Model reference · open weights

xls-r-npsc-seq2seq

Name: xls-r-npsc-seq2seq
Brand: AxForge

Available as managed deployment Audio Rolv-Arild · community Speech→text 1 variants 690 dl/mo

xls-r-npsc-seq2seq is an open-weight audio or speech model from Rolv-Arild. AxForge deploys and operates it for you on dedicated EU-owned hardware — with the licence handled where one is required.

Request deployment All audio & music

Available as managed deployment — configured and operated for you on dedicated EU hardware, quoted per deployment.

What it is

Released by	Rolv-Arild
Type	Audio & music
Task	Speech→text
Runs with	transformers
Released	2022-03-02
Popularity	690 downloads / month
Licence	Unknown

About

What xls-r-npsc-seq2seq is

It achieves the following results on the evaluation set:

Loss: 0.2965
Wer: 0.3144

Model description

More information needed

Intended uses & limitations

More information needed

Read the full model card

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0001
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 8
total_train_batch_size: 64
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 1000
num_epochs: 20.0
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
2.888	0.51	400	3.7320	0.9440
3.1636	1.02	800	2.9188	1.1916
2.773	1.53	1200	2.3347	1.0134
0.7198	2.04	1600	0.6678	0.4826
0.5255	2.55	2000	0.4605	0.4135
0.3961	3.06	2400	0.4266	0.3955
0.3424	3.57	2800	0.3786	0.3741
0.3858	4.08	3200	0.3161	0.3552
0.3218	4.59	3600	0.3029	0.3510
0.199	5.1	4000	0.2988	0.3418
0.2054	5.61	4400	0.2873	0.3434
0.1704	6.12	4800	0.3129	0.3432
0.1805	6.63	5200	0.2963	0.3413
0.2091	7.14	5600	0.2755	0.3329
0.1971	7.65	6000	0.2706	0.3309
0.1237	8.16	6400	0.2823	0.3270
0.123	8.67	6800	0.2754	0.3246
0.103	9.18	7200	0.2917	0.3272
0.1143	9.69	7600	0.2885	0.3305
0.156	10.2	8000	0.2810	0.3288
0.167	10.71	8400	0.2689	0.3232
0.0815	11.22	8800	0.2899	0.3236
0.0844	11.73	9200	0.2798	0.3225
0.0775	12.24	9600	0.2894	0.3224
0.0677	12.75	10000	0.2838	0.3204
0.1383	13.27	10400	0.2959	0.3211
0.1233	13.77	10800	0.2922	0.3213
0.0688	14.29	11200	0.2903	0.3209
0.0655	14.8	11600	0.2868	0.3182
0.0449	15.31	12000	0.2959	0.3172
0.0421	15.82	12400	0.2966	0.3180
0.0858	16.33	12800	0.2941	0.3164
0.0859	16.84	13200	0.2980	0.3165
0.0561	17.35	13600	0.2965	0.3165
0.0506	17.86	14000	0.2935	0.3148
0.0312	18.37	14400	0.2964	0.3154
0.0403	18.88	14800	0.2967	0.3160
0.0924	19.39	15200	0.2955	0.3147
0.0585	19.9	15600	0.2965	0.3144

Framework versions

Transformers 4.17.0.dev0
Pytorch 1.10.0+cu113
Datasets 1.18.1
Tokenizers 0.11.0

From the published model card. Full card on the HuggingFace links in the sidebar.

Using it via the API

Call it like any OpenAI endpoint

Once AxForge deploys xls-r-npsc-seq2seq for you, it answers on the OpenAI-compatible API — the same base URL and keys as every other model. (xls-r-npsc-seq2seq below is illustrative; you get the exact model name on deployment.)

$ curl -sS https://api.axforge.ai/v1/audio/transcriptions \
  -H "Authorization: Bearer $AXFORGE_API_KEY" \
  -F model="xls-r-npsc-seq2seq" -F file=@audio.mp3

Create an account — your API key is available in the console. 3M free tokens every 30 days with every new account.

Get API access

Sizes, precisions & builds

xls-r-300m-npsc-seq2seq

— · BF16

Weights ↗

Details