deepseek-coder-6.7b-instruct_En__translations_size_104_epochs_10_2024-06-22_06-06-55_3557999

This model is a fine-tuned version of deepseek-ai/deepseek-coder-6.7b-instruct on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 3.1144
• Accuracy: 0.047
• Chrf: 0.5
• Bleu: 0.39
• Sacrebleu: 0.4
• Rouge1: 0.501
• Rouge2: 0.239
• Rougel: 0.448
• Rougelsum: 0.497
• Meteor: 0.413

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.001
• train_batch_size: 1
• eval_batch_size: 1
• seed: 3407
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 4
• total_eval_batch_size: 4
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-06
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 104
• training_steps: 1040

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Chrf	Bleu	Sacrebleu	Rouge1	Rouge2	Rougel	Rougelsum	Meteor
0.1338	4.0	104	1.1433	0.033	0.731	0.594	0.6	0.675	0.459	0.617	0.668	0.515
0.1034	8.0	208	1.2599	0.03	0.723	0.594	0.6	0.687	0.462	0.611	0.674	0.495
0.0974	12.0	312	1.3238	0.037	0.739	0.63	0.6	0.719	0.505	0.638	0.7	0.545
0.6468	16.0	416	1.9216	0.042	0.664	0.547	0.5	0.652	0.409	0.567	0.64	0.458
0.1566	20.0	520	2.1866	0.045	0.656	0.538	0.5	0.657	0.428	0.581	0.649	0.494
0.2056	24.0	624	2.5536	0.042	0.585	0.468	0.5	0.586	0.334	0.527	0.58	0.441
0.9677	28.0	728	2.8086	0.051	0.561	0.455	0.5	0.565	0.314	0.509	0.559	0.417
0.2163	32.0	832	2.9769	0.044	0.532	0.42	0.4	0.523	0.26	0.475	0.517	0.412
0.3617	36.0	936	3.1040	0.052	0.502	0.392	0.4	0.496	0.239	0.447	0.492	0.427
0.2153	40.0	1040	3.1144	0.047	0.5	0.39	0.4	0.501	0.239	0.448	0.497	0.413

Framework versions

• Transformers 4.37.0
• Pytorch 2.2.1+cu121
• Datasets 2.20.0
• Tokenizers 0.15.2