.log_completions
super().__init__(
model=model,
args=args,
data_collator=data_collator,
train_dataset=train_dataset,
eval_dataset=eval_dataset,
processing_class=processing_class,
callbacks=callbacks,
optimizers=optimizers,
)
# Check if the per_device_train/eval_batch_size * num processes can be divided by the number of generations
num_processes = self.accelerator.num_processes
global_batch_size = args.per_device_train_batch_size * num_processes
possible_values = [
n_gen
for n_gen in range(2, global_batch_size + 1)
if (global_batch_size) % n_gen == 0
]
if self.num_generations not in possible_values:
raise ValueError(
f"The global train batch size ({num_processes} x {args.per_device_train_batch_size}) must be evenly "
f"divisible by the number of generations per prompt ({self.num_generations}). Given the current train "
f"batch size, the valid values for the number of generations are: {possible_values}."
)
if self.args.eval_strategy != "no":
global_batch_size = args.per_device_eval_batch_size * num_processes
possible_values = [
n_gen
for n_gen in range(2, global_batch_size + 1)
if (global_batch_size) % n_gen == 0
]
if self.num_generations not in possible_values:
raise ValueError(
f"The global eval batch size ({num_processes} x {args.per_device_eval_batch_size}) must be evenly "
f"divisible by the number of generations per prompt ({self.num_generations}). Given the current "
f"eval batch size, the valid values for the number of generations are: {possible_values}."
)
# Ensure each process receives a unique seed to prevent duplicate completions when generating with
# transformers if num_generations exceeds per_device_train_batch_size. We could skip it if we use vLLM, but
# it's safer to set it in all cases.
set_seed(args.seed, device_specific=True)
if self.use_vllm:
self.llm = model.vllm_engine
self._last_loaded_step = 0
self.sampling_params = SamplingParams(
temperature=args.temperature,
max_tokens=self.max_completion_length,
**getattr(
getattr(args, "vllm_sampling_params", vLLMSamplingParams()),
"_set_kwargs",
{},
),
)
else:
self.generation_config = GenerationConfig(
max_new_tokens=self.max_completion_length,
do_sample=True,
temperature=args.temperature,
pad_token_id=processing_class.pad_token_id,
)
# Gradient accumulation requires scaled loss. Normally, loss scaling in the parent class depends on whether the
# model accepts loss-related kwargs. Since we compute our own loss, this check is irrelevant. We set
# self.model_accepts_loss_kwargs to False to enable scaling.
self.model_accepts_loss_kwargs = False
# Add tags to the model
self.model.add_model_tags(self._tag_names)
if self.ref_model is not None:
if self.is_deepspeed_enabled:
self.ref_model = prepare_deepspeed(self.ref_model, self.accelerator)
else:
self.ref_model = self.accelerator.prepare_model(
self.ref_model, evaluation_mode=True
)
if args.sync_ref_model:
self.add_callback(
SyncRefModelCallback(
ref_model=self.ref_model, accelerator=self.accelerator
)
)
for i, reward_func in enumerate(self.reward_funcs):
if isinstance(reward_func, PreTrainedModel):
self.reward_funcs[i] = self.accelerator.prepare_model(
reward_func, evaluation_mode=True
)
def _set_signature_columns_if_needed(self):
# If `self.args.remove_unused_columns` is True, non-signature columns are removed.
# By default, this method sets `self._signature_columns` to the model's expected inputs.
# In GRPOTrainer, we preprocess data, so using the model's signature columns doesn't work.
# Instead, we set them to the columns expected by the `training_step` method, hence the override.
if self._signature_columns is None:
self._signature_columns = ["prompt"]
def _get_train_sampler(self) -> Sampler:
# Returns a sampler that ensures each prompt is repeated across multiple processes. This guarantees that
# identical prompts are distributed to different GPUs, allowing rewards to be computed and normalized correctly
# within each prompt group. Using the same seed across processes ensures consistent prompt assignment,
# preventing discrepancies in group formation.
return RepeatRandomSampler(
self.train_dataset, self.num_generations, seed=self.args.seed
)
def _get_eval_sampler(self, eval_dataset) -> Sampler:
# Returns a sampler that ensures each prompt is repeated across multiple processes. This guarantees that
# identical prompts are distributed to different GPUs, allowing rewards to be computed and normalized correctly
# within each prompt group. Using the same seed across processes ensures consistent prompt assignment,
# preventing discrepancies in group formation.
return RepeatRandomSampler(
eval_dataset, self.num_generations, seed=self.args.seed
)
# Get the per-token log probabilities for the completions for the model and the reference model
def _get_per_token_logps(self, model, input_ids, attention_mask, logits_to_keep):
return None # Unsloth efficient GRPO
if not hasattr(self, "_autocast_dtype"):
self._autocast_dtype = (
torch.float16
if os.environ.get("ACCELERATE_MIXED_PRECISION", "fp16") == "fp16"
else torch.bfloat16
)
with torch.amp.autocast(device_type="cuda", dtype=self._autocast_dtype):
# We add 1 to `logits_to_keep` because the last logits of the sequence is later excluded
logits = model(
input_ids=input_ids,
attention_mask=attention_mask,
logits_to_keep=logits_to_keep + 1,
).logits
logits = logits[
:, :-1, :
] # (B, L-1, V), exclude the last logit: it corresponds to the next token pred
input_ids = input_ids[:, -logits_to_keep:]
# For transformers<=4.48, logits_to_keep argument isn't supported, so here we drop logits ourselves.
# See https://github.com/huggingface/trl/issues/2770
logits = logits[:, -logits_to_keep:]
return logits
# return selective_log_softmax(logits, input_ids) # compute logprobs for the input tokens
pass
def _move_model_to_vllm(self, *args, **kwargs):
return None
def _prepare_inputs(
self, inputs: dict[str, Union[torch.Tensor, Any]]
) -> dict[str, Union[torch.Tensor, Any]]:
device = self.accelerator.device
prompts = [x["prompt"] for x in inputs]
prompts_text = [
maybe_apply_chat_template(example, self.processing_class)["prompt"]
for example in inputs
]
prompt_inputs = self.processing_class(
prompts_text,
return_tensors="pt",
padding=True,
padding_side="left",
add_special_tokens=False,
)
prompt_inputs = super()._prepare_inputs(prompt_inputs)
prompt_ids, prompt_mask = (
prompt_inputs["input_ids"],
prompt_inputs["attention_mask"],
)
if self.max_prompt_length is not None:
prompt_ids = prompt_ids[:, -self.max_prompt_length :]
prompt_mask = prompt_mask[:, -self.max_prompt_length :]
# Generate completions using either vLLM or regular generation
if self.args.use_vllm:
# First, have main process load weights if needed
if self.state.global_step != self._last_loaded_step:
self._move_model_to_vllm()
self._last_loaded_step = self.state.global_step
# Generate completions using vLLM: gather all prompts and use them in a single call in the main process
all_prompts_text = gather_object(prompts_text)
if self.accelerator.is_main_process:
print(all_prompts_text)
generate_fn = lambda prompts_text: self.llm.generate(
prompts_text,
sampling_params=self.sampling_params,
use_tqdm=False,
lora_request=self.model.load_lora(
"grpo_trainer_lora_model", load_tensors=True
),
)
if self.use_agentic_generate:
agentic_outputs = self.model.agentic_generate(
all_prompts_text,
generate_fn,
)
full_chats = agentic_outputs.full_chat_states
final_responses = agentic_outputs.final_response_str
# TODO: apply truncation here
prompt_inputs = agentic_outputs.prompt_tokens
completion_ids = agentic_outputs.response_tokens
completion_mask = agentic_outputs.response_masks
prompt_ids = pad(
prompt_inputs,
padding_value=self.processing_class.pad_token_id,
padding_side="left",
).to(device)
completion_mask = pad(
completion_mask,
padding_value=0,
padding_side="right",
).to(device)
else:
outputs = generate_fn(all_prompts_text)
completion_ids = [
out.token_ids
for completions in outputs
for out in completions.outputs
]
else:
completion_ids = [None] * len(all_prompts_text)
# Broadcast the completions from the main process to all processes, ensuring each process receives its
# corresponding slice.
completion_ids = broadcast_object_list(completion_ids, from_process=0)
process_slice = slice(
self.accelerator.process_index * len(prompts),
(self.accelerator.process_index + 1) * len(prompts),
)
completion_ids = completion_ids[process_slice]
# Pad the completions, and concatenate them with the prompts
completion_ids = [
torch.tensor(ids, device=device) for ids in completion_ids
]
completion_ids = pad(
completion_ids, padding_value=self.processing_class.pad_token_id
)
prompt_completion_ids = torch.cat([prompt_ids, completion_ids], dim=1)
else:
# Regular generation path
with unwrap_model_for_generation(
self.model, self.accelerator
) as unwrapped_model:
prompt_completion_ids = unwrapped_model.generate(
prompt_ids,
attention_mask=prompt_mask,
generation_config=self.generation_config,
)
# Compute prompt length and extract completion ids
prompt_length = prompt_ids.size(1)
prompt_ids = prompt_completion_ids[:, :prompt_length]
completion_ids = prompt_completion_ids[:, prompt_length:]
if not self.use_agentic_generate:
# Mask everything after the first EOS token
is_eos = completion_ids == self.processing_class.eos_token_id
eos_idx = torch.full(
(is_eos.size(0),), is_eos.size(1), dtype=torch.long, device=device
)
eos_idx[is_eos.any(dim=1)] = is_eos.int().argmax(dim=1)[is_eos.any(dim=1)]
sequence_indices = torch.arange(is_eos.size(1), device=device).expand(
is_eos.size(0), -1
)
completion_mask = (sequence_indices <= eos_idx.unsqueeze(1)).int()
# Concatenate prompt_mask with completion_mask for logit computation
attention_mask = torch.cat([prompt_mask, completion_mask], dim=1) # (B*G, P+C)
logits_to_keep = completion_ids.size(
1
) # we only need to compute the logits for the completion tokens
# this does nothing
with (
torch.inference_mode(),
torch.amp.autocast(
device_type="cuda",
dtype=torch.float16
if os.environ.get("ACCELERATE_MIXED_PRECISION", "fp16") == "fp16"
else torch.bfloat16,
)
if not torch.is_autocast_enabled("cuda")
else nullcontext(),
):
if self.ref_model is not None:
ref_per_token_logps = self._get_per_token_logps(
self.ref_model,
prompt_completion_ids,
attention_mask,
logits_to_keep,
)
else:
with self.accelerator.unwrap_model(
self.model, keep_fp32_wrapper=False
).disable_adapter():
ref_per_token_logps = self._get_per_token_logps(
self.model,
prompt_completion_ids,
attention_mask,
logits_to_keep,
)
# Decode the generated completions
if not self.use_agentic_generate:
completions_text = self.processing_class.batch_decode(
completion_ids, skip_special_tokens=True
)
if is_conversational(inputs[0]):
completions = []
for prompt, completion in zip(prompts, completions_text):
bootstrap = (
prompt.pop()["content"]
if prompt[-1]["role"] == "assistant"
else ""
)
completions.append(
[{"role": "assistant", "content": bootstrap + completion}]
)
else:
completions = completions_text
else:
completions = full_chats
rewards_per_func = torch.zeros(
len(prompts), len(self.reward_funcs), device=device
)
for i, (reward_func, reward_processing_class) in enumerate(
zip(self.reward_funcs, self.reward_processing_classes)
):
if isinstance(
reward_func, nn.Module
): # Module instead of PretrainedModel for compat with compiled models
if is_conversational(inputs[0]):
messages = [
{"messages": p + c} for p, c in zip(prompts, completions)
]
texts = [
apply_chat_template(x, reward_processing_class)["text"]
for x in messages
]
else:
texts = [p + c for p, c in zip(prompts, completions)]
reward_inputs = reward_processing_class(
texts,
return_tensors="pt",
padding=True,
padding_side="right",
add_special_tokens=False,
)
reward_inputs = super()._prepare_inputs(reward_inputs)
with (
torch.inference_mode(),
torch.amp.autocast(
device_type="cuda",
dtype=torch.float16
if os.environ.get("ACCELERATE_MIXED_PRECISION", "fp16")
== "fp16"
else torch.bfloat16,
)
if not torch.is_autocast_enabled("cuda")
else nullcontext(),
):
rewards_per_func[:, i] = reward_func(**reward_inputs).logits[
:, 0
] # Shape (B*G,)
else:
# Repeat all input columns (but "prompt" and "completion") to match the number of generations
keys = [key for key in inputs[0] if key not in ["prompt", "completion"]]
reward_kwargs = {
key: [example[key] for example in inputs] for key in keys
}
output_reward_func = reward_func(
prompts=prompts, completions=completions, **reward_kwargs
)
rewards_per_func[:, i] = torch.tensor(
output_reward_func, dtype=torch.float32, device=device
)
# Gather the reward per function: this part is crucial, because the rewards are normalized per group and the
# completions may be distributed across processes
rewards_per_func = gather(rewards_per_func)
# Apply weights to each reward function's output and sum
rewards = (rewards_per_func * self.reward_weights.to(device).unsqueeze(0)).sum(
dim=1
)
# else:
# reward_fn = self.reward_funcs[0]
# rewards = reward_fn(
# prompts=prompts_text,
# student_answers=final_responses,
# answers=[input["answer"] for input in inputs],
# )
# rewards = torch.tensor(rewards, dtype=torch.float16, device=device)
# rewards_per_func = rewards.unsqueeze(1)
# Compute grouped-wise rewards
mean_grouped_rewards = rewards.view(-1, self.num_generations).mean(dim=1)
std_grouped_rewards = rewards.view(-1, self.num_generations).std(dim=1)
# Normalize the rewards to compute the advantages
mean_grouped_rewards = mean_grouped_rewards.repeat_interleave(
self.num_generations, dim=0
)
std_grouped_rewards = std_grouped_rewards.repeat_interleave(
self.num_generations, dim=0
)
advantages = (rewards - mean_grouped_rewards) / (std_grouped_rewards + 1e-4)
# Slice to keep only the local part of the data
process_slice = slice(
self.accelerator.process_index * len(prompts),
(self.accelerator.process_index + 1) * len(prompts),
)
advantages = advantages[process_slice]
# Log the metrics
reward_per_func = rewards_per_func.mean(0)
print("rewards_per_func:", reward_per_func)
for i, reward_func in enumerate(self.reward_funcs):
if isinstance(
reward_func, nn.Module
): # Module instead of PretrainedModel for compat with compiled models
reward_func_name = reward_func.config._name_or_path.split("/")[-1]
else:
reward_func_name = reward_func.__name__
self._metrics[f"rewards/{reward_func_name}"].append(
reward_per_func[i].item()
)
self._metrics["reward"].append(rewards.mean().item())
self._metrics["reward_std"].append(std_grouped_rewards.mean().item())
if (
self.log_completions
and self.state.global_step % self.args.logging_steps == 0
and "wandb" in self.args.report_to
):
import pandas as pd
# For logging
table = {
"step": [str(self.state.global_step)] * len(rewards),
"prompt": gather_object(prompts_text),
"completion": gather_object(completions_text),
"reward": rewards.tolist(),
}
df = pd.DataFrame(table)
if wandb.run is not None and self.accelerator.is_main_process:
wandb.log({"completions": wandb.Table(dataframe=df)})
return {
"prompt_ids": prompt_ids,
"prompt_mask": prompt_mask,
"completion_ids": completion_ids,
"completion_mask": completion_mask,
"ref_per_token_logps": ref_per_token_logps,
"advantages": advantages,
}
def compute_loss(
self, model, inputs, return_outputs=False, num_items_in_batch=None
):
if return_outputs:
raise ValueError("The GRPOTrainer does not support returning outputs")
# Compute the per-token log probabilities for the model
prompt_ids, prompt_mask = inputs["prompt_ids"], inputs["prompt_mask"]
completion_ids, completion_mask = (
inputs["completion_ids"],
inputs["completion_mask"],
)
input_ids = torch.cat([prompt_ids, completion_ids], dim=1)
bsz, qlen = input_ids.shape
# attention_mask = torch.cat([prompt_mask, completion_mask], dim=1)
attention_mask = None
logits_to_keep = completion_ids.size(
1
) # we only need to compute the logits for the completion tokens
_input_ids = input_ids
_logits_to_keep = logits_to_keep
per_token_logps = self._get_per_token_logps(
model, input_ids, attention_mask, logits_to_keep
)
# Compute the KL divergence between the model and the reference model
ref_per_token_logps = inputs["ref_per_token_logps"]
# per_token_kl = torch.exp(ref_per_token_logps - per_token_logps) - (ref_per_token_logps - per_token_logps) - 1
# x - x.detach() allows for preserving gradients from x
advantages = inputs["advantages"]
# per_token_loss = torch.exp(per_token_logps - per_token_logps.detach()) * advantages.unsqueeze(1)
# per_token_loss = -(per_token_loss - self.beta * per_token_kl)
# loss = ((per_token_loss * completion_mask).sum(dim=1) / completion_mask.sum(dim=1)).mean()
input_ids = input_ids[:, -logits_to_keep:]
if False: # per_token_logps is not None:
loss, completion_length, mean_kl = grpo_compute_loss(
ref_per_token_logps,
per_token_logps,
input_ids,
completion_mask,
self.beta,
advantages,
)
else:
loss, completion_length, mean_kl = grpo_accumulated_loss(
self,
_input_ids,
logits_to_keep,
completion_mask,
advantages,
n_chunks=self.args.unsloth_num_chunks,
)
# Log the metrics
# completion_length = self.accelerator.gather_for_metrics(completion_mask.sum(1)).float().mean().item()
self._metrics["completion_length"].append(completion_length.item())
# mean_kl = ((per_token_kl * completion_mask).sum(dim=1) / completion_mask.sum(dim=1)).mean()
# self._metrics["kl"].append(self.accelerator.gather_for_metrics(mean_kl).mean().item())
self._metrics["kl"].append(mean_kl.item())
return loss
def prediction_step(
self,
model,
inputs,
prediction_loss_only,
ignore_keys: Optional[list[str]] = None,
):
inputs = self._prepare_inputs(inputs)
with torch.no_grad():
with self.compute_loss_context_manager():
loss = self.compute_loss(model, inputs)
loss = loss.mean().detach()
return loss, None, None
def log(self, logs: dict[str, float], start_time: Optional[float] = None) -> None:
metrics = {
key: sum(val) / len(val) for key, val in self._metrics.items()
} # average the metrics
# This method can be called both in training and evaluation. When called in evaluation, the keys in `logs`
# start with "eval_". We need to add the prefix "eval_" to the keys in `metrics` to match the format.
if next(iter(logs.keys())).startswith("eval_"):
metrics = {f"eval_{key}": val for key, val in metrics.items()}
logs = {**logs, **metrics}
if version.parse(transformers.__version__) >= version.parse("4.47.0.dev0"):
super().log(logs, start_time)
else: # transformers<=4.46
super().log(logs)
self._metrics.clear()
def create_model_card(
self,
model_name: Optional[str] = None,
dataset_name: Optional[str] = None,
tags: Union[str, list[str], None] = None,
):
"""
Creates a draft of a model card using the information available to the `Trainer`.
Args:
model_name (`str` or `None`, *optional*, defaults to `None`):
Name of the model.
dataset_name (`str` or `None`, *optional*, defaults to `None`):
Name of the dataset used for training.
tags (`str`, `list[str]` or `None`, *optional*, defaults to `None`):
Tags to be associated with the model card.
"""
if not self.is_world_process_zero():
return
if hasattr(self.model.config, "_name_or_path") and not os.path.isdir(
self.model.config._name_or_path
):
base_model = self.model.config._name_or_path
else:
base_model = None
tags = tags or []
if isinstance(tags, str):
tags = [tags]
if hasattr(self.model.config, "unsloth_version"):
tags.append("unsloth")
citation = textwrap.dedent(
"""\
@article{zhihong2024deepseekmath,
title = {{DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models}},
author = {Zhihong Shao and Peiyi Wang and Qihao Zhu and Runxin Xu and Junxiao Song and Mingchuan Zhang and Y. K. Li and Y. Wu and Daya Guo},
year = 2024,
eprint = {arXiv:2402.03300},
}
"""
)
model_card = generate_model_card(
base_model=base_model,
model_name=model_name,
hub_model_id=self.hub_model_id,
dataset_name=dataset_name,
tags=tags,
wandb_url=wandb.run.get_url()
if is_wandb_available() and wandb.run is not None
else None,
comet_url=get_comet_experiment_url(),
trainer_name="GRPO",
trainer_citation=citation,
paper_title="DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models",
paper_id="2402.03300",
)
model_card.save(os.path.join(self.args.output_dir, "README.md"))
class UnslothGRPOTrainer(_UnslothGRPOTrainer):
"""
Trainer for the Group Relative Policy Optimization (GRPO) method. This algorithm was initially proposed in the
paper [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://huggingface.co/papers/2402.03300).
Example:
```python
from datasets import load_dataset
from trl import GRPOTrainer
dataset = load_dataset("trl-lib/tldr", split="train")
def reward_func(completions, **kwargs):
# Dummy reward function that rewards completions with more unique letters.
return [float(len(set(completion))) for completion in completions]
trainer = GRPOTrainer(
model="Qwen/Qwen2-0.5B-Instruct",
reward_funcs=reward_func,
train_dataset=dataset,
)
trainer.train()
```
Args:
model (`Union[str, PreTrainedModel]`):
Model to be trained. Can be either:
- A string, being the *model id* of a pretrained model hosted inside a model repo on huggingface.co, or
a path to a *directory* containing model weights saved using
[`~transformers.PreTrainedModel.save_pretrained`], e.g., `'./my_model_directory/'`. The model is
loaded using [`~transformers.AutoModelForCausalLM.from_pretrained`] with the keywork arguments
in `args.model_init_kwargs`.
- A [`~transformers.PreTrainedModel`] object. Only causal language models are supported.
reward_funcs (`Union[RewardFunc, list[RewardFunc]]`):
Reward functions to be used for computing the rewards. To compute the rewards, we call all the reward
functions with the prompts and completions and sum the rewards. Can be either:
- A single reward function, such as:
- A string: The *model ID* of a pretrained model hosted inside a model repo on huggingface.co, or a
path to a *directory* containing model weights saved using
[`~transformers.PreTrainedModel.save_pretrained`], e.g., `'./my_model_directory/'`. The model is loaded
using [`~transformers.AutoModelForSequenceClassification.from_pretrained`] with `num_labels=1` and the
keyword arguments in `args.model_init_kwargs`.
- A [`~transformers.PreTrainedModel`] object: Only sequence classification models are supported.
- A custom reward function: The function is provided with the prompts and the generated completions,
plus any additional columns in the dataset. It should return a list of rewards. For more details, see
[Using a custom reward function](#using-a-custom-reward-function).
- A list of reward functions, where each item can independently be any of the above types. Mixing different
types within the list (e.g., a string model ID and a custom reward function) is allowed.
args ([`GRPOConfig`], *optional*, defaults to `None`):
Configuration for this trainer. If `None`, a default configuration is used.
train_dataset ([`~datasets.Dataset`] or [`~datasets.IterableDataset`]):
Dataset to use for training. It must include a column `"prompt"`. Any additional columns in the dataset is
ignored. The format of the samples can be either:
- [Standard](dataset_formats#standard): Each sample contains plain text.
- [Conversational](dataset_formats#conversational): Each sample contains structured messages (e.g., role
and content).
eval_dataset ([`~datasets.Dataset`], [`~datasets.IterableDataset`] or `dict[str, Union[Dataset, IterableDataset]]`):
Dataset to use for evaluation. It must meet the same requirements as `train_dataset`.
processing_class ([`~transformers.PreTrainedTokenizerBase`], *optional*, defaults to `None`):
Processing class used to process the data. The padding side must be set to "left". If `None`, the
processing class is loaded from the model's name with [`~transformers.AutoTokenizer.from_pretrained`].
reward_processing_classes (`Union[PreTrainedTokenizerBase, list[PreTrainedTokenizerBase]]`, *optional*, defaults to `None`):
Processing classes corresponding to the reward functions specified in `reward_funcs`. Can be either:
- A single processing class: Used when `reward_funcs` contains only one reward function.
- A list of processing classes: Must match the order and length of the reward functions in `reward_funcs`.
If set to `None`, or if an element of the list corresponding to a [`~transformers.PreTrainedModel`] is
`None`, the tokenizer for the model is automatically loaded using [`~transformers.AutoTokenizer.from_pretrained`].
For elements in `reward_funcs` that are custom reward functions (not [`~transformers.PreTrainedModel`]),
the corresponding entries in `reward_processing_classes` are ignored.
callbacks (list of [`~transformers.TrainerCallback`], *optional*, defaults to `None`):
List of callbacks to customize the training loop. Will add those to the list of default callbacks
detailed in [here](https://huggingface.co/docs/transformers/main_classes/callback).
If you want to remove one of the default callbacks used, use the [`~transformers.Trainer.remove_callback`]
method.
optimizers (`tuple[torch.optim.Optimizer, torch.optim.lr_scheduler.LambdaLR]`, *optional*, defaults to `(None, None)`):
A tuple containing the optimizer and the scheduler to use. Will default to an instance of [`AdamW`] on your
model and a scheduler given by [`get_linear_schedule_with_warmup`] controlled by `args`.
peft_config ([`~peft.PeftConfig`], *optional*, defaults to `None`):
PEFT configuration used to wrap the model. If `None`, the model is not wrapped.
"""
def __init__(
self,
model,
reward_funcs,
args=None,
train_dataset=None,
eval_dataset=None,
processing_class=None,
reward_processing_classes=None,
callbacks=None,
peft_config=None,
**kwargs,
):
if args is None:
args = UnslothGRPOConfig()
use_bf16 = getattr(args, "bf16", False)
use_fp16 = getattr(args, "fp16", False)
dtype = getattr(model.config, "torch_dtype", None)
if dtype is None:
dtype = model.get_input_embeddings().dtype
from unsloth_zoo.utils import _get_dtype
dtype = _get_dtype(dtype)
float16 = dtype == torch.float16
if float16 and use_bf16:
raise TypeError(
"Unsloth: Model is in float16 precision but you want to use bfloat16 precision. Set fp16 to `True` and bf16 to `False`"
)
if not float16 and use_fp16:
raise TypeError(
"Unsloth: Model is in bfloat16 precision but you want to use float16 precision. Set fp16 to `False` and bf16 to `True`"
)
if not use_bf16 and not use_fp16:
args.fp16 = float16
args.bf16 = not float16
os.environ["ACCELERATE_MIXED_PRECISION"] = "fp16" if float16 else "bf16"
if (
getattr(args, "eval_dataset", None) is not None
and getattr(args, "eval_strategy", "no") == "no"
):
args.eval_strategy = "steps"
if getattr(args, "eval_steps", None) is None:
args.eval_steps = 0.1
ga_steps = getattr(args, "gradient_accumulation_steps", None)
if ga_steps is not None and ga_steps > 1:
from transformers import __version__ as transformers_version
if Version(transformers_version) <= Version("4.45.2"):
print(
"**** Unsloth: Please use our fixed gradient_accumulation_steps by updating transformers, TRL and Unsloth!\n"
"`pip install --upgrade --no-cache-dir --force-reinstall --no-deps unsloth transformers trl unsloth_zoo`"
)
if getattr(args, "eval_strategy", "no") != "no":
eval_bsz = getattr(args, "per_device_eval_batch_size", 8)
if eval_bsz == 8 and args.per_device_train_batch_size < eval_bsz:
args.per_device_eval_batch_size = args.per_device_train_batch_size
if (
getattr(args, "eval_accumulation_steps", None) is None
and ga_steps is not None
):
args.eval_accumulation_steps = ga_steps
fp16_full_eval = getattr(args, "fp16_full_eval", False)
bf16_full_eval = getattr(args, "bf16_full_eval", False)
if args.fp16 and bf16_full_eval:
args.bf16_full_eval = False
args.fp16_full_eval = True
if args.bf16 and fp16_full_eval:
args.bf16_full_eval = True
args.fp16_full_eval = False
if not bf16_full_eval and not fp16_full_eval:
args.bf16_full_eval = args.bf16
args.fp16_full_eval = args.fp16
if "max_seq_length" not in locals() and not hasattr(args, "max_seq_length"):
pass
else:
model_max_seq_length = getattr(model, "max_seq_length", None)
args_max_seq_length = getattr(args, "max_seq_length", None)
if args_max_seq_length is None and model_max_seq_length is not None:
max_seq_length = model.max_seq_length
if hasattr(args, "max_seq_length"):
args.max_seq_length = max_seq_length
if model is not None and hasattr(model, "for_training"):
model.for_training()
if "tokenizer" in locals() and hasattr(tokenizer, "padding_side"):
tokenizer.padding_side = "right"
if "processing_class" in locals():
if hasattr(processing_class, "padding_side"):
processing_class.padding_side = "right"
if hasattr(processing_class, "tokenizer") and hasattr(
processing_class.tokenizer, "padding_side"
):
processing_class.tokenizer.padding_side = "right"
other_metrics = []
if not isinstance(reward_funcs, list):
_reward_funcs = [reward_funcs]
else:
_reward_funcs = reward_funcs
for reward_func in _reward_funcs:
try:
reward_func_name = reward_func.__name__
other_metrics.append(f"rewards/{reward_func_name}")
except:
pass
from unsloth_zoo.logging_utils import PatchRLStatistics
PatchRLStatistics("grpo_trainer", other_metrics)
super().__init__(
model=model,
reward_funcs=reward_funcs,
args=args,
train_dataset=train_dataset,
eval_dataset=eval_dataset,
processing_class=processing_class,
reward_processing_classes=reward_processing_classes,
callbacks=callbacks,
peft_config=peft_config,
**kwargs,
)
if hasattr(self, "neftune_hook_handle"):
self.neftune_hook_handle.remove()
if hasattr(self, "neftune_hook_handle"):
del self.neftune_hook_handle
if getattr(args, "neftune_noise_alpha", None) is not None:
model.get_input_embeddings().neftune_noise_alpha = self.neftune_noise_alpha
pass
pass