# Reproducing DistilRoBERTa
**URL:**
**Author:**  [@DavidHarrison](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2FDavidHarrison)%5C
**Post date:** [March 30, 2021, 7:41pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Freproducing-distilroberta%2F5217%2F1 "2021-03-30T19:41:54Z")
I’ve been trying to retrain DistilRoBERTa from the information given [here](https://huggingface.co/distilroberta-base) along with the example code/documentation [here](https://github.com/huggingface/transformers/tree/master/examples/research_projects/distillation).
I’m a bit unclear on the exact configuration used to train the DistilRoBERTa model. I have been assuming it uses the same configuration as the DistilBERT model with minor changes, though some things, such as the loss coefficients are still a bit ambiguous.
**Would it be possible to share the exact command/configuration to train DistilRoBERTa?**
I’ve been able to replicate DistilRoBERTa to similar evaluation MLM perplexity but there still seems to be a small but statistically significant difference, I can share the full config if it’s helpful.
Thank you!
---
**Author:**  [@DavidHarrison](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2FDavidHarrison)%5C
**Post date:** [March 30, 2021, 7:43pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Freproducing-distilroberta%2F5217%2F2 "2021-03-30T19:43:36Z")
@VictorSanh it was [mentioned](https://github.com/huggingface/transformers/issues/10974#issuecomment-810525398) that you might be the best person to ask about this.
---
**Author:**  [@VictorSanh](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2FVictorSanh)%5C
**Post date:** [March 31, 2021, 2:06am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Freproducing-distilroberta%2F5217%2F3 "2021-03-31T02:06:07Z")
Hi @DavidHarrison ,
Here’s the exact configuration I used for DistilRoBERTa:
```
{
"force": true,
"dump_path": "serialization_dir/distilroberta-base",
"data_file": "data/dump_openwebtext.roberta-base.pickle",
"student_type": "roberta",
"student_config": "training_configs/roberta-base.json",
"student_pretrained_weights": "serialization_dir/tf_roberta-base_0247911.pth",
"teacher_type": "roberta",
"teacher_name": "roberta-base",
"temperature": 2.0,
"alpha_ce": 5.0,
"alpha_mlm": 2.0,
"alpha_clm": 0.0,
"alpha_mse": 0.0,
"alpha_cos": 1.0,
"mlm": true,
"mlm_mask_prop": 0.15,
"word_mask": 0.8,
"word_keep": 0.1,
"word_rand": 0.1,
"mlm_smoothing": 0.7,
"token_counts": "data/token_counts_openwebtext.roberta-base.pickle",
"restrict_ce_to_mask": false,
"freeze_pos_embs": true,
"freeze_token_type_embds": true,
"n_epoch": 4,
"batch_size": 5,
"group_by_size": true,
"gradient_accumulation_steps": 100,
"warmup_prop": 0.05,
"weight_decay": 0.0,
"learning_rate": 0.00025,
"adam_epsilon": 1e-06,
"max_grad_norm": 5.0,
"initializer_range": 0.02,
"fp16": false,
"fp16_opt_level": "O1",
"n_gpu": 8,
"local_rank": 0,
"seed": 56,
"log_interval": 400,
"checkpoint_interval": 20000,
"world_size": 8,
"n_gpu_per_node": 8,
"global_rank": 0,
"n_nodes": 1,
"node_id": 0,
"multi_gpu": true,
"is_master": true,
"multi_node": false
}
```
Victor
---
**Author:**  [@DavidHarrison](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2FDavidHarrison)%5C
**Post date:** [March 31, 2021, 6:23pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Freproducing-distilroberta%2F5217%2F4 "2021-03-31T18:23:26Z")
Thank you @VictorSanh!
---
**Author:**  [@theyorubayesian](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Ftheyorubayesian)%5C
**Post date:** [July 1, 2021, 3:56pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Freproducing-distilroberta%2F5217%2F5 "2021-07-01T15:56:41Z")
Hi @VictorSanh ,
For DistillGPT2, I know both the CLM loss and the distillation loss are used. Can you clarify whether the MSE loss is also used? Or the Cosine distance loss?
---
**Author:**  [@VictorSanh](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2FVictorSanh)%5C
**Post date:** [July 2, 2021, 8:15am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Freproducing-distilroberta%2F5217%2F6 "2021-07-02T08:15:14Z")
Hey @theyorubayesian !
Very much the same as distilroberta. we used the cosine loss, no mse loss!
Victor