Yili Hong comments

Results 9 comments of


                                            Yili Hong

Inconsistent number of instructions for sciworld_test.json on HF dataset

Same question. The [sciworld_test.json](https://huggingface.co/datasets/AgentGym/AgentEval/blob/main/sciworld_test.json) is even in the format of training set. Could you please update it to the correct version?

Quickstart PPO training error

I have the same error. Have you solved the issue?

Could not find the transformer layer class to wrap in the model.

@juliaparedesq I tried to remove `--fsdp_transformer_layer_cls_to_wrap 'LlamaDecoderLayer'` and the exception was gone. But after finetuning, the model's ability declined significantly. It seems that fastchat can only be used to deploy...

New binaries release needed for PyTorch 2.7.0 (torch2.7.0cu128 / torch2.6.0cu126 + flash_attn-2.7.4.post1 seem broken because PyTorch changed ABI)

So does any one have a solution for cuda12.8 + torch 2.7 with pip?

Having issues with vLLM for GRPO

Same issue

Having issues with vLLM for GRPO

```bash ValueError: vllm version 0.6.3.post1 not supported. Currently supported versions are 0.3.1, 0.4.2, 0.5.4, 0.6.3 and 0.7.0+ ```

OOM when using Reinforce++ with rule-based reward on1x8 80G H800 GPU

How to set rule-based rewards? I only find model-based reward examples.

OOM when using Reinforce++ with rule-based reward on1x8 80G H800 GPU

```python def reward_func(queries, prompts, labels): # queries is prompts + responses # labels is answers print(queries) return torch.randn(len(queries)) ``` @dubanx Could you give me an example of `prompts`,`queries` and `labels`?...

ImportError: /lib/x86_64-linux-gnu/libc.so.6: version `GLIBC_2.32' not found

Same issue, has any one found a solution?