Project5_V3_Mistral8x7b_V2.2.5

This model is a fine-tuned version of mistralai/Mixtral-8x7B-v0.1 on the generator dataset. It achieves the following results on the evaluation set:

• Loss: 2.2430

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: 2e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 128
• total_train_batch_size: 256
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• num_epochs: 40

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.63	1	2.3024
No log	1.88	3	2.2997
No log	2.51	4	2.2977
2.2951	3.76	6	2.2935
2.2951	4.39	7	2.2905
2.2951	5.65	9	2.2853
2.2845	6.9	11	2.2789
2.2845	7.53	12	2.2757
2.2845	8.78	14	2.2696
2.2729	9.41	15	2.2667
2.2729	10.67	17	2.2619
2.2729	11.92	19	2.2576
2.2599	12.55	20	2.2555
2.2599	13.8	22	2.2519
2.2599	14.43	23	2.2504
2.2523	15.69	25	2.2479
2.2523	16.94	27	2.2462
2.2523	17.57	28	2.2454
2.2471	18.82	30	2.2442
2.2471	19.45	31	2.2437
2.2471	20.71	33	2.2432
2.2444	21.96	35	2.2427
2.2444	22.59	36	2.2428
2.2444	23.84	38	2.2429
2.2444	24.47	39	2.2429
2.2427	25.1	40	2.2430

Framework versions

• PEFT 0.8.2
• Transformers 4.38.1
• Pytorch 2.2.1+cu121
• Datasets 2.17.1
• Tokenizers 0.15.2