Datasets:

qgallouedec

/

trl-metrics

like 1

HuggingFaceDocBuilderDev

The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/trl/pr_2348). All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

muellerzr

Beautiful! 🔥

HuggingFaceDocBuilderDev

The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/trl/pr_2347). All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

qgallouedec

Can you point the "previous version" you are refering to?

qgallouedec

I think it has been like this from the initial implementation (see #2020)

Galaxy-Husky

> I think it has been like this from the initial implementation (see #2020) Sorry, I didn't say that right. I mean before v0.11.0, there was no `maybe_apply_chat_template` back then. For example, the dpo dataset was preprocessed like: https://github.com/huggingface/trl/blob/55cc4b1076144b74a6ce5d07557b7f664b1de8d9/examples/scripts/dpo.py#L156-L160 Since the code has been refactored , I'm not sure if there was generation prompt or not. If so, could you please point out where it was implemented?

qgallouedec

Yes the example code was wrong, you need to add a generation prompt at the end of the prompt.

Galaxy-Husky

> Yes the example code was wrong, you need to add a generation prompt at the end of the prompt. I see. Thanks a lot!

HuggingFaceDocBuilderDev

The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/trl/pr_2345). All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

qgallouedec

Why do you need the model to be un eval mode? Can we use the inference mode in forward instead?

kashif

@ qgallouedec using inference mode so there should be no unexpected behaviour

qgallouedec

very nice @ccs96307! looking into details

HuggingFaceDocBuilderDev

The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/trl/pr_2344). All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

qgallouedec

Thanks a lot @ccs96307 for your contribution!

HuggingFaceDocBuilderDev

The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/trl/pr_2343). All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

qgallouedec

It should be fixed by #2325. Could you confirm?

asparius

Saving issue is solved but training time duration has increased significantly, 1 million episodes taking 300+ hours on A100. Is this expected, is there any reference number to compare with?

qgallouedec

I can't reproduce: ``` # v0.12.1 (includes the fix); transformers 4.47 dev (blue) /fsx/qgallouedec/trl/examples/scripts/rloo/rloo_tldr.py --output_dir models/minimal/rloo_tldr --dataset_name trl-internal-testing/tldr-preference-sft-trl-style --dataset_test_split validation --num_ppo_epochs 2 --num_mini_batches 2 --learning_rate 3e-6 --per_device_train_batch_size 4 --gradient_accumulation_steps 16 --total_episodes 1000 --model_name_or_path EleutherAI/pythia-1b-deduped --sft_model_path cleanrl/EleutherAI_pythia-1b-deduped__sft__tldr --reward_model_path cleanrl/EleutherAI_pythia-1b-deduped__reward__tldr --local_rollout_forward_batch_size 16 --missing_eos_penalty 1.0 --stop_token eos --kl_coef 0.03 --save_strategy steps --save_steps 10000 --eval_strategy steps --eval_steps 1000 --report_to wandb ``` ``` # TRL v0.11 (doesn't include the fix); transformers v4.45 (red) /fsx/qgallouedec/trl/examples/scripts/rloo/rloo_tldr.py --output_dir models/minimal/rloo_tldr --num_ppo_epochs 2 --num_mini_batches 2 --learning_rate 3e-6 --per_device_train_batch_size 4 --gradient_accumulation_steps 16 --total_episodes 1000 --model_name_or_path EleutherAI/pythia-1b-deduped --sft_model_path cleanrl/EleutherAI_pythia-1b-deduped__sft__tldr --reward_model_path cleanrl/EleutherAI_pythia-1b-deduped__reward__tldr --local_rollout_forward_batch_size 16 --missing_eos_penalty 1.0 --stop_token eos --kl_coef 0.03 --save_strategy steps --save_steps 10000 --eval_strategy steps --eval_steps 1000 --report_to wandb ``` 

sahandrez

This issue still persists on `trl==0.12.1`. The network usage after 100k starts to spike and checkpoints are saved every 2 steps, regardless of the value of for saving steps. 

Shreyas-Bhat

Hi @shashankg7 , I have the exact same question. Do you have the answer to this? Thanks

shashankg7

Kind of. To train in a mini-batch and multi-epoch mode with the samples collected from the current policy, plain REINFORCE/policy-gradient will not work, since the model changes from the policy used to collect the data. Importance sampling trick is required to account for the change in action distribution. But that's just my guess, there might be some other reason as well.

Shreyas-Bhat

Thanks a lot for your prompt response, @shashankg7 ! That makes more sense now. I had another question and was wondering if you face the same: during training, do your model logits tend to high negative values (often -inf)?

shashankg7

Hey @Shreyas-Bhat, missed your post. I am trying out RLOO in a different context, so I didn't try with the current setup, sorry. Did you manage to control/resolve the high negative error?

qgallouedec

We know that a lot of notebooks/docs are outdated. Sorry for the inconvenience. It was a deliberate choice that has allowed us to move faster on the lib evolution. For more information, see https://github.com/huggingface/trl/pull/2174#issuecomment-2399843454. But you can be sure that it will soon be completely up to date. Most doc and notebooks should work with `trl==0.11` I agree with you that the notebooks should mention it. Feel free to open a PR it that sense if you wan't to contribute

Debolena7

Thank you so much for your prompt reply. Changing the package trl version resolved the errors. I have been trying several code examples of rlhf from huggingface and also from youtube for a week now, and all had multiple issues. Was stuck for so many days. Thanks again..

Mrinh212375

@Debolena7 @qgallouedec ... ```` config = PPOConfig( #model_name="google/gemma-2-2b-it", learning_rate=1.41e-5, mini_batch_size=5, batch_size=20, output_dir='/kaggle/working/' ) ppo_trainer = PPOTrainer(config=config, processing_class = 'PreTrainedTokenizerBase' , policy = model, ref_policy = ref_model, reward_model = rm_model, #tokenizer=tokenizer, train_dataset=ppo_training_dataset, data_collator=collator) ```` when I'm trying to run the above code snippet, I'm getting the following error -  How to pass the module from the HF preTrainedWrapper class ?

ioana-ghiban-arm

hi! I'm facing quite a few errors when attempting running the 'toxicity' example as well. Currently stuck on this error: `TypeError: PPOTrainer.__init__() got multiple values for argument 'processing_class'`. Would immensely appreciate an updated end-to-end working demo of this. Thank you in advance.

Debolena7

> policy = model, > ref_policy = ref_model, > reward_model = rm_model, @Mrinh212375 I faced the same issue. So this error is basically caused because, the value model is not being passed in the 'PPOTrainer' arguments. So, by default, the value_model is None, which leads to the error. To solve it, you can either initialize a value model like: `value_model = AutoModelForSequenceClassification.from_pretrained("model_name")` and pass the value model into the 'PPOTrainer' , OR just simply use old `trl==0.11.0`

Debolena7

> hi! I'm facing quite a few errors when attempting running the 'toxicity' example as well. Currently stuck on this error: `TypeError: PPOTrainer.__init__() got multiple values for argument 'processing_class'`. Would immensely appreciate an updated end-to-end working demo of this. Thank you in advance. @ioana-ghiban-arm You can pass your model tokenizer into the 'processing_class' argument of PPOTrainer. `tokenizer = AutoTokenizer.from_pretrained(model_id)` ``` ppo_trainer = PPOTrainer(config=config, processing_class = tokenizer, .................) ```

ioana-ghiban-arm

@Debolena7 thank you for your help! you're right, I tried your suggestion and I think the execution got further. Now I'm getting the error I'd see when running a simplified version of the script. Do you perhaps have some troubleshooting steps for this error: `AttributeError: 'AutoModelForCausalLMWithValueHead' object has no attribute 'generation_config'`? TIA

Debolena7

it seems you have used something like: `model = AutoModelForCausalLMWithValueHead.from_pretrained(model_id)` which lead to the error.. you can use: `from transformers import GenerationConfig ` `model.generation_config = GenerationConfig()` , after initialization. But i would suggest it is best to use an old trl==0.11.0. otherwise, you will encounter more errors.

ioana-ghiban-arm

thank you for your help. Indeed, changing to `trl==0.11` does get the training going. However, I'm seeing this warning: `UserWarning: The average ratio of batch (...) exceeds threshold 10.00. Skipping batch.` which as mentioned [here](https://github.com/huggingface/trl/issues/1031) _suggests that the updates to the policy are too large, which could lead to instability in the training_. The maintainer suggested using [ppo.py](https://github.com/huggingface/trl/blob/main/examples/scripts/ppo/ppo.py) instead, so I tried adapting that script to use the toxicity model and dataset. However, as that is an updated script I'm assuming it should be ran with the latest version of trl provided by the repo. That leads me to the error that this thread started with.. Any suggestion to help me stop going in circles and be able to run a first round of fine-tuning on this model would be greatly appreciated, thank you.

Charley-xiao

I'm also getting the same `UserWarning: The average ratio of batch (13.00) exceeds threshold 10.00. Skipping batch.` kind of warning, using the exact toxicity example provided in the repo. Not sure if it would affect the results. I'd really appreciate it if someone could explain this🤔

imrankh46

@kashif any suggestions?

Sunrepe

### I encountered the same problem. My System Info is: ''' - Python version: 3.10.14 - PyTorch version: 2.4.1 - CUDA device(s): NVIDIA A800-SXM4-80GB, NVIDIA A800-SXM4-80GB, NVIDIA A800-SXM4-80GB, NVIDIA A800-SXM4-80GB, NVIDIA A100-SXM4-80GB, NVIDIA A100-SXM4-80GB, NVIDIA A100-SXM4-80GB - Transformers version: 4.46.2 - Accelerate version: 0.34.2 - Accelerate config: not found - Datasets version: 3.0.1 - HF Hub version: 0.25.1 - TRL version: 0.12.0 - bitsandbytes version: not installed - DeepSpeed version: 0.15.1 - Diffusers version: not installed - Liger-Kernel version: not installed - LLM-Blender version: not installed - OpenAI version: 0.28.0 - PEFT version: 0.13.0 ''' Here’s a revised version of your text with the grammar corrected: --- I am using the code in `example/script/sft.py`. I have downloaded the dataset and model locally. So, I run the following terminal command: ```bash python sft.py \ --model_name_or_path /data1/llm_models/qwen-05B \ --dataset_name /data1/datasets/trl-lib/Capybara \ --learning_rate 2.0e-4 \ --num_train_epochs 1 \ --packing \ --per_device_train_batch_size 2 \ --gradient_accumulation_steps 8 \ --gradient_checkpointing \ --logging_steps 25 \ --eval_strategy steps \ --eval_steps 100 \ --use_peft \ --lora_r 32 \ --lora_alpha 16 \ --output_dir Qwen2-0.5B-SFT ``` ## However, I am encountering the following issue: ```python Traceback (most recent call last): File "/data1/tmpzxf/research/SwiftSage/df_models/sft.py", line 106, in <module> trainer.train() File "/data1/envs/miniconda3/envs/tdt/lib/python3.10/site-packages/transformers/trainer.py", line 2123, in train return inner_training_loop( File "/data1/envs/miniconda3/envs/tdt/lib/python3.10/site-packages/transformers/trainer.py", line 2481, in _inner_training_loop tr_loss_step = self.training_step(model, inputs, num_items_in_batch) File "/data1/envs/miniconda3/envs/tdt/lib/python3.10/site-packages/transformers/trainer.py", line 3579, in training_step loss = self.compute_loss(model, inputs, num_items_in_batch=num_items_in_batch) File "/data1/envs/miniconda3/envs/tdt/lib/python3.10/site-packages/transformers/trainer.py", line 3633, in compute_loss outputs = model(**inputs) File "/data1/envs/miniconda3/envs/tdt/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1553, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/data1/envs/miniconda3/envs/tdt/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1562, in _call_impl return forward_call(*args, **kwargs) File "/data1/envs/miniconda3/envs/tdt/lib/python3.10/site-packages/torch/nn/parallel/data_parallel.py", line 176, in forward inputs, module_kwargs = self.scatter(inputs, kwargs, self.device_ids) File "/data1/envs/miniconda3/envs/tdt/lib/python3.10/site-packages/torch/nn/parallel/data_parallel.py", line 198, in scatter return scatter_kwargs(inputs, kwargs, device_ids, dim=self.dim) File "/data1/envs/miniconda3/envs/tdt/lib/python3.10/site-packages/torch/nn/parallel/scatter_gather.py", line 78, in scatter_kwargs scattered_kwargs = scatter(kwargs, target_gpus, dim) if kwargs else [] File "/data1/envs/miniconda3/envs/tdt/lib/python3.10/site-packages/torch/nn/parallel/scatter_gather.py", line 64, in scatter res = scatter_map(inputs) File "/data1/envs/miniconda3/envs/tdt/lib/python3.10/site-packages/torch/nn/parallel/scatter_gather.py", line 55, in scatter_map return [type(obj)(i) for i in zip(*map(scatter_map, obj.items()))] File "/data1/envs/miniconda3/envs/tdt/lib/python3.10/site-packages/torch/nn/parallel/scatter_gather.py", line 51, in scatter_map return list(zip(*map(scatter_map, obj))) File "/data1/envs/miniconda3/envs/tdt/lib/python3.10/site-packages/torch/nn/parallel/scatter_gather.py", line 47, in scatter_map return Scatter.apply(target_gpus, None, dim, obj) File "/data1/envs/miniconda3/envs/tdt/lib/python3.10/site-packages/torch/autograd/function.py", line 574, in apply return super().apply(*args, **kwargs) # type: ignore[misc] File "/data1/envs/miniconda3/envs/tdt/lib/python3.10/site-packages/torch/nn/parallel/_functions.py", line 96, in forward outputs = comm.scatter(input, target_gpus, chunk_sizes, ctx.dim, streams) File "/data1/envs/miniconda3/envs/tdt/lib/python3.10/site-packages/torch/nn/parallel/comm.py", line 188, in scatter return tuple(torch._C._scatter(tensor, devices, chunk_sizes, dim, streams)) RuntimeError: chunk expects at least a 1-dimensional tensor ```

qGentry

Looks like "num_items_in_batch" is getting added to the batch dict at some point by trl/tokenizer/collator and it is a 0-dim constant that is getting scattered across data parallel replicas but it can't.

hua-777

Isolating my training to 1 GPU fixed this problem for me. ``` import os os.environ["CUDA_VISIBLE_DEVICES"] = "0"

Leo-T-Zang

try transformers 4.45.1?

oscar50513

I successfully tested Transformers 4.46.0！！！！

imrankh46

I have some Nan entry in the dataset. And also change the code a little bit so it working for me.

yxdr

I encountered the same problem when I used the following command to run my training script. ``` CUDA_VISIBLE_DEVICES=0,1 python train.py \ --seed=1 \ --model_path=$MODEL_PATH \ --processed_data_dir=$PROCESSED_DATA_DIR \ --output_dir=$OUTPUT_DIR \ --learning_rate=5e-6 \ --epochs=1 \ --save_freq=10 \ --eval_freq=10 \ --num_warmup_steps=30 ``` But when I switched to using Huggingface Accelerate to run it, the problem disappeared. ``` CUDA_VISIBLE_DEVICES=0,1 accelerate launch --num_processes 2 train.py \ --seed=1 \ --model_path=$MODEL_PATH \ --processed_data_dir=$PROCESSED_DATA_DIR \ --output_dir=$OUTPUT_DIR \ --learning_rate=5e-6 \ --epochs=1 \ --save_freq=10 \ --eval_freq=10 \ --num_warmup_steps=30 ``` Additionally, if you use only one GPU, there should be no problem either.

Suman-punshi

I tried all the solutions above, reverting to single GPU and using accelerate, but it is still not solving the problem for me

kashif

@Suman-punshi what is your TRL Env and versions?

Suman-punshi

@kashif my TRL version 0.12.0

hojin-koh

> Looks like "num_items_in_batch" is getting added to the batch dict at some point by trl/tokenizer/collator and it is a 0-dim constant that is getting scattered across data parallel replicas but it can't. Got the same problem in our training environment with 2 GPUs, with trl 0.12.1 and transformer 4.46.3. I was using SFTTrainer with DataCollatorForCompletionOnlyLM on llama3.1-8b base model. After some tracing it is indeed that `num_items_in_batch` (it's just a plain number) causing problems. Trying to split a scalar between two GPUs can't be good lol Stopping `Trainer.compute_loss()` in `trainer.py` from adding `num_items_in_batch` to `loss_kwargs` solved the issue, although I don't know if there are any bad side-effects in doing this...

yxdr

> > Looks like "num_items_in_batch" is getting added to the batch dict at some point by trl/tokenizer/collator and it is a 0-dim constant that is getting scattered across data parallel replicas but it can't. > > Got the same problem in our training environment with 2 GPUs, with trl 0.12.1 and transformer 4.46.3. I was using SFTTrainer with DataCollatorForCompletionOnlyLM on llama3.1-8b base model. After some tracing it is indeed that `num_items_in_batch` (it's just a plain number) causing problems. Trying to split a scalar between two GPUs can't be good lol > > Stopping `Trainer.compute_loss()` in `trainer.py` from adding `num_items_in_batch` to `loss_kwargs` solved the issue, although I don't know if there are any bad side-effects in doing this... It does have side-effects when you set gradient_accumulation>1, because `num_items_in_batch` is used for averaging the losses when gradient_accumulation>1.

ag8

Using `accelerate launch` and adding the `--ddp_find_unused_parameters False` flag fixed the issue for me!

arivero

Interestingly, `num_items_in_batch` is also the cause of other problem, `loss = loss / num_items_in_batch` fails telling that tensors should to be on the same device. Problems seem to disappear if ds_accelerator is installed too

arslion

Nothing from above fixed my problem in Kaggle notebook. I want to use both GPU. Single GPU does fix the problem. But I need multi-gpu support. transformers version 4.46.3

ahuizxc

> Nothing from above fixed my problem in Kaggle notebook. I want to use both GPU. Single GPU does fix the problem. But I need multi-gpu support. transformers version 4.46.3 try downgrade transformers to 4.45.1, it works for me :)

qgallouedec

I agree. Not sure what's the best way to do that though, because it still has to work with the precomputing of ref logprobs. (that's why we initially set `"shuffle": False`). Any idea?

sagie-dekel

Hi Does anyone know how to solve it? how to set "shuffle": True in the trainer Dataloader

HuggingFaceDocBuilderDev

The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/trl/pr_2336). All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

littleshutong

trl/trainer/ppo_trainer.py  However, it is necessary to consider passing the parameters over.

ccs96307

I encountered this issue previously and temporarily worked around it by adjusting the accelerate version to 0.34.2. Here are the versions I used: - accelerate==0.34.2 - torch==2.5.1 - transformers==4.46.2 - deepspeed==0.15.4

Galaxy-Husky

@qgallouedec hi, do you have any suggestions?

qgallouedec

As far as I understand, the grad accum thing is only an issue with SFT right?

kashif

right i think its more about the updated kernels

HuggingFaceDocBuilderDev

The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/trl/pr_2333). All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

ByronHsu

Yes grad accum is only used for sft. Beside grad accum, we also have other improvement

qgallouedec

I approve, as this is an important issue affecting the most widely used trainer. (Thanks for solving it!) For the record, generally speaking, I won’t raise the minimum version requirement unless a new feature from the dependency is needed in our codebase.

kashif

thanks @yanghh2000 would it be possible to add a test?

yanghh2000

Hi, I am glad to help, but I am not sure how to add a test for this. Is there any guideline to test a PR?

yanghh2000

Oh, I have read the guideline in trl/CONTRIBUTING.md, and what I need to do is add a test.py and commit it under test/ dir?

kashif

yes in the `dpo_trainer` tests file

qgallouedec

Tbh I'm not sure it is possible to test it considering it's in a middle of the method.

qgallouedec

Good catch! Thanks! Do you mind opening a PR to fix that?

naskimed

Hey, I have the same issue using PPOTrainer: "ValueError: Please make sure to properly initialize your accelerator via `accelerator = Accelerator()` before using any functionality from the `accelerate` library". trl: 0.13.0.dev0 transformers: 4.46.2 accelerate: 1.1.0.dev0 

KAKSIS

> Hey, I have the same issue using PPOTrainer: "ValueError: Please make sure to properly initialize your accelerator via `accelerator = Accelerator()` before using any functionality from the `accelerate` library". > > trl: 0.13.0.dev0 transformers: 4.46.2 accelerate: 1.1.0.dev0  I have the same problem

kongjiellx

+1 with PPOTrainer

macheng6

After using the version configuration below, the code can be run: trl==0.11.4 accelerate==0.33.0，

leobianco

Having the same problem with `RLOOTrainer`.

zwhe99

+1

yananchen1989

same error with `/home/yanan/trl/examples/scripts/ppo/ppo.py`. accelerate: 1.2.0.dev0 trl: 0.13.0 transformers: 4.46.1

yaswanthchittepu

Same error with the ppo example script provided at the huggingface trl repo trl/examples/scripts/ppo/ppo_tldr.py, when using deepspeed zero2 acceelrate: 1.1.1 trl: 0.13.0.dev0 transformers: 4.47.0 deepspeed: 0.15.4

HuggingFaceDocBuilderDev

The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/trl/pr_2328). All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

HuggingFaceDocBuilderDev

The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/trl/pr_2327). All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

qgallouedec

We've have an example script to train VLM with DPO [here](https://github.com/huggingface/trl/blob/main/examples/scripts/dpo_vlm.py). Have you tried to run it with MiniCPM-V? At present, we're not claiming that you can use it with any VLM, as the level of standardization of VLMs is lower than that of LLMs. But it's definitely worth giving this one a try.

DarioPTWR

Alright cool! Will try it out and provide an update, thanks for your response!

DarioPTWR

Hi, I've tried to run the script with MiniCPM-v, but came across this error: (base) PS C:\Users\userAdmin\RLHF_V_MiniCPMV> accelerate launch dpo_vlm_2.py The following values were not passed to `accelerate launch` and had defaults used instead: `--num_processes` was set to a value of `0` `--num_machines` was set to a value of `1` `--mixed_precision` was set to a value of `'no'` `--dynamo_backend` was set to a value of `'no'` To avoid this warning pass in values for each of the problematic parameters or run `accelerate config`. MiniCPMForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly overwritten. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). - If you are not the owner of the model architecture class, please contact the model code owner to update it. Loading checkpoint shards: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 6.69it/s] Traceback (most recent call last): File "C:\Users\userAdmin\RLHF_V_MiniCPMV\dpo_vlm_2.py", line 78, in <module> main() File "C:\Users\userAdmin\RLHF_V_MiniCPMV\dpo_vlm_2.py", line 66, in main trainer = DPOTrainer( ^^^^^^^^^^^ File "c:\Users\userAdmin\RLHF_V_MiniCPMV\.venv\Lib\site-packages\huggingface_hub\utils\_deprecation.py", line 101, in inner_f return f(*args, **kwargs) ^^^^^^^^^^^^^^^^^^ File "c:\Users\userAdmin\RLHF_V_MiniCPMV\.venv\Lib\site-packages\transformers\utils\deprecation.py", line 165, in wrapped_func return func(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^ File "c:\Users\userAdmin\RLHF_V_MiniCPMV\.venv\Lib\site-packages\trl\trainer\dpo_trainer.py", line 367, in __init__ model.enable_input_require_grads() File "c:\Users\userAdmin\RLHF_V_MiniCPMV\.venv\Lib\site-packages\transformers\modeling_utils.py", line 18 self File "c:\Users\userAdmin\RLHF_V_MiniCPMV\.venv\Lib\site-packages\transformers\modeling_utils.py", line 1873, in get_input_embeddings File "c:\Users\userAdmin\RLHF_V_MiniCPMV\.venv\Lib\site-packages\transformers\modeling_utils.py", line 1873, in get_input_embeddings raise NotImplementedError NotImplementedError File "c:\Users\userAdmin\RLHF_V_MiniCPMV\.venv\Lib\site-packages\transformers\modeling_utils.py", line 1873, in get_input_embeddings raise NotImplementedError et_input_embeddings raise NotImplementedError raise NotImplementedError NotImplementedError Traceback (most recent call last): File "<frozen runpy>", line 198, in _run_module_as_main File "<frozen runpy>", line 88, in _run_code File "C:\Users\userAdmin\RLHF_V_MiniCPMV\.venv\Scripts\accelerate.exe\__main__.py", line 7, in <module> File "c:\Users\userAdmin\RLHF_V_MiniCPMV\.venv\Lib\site-packages\accelerate\commands\accelerate_cli.py", line 48, in main args.func(args) File "c:\Users\userAdmin\RLHF_V_MiniCPMV\.venv\Lib\site-packages\accelerate\commands\launch.py", line 1168, in launch_command simple_launcher(args) File "c:\Users\userAdmin\RLHF_V_MiniCPMV\.venv\Lib\site-packages\accelerate\commands\launch.py", line 763, in simple_launcher raise subprocess.CalledProcessError(returncode=process.returncode, cmd=cmd) subprocess.CalledProcessError: Command '['c:\\Users\\userAdmin\\RLHF_V_MiniCPMV\\.venv\\Scripts\\python.exe', 'dpo_vlm_2.py']' returned non-zero exit status 1. Seems like it has something to do with the GenerationMixin, is there any way to solve this? Thanks.

HuggingFaceDocBuilderDev

The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/trl/pr_2325). All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

HuggingFaceDocBuilderDev

The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/trl/pr_2324). All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

qgallouedec

Thanks @fanconic! Do you have reference results to share?

qgallouedec

Thanks for contributing @fanconic 👊

HuggingFaceDocBuilderDev

The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/trl/pr_2323). All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

HuggingFaceDocBuilderDev

The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/trl/pr_2322). All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

qgallouedec

> Thanks, do you mind giving a little more detail in the description about why this is needed? Done, sorry about that

HuggingFaceDocBuilderDev

The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/trl/pr_2321). All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

HuggingFaceDocBuilderDev

The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/trl/pr_2320). All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

HuggingFaceDocBuilderDev

The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/trl/pr_2319). All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

HuggingFaceDocBuilderDev

The docs for this PR live [here](https://moon-ci-docs.huggingface.co/docs/trl/pr_2318). All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

qgallouedec

Closed by #2318, thanks for reporting @LuisVasquezBSC

chenyang399

i find the problem , its because someone update the ppotrainer and ppoconfig, but didn't update the notebook, we need to pip install trl==0.11.3 to restore a older version.

chenyang399

i hope the community to update the notebook

qgallouedec

Yes, indeed, this has been discussed here, at https://github.com/huggingface/trl/pull/2174#issuecomment-2399843454. Sorry for the inconvenience. We're doing our best to update all the documentation, but it's a lot of work and help from the community would be greatly appreciated.

ZNP8b

I've had the same problem with SFTTrainer. And i found trl docs: https://huggingface.co/docs/trl/index Here is PPO docs: https://huggingface.co/docs/trl/ppo_trainer#trl.PPOTrainer There is 2 classes PPOtrainer and PPOconfig. And i think updated version is expecting model_name inside PPOtrainer and not PPOconfig:  Same thing happened with SFTtrainer: old SFTtrainer: ```Python trainer = SFTTrainer( model = model, tokenizer = tokenizer, train_dataset = dataset, dataset_text_field = "text", # keyword error max_seq_length = max_seq_length, # keyword error data_collator = DataCollatorForSeq2Seq(tokenizer = tokenizer), # keyword error dataset_num_proc = 2, # keyword error packing = False, # keyword error args = TrainingArguments( per_device_train_batch_size = 2, gradient_accumulation_steps = 4, warmup_steps = 5, # num_train_epochs = 1, max_steps = 60, learning_rate = 2e-4, fp16 = not is_bfloat16_supported(), bf16 = is_bfloat16_supported(), logging_steps = 1, optim = "adamw_8bit", weight_decay = 0.01, lr_scheduler_type = "linear", seed = 3407, output_dir = "outputs", report_to = "none", ), ) ``` Everything with keyword error is going inside TrainingArguments (now its SFTConfig) ```Python trainer = SFTTrainer( model=model, tokenizer=tokenizer, train_dataset=dataset, # dataset_text_field="text", # max_seq_length=max_seq_length, # dataset_num_proc=2, # packing=False, args=SFTConfig( # SFTConfig instead TrainingArguments dataset_text_field="text", # Here now max_seq_length=max_seq_length, # Here now dataset_num_proc=2, # Here now packing=False, # Here now per_device_train_batch_size=2, gradient_accumulation_steps=4, warmup_steps=50, num_train_epochs=20, # max_steps=5, learning_rate=1e-4, fp16=not is_bfloat16_supported(), bf16=is_bfloat16_supported(), logging_steps=1, optim="adamw_8bit", weight_decay=0.01, lr_scheduler_type="linear", seed=3407, output_dir="outputs", report_to="none", ), ) ``` Try moving model_name outside PPOConfig inside PPOTrainer

qgallouedec

Not sure to get your point.`model_name` is an argument of `PPOTrainer.create_model_card` (see your screenshot). Not `PPOConfig` nor `PPOTrainer`

chenyang399

thanks guys

Debolena7

All the notebooks are giving a lot of errors. `The` PPOTrainer class also has a lot of other arguments that are required to be passed. for example, 'processing_class' instead of 'tokenizer', 'policy', 'ref_policy', 'reward_model', 'value_model', 'train_dataset' and NO 'optimizer'. I resolved all of these by mentioning the correct arguments. But now i am stuck at a new error: ``` Traceback (most recent call last): File "/u/student/2020/ai20resch11003/miniconda3/envs/rlhf_new/lib/python3.9/site-packages/transformers/tokenization_utils_base.py", line 782, in convert_to_tensors tensor = as_tensor(value) File "/u/student/2020/ai20resch11003/miniconda3/envs/rlhf_new/lib/python3.9/site-packages/transformers/tokenization_utils_base.py", line 738, in as_tensor return torch.tensor(value) ValueError: too many dimensions 'str' The above exception was the direct cause of the following exception: Traceback (most recent call last): File "/u/student/2020/ai20resch11003/RLHF_new/hf_example.py", line 235, in <module> for epoch, batch in tqdm(enumerate(ppo_trainer.dataloader)): File "/u/student/2020/ai20resch11003/miniconda3/envs/rlhf_new/lib/python3.9/site-packages/tqdm/std.py", line 1181, in __iter__ for obj in iterable: File "/u/student/2020/ai20resch11003/miniconda3/envs/rlhf_new/lib/python3.9/site-packages/accelerate/data_loader.py", line 552, in __iter__ current_batch = next(dataloader_iter) File "/u/student/2020/ai20resch11003/miniconda3/envs/rlhf_new/lib/python3.9/site-packages/torch/utils/data/dataloader.py", line 630, in __next__ data = self._next_data() File "/u/student/2020/ai20resch11003/miniconda3/envs/rlhf_new/lib/python3.9/site-packages/torch/utils/data/dataloader.py", line 673, in _next_data data = self._dataset_fetcher.fetch(index) # may raise StopIteration File "/u/student/2020/ai20resch11003/miniconda3/envs/rlhf_new/lib/python3.9/site-packages/torch/utils/data/_utils/fetch.py", line 55, in fetch return self.collate_fn(data) File "/u/student/2020/ai20resch11003/miniconda3/envs/rlhf_new/lib/python3.9/site-packages/transformers/data/data_collator.py", line 271, in __call__ batch = pad_without_fast_tokenizer_warning( File "/u/student/2020/ai20resch11003/miniconda3/envs/rlhf_new/lib/python3.9/site-packages/transformers/data/data_collator.py", line 66, in pad_without_fast_tokenizer_warning padded = tokenizer.pad(*pad_args, **pad_kwargs) File "/u/student/2020/ai20resch11003/miniconda3/envs/rlhf_new/lib/python3.9/site-packages/transformers/tokenization_utils_base.py", line 3548, in pad return BatchEncoding(batch_outputs, tensor_type=return_tensors) File "/u/student/2020/ai20resch11003/miniconda3/envs/rlhf_new/lib/python3.9/site-packages/transformers/tokenization_utils_base.py", line 240, in __init__ self.convert_to_tensors(tensor_type=tensor_type, prepend_batch_axis=prepend_batch_axis) File "/u/student/2020/ai20resch11003/miniconda3/envs/rlhf_new/lib/python3.9/site-packages/transformers/tokenization_utils_base.py", line 798, in convert_to_tensors raise ValueError( ValueError: Unable to create tensor, you should probably activate truncation and/or padding with 'padding=True' 'truncation=True' to have batched tensors with the same length. Perhaps your features (`filename` in this case) have excessive nesting (inputs type `list` where type `int` is expected). ``` I am trying this notebook: https://github.com/huggingface/trl/blob/main/examples/research_projects/toxicity/scripts/gpt-j-6b-toxicity.py if these notebooks are outdated, please mention the correct 'trl' and 'transformers' package versions that are supposed to be installed before using these. that would help a lot. please help :(

qgallouedec

Thanks for sharing this. We rely on `transformers.Trainer` to save checkpoints and push on the hub. I think this issue would be more relevant to [huggingface/transformers](https://github.com/huggingface/transformers).