[bug] TypeError: patch_gpt_oss.<locals>.Mxfp4GptOssExperts.forward() got an unexpected keyword argument 'router_indices'

[shelterwff-byte]

from unsloth import FastLanguageModel
import torch
max_seq_length = 64*1024 # Set long context sequence length here!
dtype = None
model, tokenizer = FastLanguageModel.from_pretrained(
    model_name = "./models/unsloth/gpt-oss-120b",
    dtype = dtype, # None for auto detection
    max_seq_length = max_seq_length, # Choose any for long context!
    load_in_4bit = True,  # 4 bit quantization to reduce memory
    full_finetuning = False, # [NEW!] We have full finetuning now!
    unsloth_tiled_mlp = True, # Super long context >500K finetuning!
    trust_remote_code=True,
)
model = FastLanguageModel.get_peft_model(
    model,
    r = 16, # Choose any number > 0 ! Suggested 8, 16, 32, 64, 128
    target_modules = ["q_proj", "k_proj", "v_proj", ], # "o_proj", "gate_proj", "up_proj", "down_proj",
    lora_alpha = 32,
    lora_dropout = 0, # Supports any, but = 0 is optimized
    bias = "none",    # Supports any, but = "none" is optimized
    # [NEW] "unsloth" uses 30% less VRAM, fits 2x larger batch sizes!
    use_gradient_checkpointing = "unsloth", # True or "unsloth" for very long context
    random_state = 3407,
)

🦥 Unsloth: Will patch your computer to enable 2x faster free finetuning.
🦥 Unsloth Zoo will now patch everything to make training faster!
Unsloth: WARNING `trust_remote_code` is True.
Are you certain you want to do remote code execution?
==((====))==  Unsloth 2025.12.5: Fast Gpt_Oss patching. Transformers: 4.57.3.
   \\   /|    NVIDIA H200. Num GPUs = 1. Max memory: 139.812 GB. Platform: Linux.
O^O/ \_/ \    Torch: 2.9.1+cu128. CUDA: 9.0. CUDA Toolkit: 12.8. Triton: 3.4.0
\        /    Bfloat16 = TRUE. FA [Xformers = 0.0.33.post2. FA2 = False]
 "-____-"     Free license: http://github.com/unslothai/unsloth
Unsloth: Fast downloading is enabled - ignore downloading bars which are red colored!
Unsloth: Gpt_Oss does not support SDPA - switching to fast eager.
Loading checkpoint shards: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 15/15 [00:16<00:00,  1.11s/it]
Unsloth: Making `model.base_model.model.model` require gradients
GPU = NVIDIA H200. Max memory = 139.812 GB.
71.789 GB of memory reserved.
The model is already on multiple devices. Skipping the move to device specified in `args`.
  0%|                                                                                                                                                                                 | 0/5380 [00:00<?, ?it/s]Unsloth: Will smartly offload gradients to save VRAM!

发生异常: TypeError       (note: full exception trace is shown but execution is paused at: _run_module_as_main)
patch_gpt_oss.<locals>.Mxfp4GptOssExperts.forward() got an unexpected keyword argument 'router_indices'
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/transformers/utils/generic.py", line 1079, in wrapper
    outputs = func(self, *args, **kwargs_without_recordable)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/transformers/models/gpt_oss/modeling_gpt_oss.py", line 507, in forward
    hidden_states,

                attention_mask=causal_mask_mapping[decoder_layer.attention_type],

                position_ids=position_ids,

                past_key_values=past_key_values,

                use_cache=use_cache,

                cache_position=cache_position,

                position_embeddings=position_embeddings,

                **kwargs,

            )

        hidden_states = self.norm(hidden_states)
                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/transformers/modeling_layers.py", line 93, in __call__
    return self._gradient_checkpointing_func(partial(super().__call__, **kwargs), *args)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/_compile.py", line 53, in inner
    return disable_fn(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/_dynamo/eval_frame.py", line 1044, in _fn
    return fn(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/utils/checkpoint.py", line 496, in checkpoint
    return CheckpointFunction.apply(function, preserve, *args)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/autograd/function.py", line 581, in apply
    return super().apply(*args, **kwargs)  # type: ignore[misc]
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/unsloth_zoo/gradient_checkpointing.py", line 492, in forward
    outputs = run_function(*args)
              ^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl
    return forward_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/transformers/utils/deprecation.py", line 172, in wrapped_func
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/transformers/models/gpt_oss/modeling_gpt_oss.py", line 386, in forward
    hidden_states, _ = self.mlp(hidden_states)  # diff with llama: router scores
                       ^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl
    return forward_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/unsloth_zoo/tiled_mlp.py", line 256, in tiled_forward_arctic_size
    return TiledMLP.apply(inner_forward, mlp_module, x, preserve_rng_state, n_shards, chunk_size)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/autograd/function.py", line 581, in apply
    return super().apply(*args, **kwargs)  # type: ignore[misc]
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/amp/autocast_mode.py", line 527, in decorate_fwd
    return fwd(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/unsloth_zoo/tiled_mlp.py", line 143, in forward
    out = TiledMLP.handle_output(mlp_forward(x_split), extra_outputs)
                                 ^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/transformers/models/gpt_oss/modeling_gpt_oss.py", line 170, in forward
    routed_out = self.experts(hidden_states, router_indices=router_indices, routing_weights=router_scores)
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl
    return forward_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
TypeError: patch_gpt_oss.<locals>.Mxfp4GptOssExperts.forward() got an unexpected keyword argument 'router_indices'

During handling of the above exception, another exception occurred:

  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl
    return forward_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/transformers/models/gpt_oss/modeling_gpt_oss.py", line 170, in forward
    routed_out = self.experts(hidden_states, router_indices=router_indices, routing_weights=router_scores)
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/unsloth_zoo/tiled_mlp.py", line 143, in forward
    out = TiledMLP.handle_output(mlp_forward(x_split), extra_outputs)
                                 ^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/amp/autocast_mode.py", line 527, in decorate_fwd
    return fwd(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/autograd/function.py", line 581, in apply
    return super().apply(*args, **kwargs)  # type: ignore[misc]
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/unsloth_zoo/tiled_mlp.py", line 256, in tiled_forward_arctic_size
    return TiledMLP.apply(inner_forward, mlp_module, x, preserve_rng_state, n_shards, chunk_size)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl
    return forward_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/transformers/models/gpt_oss/modeling_gpt_oss.py", line 386, in forward
    hidden_states, _ = self.mlp(hidden_states)  # diff with llama: router scores
                       ^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/transformers/utils/deprecation.py", line 172, in wrapped_func
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl
    return forward_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/unsloth_zoo/gradient_checkpointing.py", line 492, in forward
    outputs = run_function(*args)
              ^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/autograd/function.py", line 581, in apply
    return super().apply(*args, **kwargs)  # type: ignore[misc]
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/utils/checkpoint.py", line 496, in checkpoint
    return CheckpointFunction.apply(function, preserve, *args)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/_dynamo/eval_frame.py", line 1044, in _fn
    return fn(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/_compile.py", line 53, in inner
    return disable_fn(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/transformers/modeling_layers.py", line 93, in __call__
    return self._gradient_checkpointing_func(partial(super().__call__, **kwargs), *args)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/transformers/models/gpt_oss/modeling_gpt_oss.py", line 507, in forward
    hidden_states,

                attention_mask=causal_mask_mapping[decoder_layer.attention_type],

                position_ids=position_ids,

                past_key_values=past_key_values,

                use_cache=use_cache,

                cache_position=cache_position,

                position_embeddings=position_embeddings,

                **kwargs,

            )

        hidden_states = self.norm(hidden_states)
                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/transformers/utils/generic.py", line 1081, in wrapper
    raise original_exception
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/transformers/utils/generic.py", line 1081, in wrapper
    raise original_exception
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl
    return forward_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/transformers/models/gpt_oss/modeling_gpt_oss.py", line 668, in forward
    input_ids=input_ids,

            attention_mask=attention_mask,

            position_ids=position_ids,

            past_key_values=past_key_values,

            inputs_embeds=inputs_embeds,

            use_cache=use_cache,

            output_router_logits=output_router_logits,

            cache_position=cache_position,

            **kwargs,

        )
                                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/transformers/utils/generic.py", line 918, in wrapper
    output = func(self, *args, **kwargs)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/peft/tuners/tuners_utils.py", line 308, in forward
    return self.model.forward(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl
    return forward_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/peft/peft_model.py", line 1923, in forward
    input_ids=input_ids,

                    attention_mask=attention_mask,

                    inputs_embeds=inputs_embeds,

                    labels=labels,

                    output_attentions=output_attentions,

                    output_hidden_states=output_hidden_states,

                    return_dict=return_dict,

                    **kwargs,

                )
     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/amp/autocast_mode.py", line 44, in decorate_autocast
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/accelerate/utils/operations.py", line 807, in __call__
    return convert_to_fp32(self.model_forward(*args, **kwargs))
                           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/accelerate/utils/operations.py", line 819, in forward
    return model_forward(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1786, in _call_impl
    return forward_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1775, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "<string>", line 36, in compute_loss
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/unsloth/models/_utils.py", line 1651, in _unsloth_pre_compute_loss
    outputs = self._old_compute_loss(model, inputs, *args, **kwargs)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/shelterw/unsloth_compiled_cache/UnslothSFTTrainer.py", line 1071, in compute_loss
    model,

            inputs,

            return_outputs = return_outputs,

            num_items_in_batch = num_items_in_batch,

        )

        return outputs
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "<string>", line 40, in _unsloth_training_step
  File "/mnt/data1/xmw/shelterw/unsloth_compiled_cache/UnslothSFTTrainer.py", line 1082, in training_step
    return super().training_step(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/transformers/trainer.py", line 2674, in _inner_training_loop
    tr_loss_step = self.training_step(model, inputs, num_items_in_batch)
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/site-packages/transformers/trainer.py", line 2325, in train
    args=args,

                resume_from_checkpoint=resume_from_checkpoint,

                trial=trial,

                ignore_keys_for_eval=ignore_keys_for_eval,

            )
     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/shelterw/unsloth_compiled_cache/UnslothSFTTrainer.py", line 55, in wrapper
    output = f(self, *args, **kwargs)
             ^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/xmw/data/shelterw/train_oss.py", line 81, in <module>
    trainer_stats = trainer.train()
                    ^^^^^^^^^^^^^^^
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/runpy.py", line 88, in _run_code
    exec(code, run_globals)
  File "/mnt/data1/xmw/miniconda3/envs/unsloth/lib/python3.11/runpy.py", line 198, in _run_module_as_main (Current frame)
    "__main__", mod_spec)
TypeError: patch_gpt_oss.<locals>.Mxfp4GptOssExperts.forward() got an unexpected keyword argument 'router_indices'

[mmathew23]

@shelterwff-byte are you using unsloth/gpt-oss-120b instead of unsloth/gpt-oss-120b-BF16? I would try that first. But if that doesn't fix the issue, then maybe try reinstall unsloth in a fresh venv with just pip install unsloth. For reference this script should run as a proof of concept.

from unsloth import FastLanguageModel
import torch
max_seq_length = 64*1024 # Set long context sequence length here!
dtype = None
model, tokenizer = FastLanguageModel.from_pretrained(
    model_name = "unsloth/gpt-oss-120b",
    dtype = dtype, # None for auto detection
    max_seq_length = max_seq_length, # Choose any for long context!
    load_in_4bit = True,  # 4 bit quantization to reduce memory
    full_finetuning = False, # [NEW!] We have full finetuning now!
    unsloth_tiled_mlp = True, # Super long context >500K finetuning!
)
model = FastLanguageModel.get_peft_model(
    model,
    r = 16, # Choose any number > 0 ! Suggested 8, 16, 32, 64, 128
    target_modules = ["q_proj", "k_proj", "v_proj", ], # "o_proj", "gate_proj", "up_proj", "down_proj",
    lora_alpha = 32,
    lora_dropout = 0, # Supports any, but = 0 is optimized
    bias = "none",    # Supports any, but = "none" is optimized
    # [NEW] "unsloth" uses 30% less VRAM, fits 2x larger batch sizes!
    use_gradient_checkpointing = "unsloth", # True or "unsloth" for very long context
    random_state = 3407,
)

import requests
from dataclasses import dataclass, field
from typing import List
@dataclass
class Book:
    url: str
    text: str = ""
    input_ids: List[int] = field(default_factory = list)
    length: int = 0
books = {
    # "Frankenstein" : Book("https://www.gutenberg.org/ebooks/84.txt.utf-8"),
    "Moby Dick" : Book("https://www.gutenberg.org/ebooks/2701.txt.utf-8"),
    "Pride and Prejudice" : Book("https://www.gutenberg.org/ebooks/1342.txt.utf-8"),
    # "The King in Yellow" : Book("https://www.gutenberg.org/ebooks/8492.txt.utf-8"),
    # "Romeo and Juliet" : Book("https://www.gutenberg.org/ebooks/1513.txt.utf-8"),
}
for book_name, book_data in books.items():
    book_data.text = requests.get(book_data.url).text
    book_data.input_ids = tokenizer.encode(book_data.text)
    book_data.length = len(book_data.input_ids)
    print(f"The book '{book_name}' has context length = {book_data.length:,}")


from datasets import Dataset
import itertools
messages = []
for book_name, book_data in books.items():
    messages += [
        {"role" : "user", "content" : f"Print out the contents for the book: '{book_name}'"},
        {"role" : "assistant", "content" : book_data.text},
    ]
dataset = Dataset.from_list([{"text" : tokenizer.apply_chat_template(messages, tokenize = False) + tokenizer.eos_token}])

from trl import SFTConfig, SFTTrainer
trainer = SFTTrainer(
    model = model,
    tokenizer = tokenizer,
    train_dataset = dataset,
    args = SFTConfig(
        per_device_train_batch_size = 1,
        gradient_accumulation_steps = 1,
        warmup_steps = 1,
        # num_train_epochs = 1, # Set this for 1 full training run.
        max_steps = 1,
        learning_rate = 2e-4,
        logging_steps = 1,
        optim = "adamw_8bit",
        weight_decay = 0.001,
        lr_scheduler_type = "linear",
        seed = 3407,
        output_dir = "outputs",
        report_to = "none", # Use TrackIO/WandB etc
    ),
)

trainer_stats = trainer.train()