Model reference · open weights

clip-roberta-finetuned

Name: clip-roberta-finetuned
Brand: AxForge

Available as managed deployment Embeddings davanstrien · community Embeddings 1 variants 580 dl/mo

clip-roberta-finetuned is an open-weight embedding model from davanstrien. AxForge deploys and operates it for you on dedicated EU-owned hardware — with the licence handled where one is required.

Request deployment All embedding models

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

What it is

Released by	davanstrien
Type	Embedding models
Task	Embeddings
Context	514 tokens
Runs with	transformers
Released	2022-07-13
Popularity	580 downloads / month
Licence	Unknown

About

What clip-roberta-finetuned is

This model is a fine-tuned version of ./clip-roberta on the davanstrien/manuscript_noisy_labels_iiif dataset. It achieves the following results on the evaluation set:

Loss: 2.5792

Read the full model card

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: 5e-05
train_batch_size: 128
eval_batch_size: 256
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 10.0
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
2.9841	0.07	500	3.4112
2.72	0.15	1000	3.3430
2.6319	0.22	1500	3.2295
2.5781	0.29	2000	3.1645
2.5339	0.36	2500	3.1226
2.503	0.44	3000	3.0856
2.4581	0.51	3500	3.0639
2.4494	0.58	4000	3.0415
2.4275	0.65	4500	3.0245
2.3909	0.73	5000	2.9991
2.3902	0.8	5500	2.9931
2.3741	0.87	6000	2.9612
2.3536	0.95	6500	2.9509
2.3392	1.02	7000	2.9289
2.3083	1.09	7500	2.9214
2.3094	1.16	8000	2.9153
2.2864	1.24	8500	2.9034
2.2893	1.31	9000	2.8963
2.2697	1.38	9500	2.8847
2.2762	1.46	10000	2.8665
2.2667	1.53	10500	2.8536
2.2548	1.6	11000	2.8472
2.238	1.67	11500	2.8491
2.2423	1.75	12000	2.8257
2.2406	1.82	12500	2.8287
2.2248	1.89	13000	2.8193
2.223	1.96	13500	2.8101
2.1995	2.04	14000	2.8027
2.1834	2.11	14500	2.7880
2.1723	2.18	15000	2.7783
2.1651	2.26	15500	2.7739
2.1575	2.33	16000	2.7825
2.1598	2.4	16500	2.7660
2.1667	2.47	17000	2.7578
2.1565	2.55	17500	2.7580
2.1558	2.62	18000	2.7561
2.1642	2.69	18500	2.7512
2.1374	2.77	19000	2.7361
2.1402	2.84	19500	2.7385
2.1326	2.91	20000	2.7235
2.1272	2.98	20500	2.7183
2.0954	3.06	21000	2.7156
2.0842	3.13	21500	2.7065
2.0859	3.2	22000	2.7089
2.0856	3.27	22500	2.6962
2.0775	3.35	23000	2.6931
2.0821	3.42	23500	2.6933
2.0706	3.49	24000	2.7011
2.0689	3.57	24500	2.7009
2.0807	3.64	25000	2.6825
2.0639	3.71	25500	2.6744
2.0742	3.78	26000	2.6777
2.0789	3.86	26500	2.6689
2.0594	3.93	27000	2.6566
2.056	4.0	27500	2.6676
2.0223	4.08	28000	2.6711
2.0185	4.15	28500	2.6568
2.018	4.22	29000	2.6567
2.0036	4.29	29500	2.6545
2.0238	4.37	30000	2.6559
2.0091	4.44	30500	2.6450
2.0096	4.51	31000	2.6389
2.0083	4.58	31500	2.6401
2.0012	4.66	32000	2.6399
2.0166	4.73	32500	2.6289
1.9963	4.8	33000	2.6348
1.9943	4.88	33500	2.6240
2.0099	4.95	34000	2.6190
1.9895	5.02	34500	2.6308
1.9581	5.09	35000	2.6385
1.9502	5.17	35500	2.6237
1.9485	5.24	36000	2.6248
1.9643	5.31	36500	2.6279
1.9535	5.38	37000	2.6185
1.9575	5.46	37500	2.6146
1.9475	5.53	38000	2.6093
1.9434	5.6	38500	2.6090
1.954	5.68	39000	2.6027
1.9509	5.75	39500	2.6107
1.9454	5.82	40000	2.5980
1.9479	5.89	40500	2.6016
1.9539	5.97	41000	2.5971
1.9119	6.04	41500	2.6228
1.8974	6.11	42000	2.6169
1.9038	6.19	42500	2.6027
1.9008	6.26	43000	2.6027
1.9142	6.33	43500	2.6011
1.8783	6.4	44000	2.5960
1.8896	6.48	44500	2.6111
1.8975	6.55	45000	2.5889
1.9048	6.62	45500	2.6007
1.9049	6.69	46000	2.5972
1.8969	6.77	46500	2.6053
1.9105	6.84	47000	2.5893
1.8921	6.91	47500	2.5883
1.8918	6.99	48000	2.5792
1.8671	7.06	48500	2.6041
1.8551	7.13	49000	2.6070
1.8555	7.2	49500	2.6148
1

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 clip-roberta-finetuned for you, it answers on the OpenAI-compatible API — the same base URL and keys as every other model. (clip-roberta-finetuned below is illustrative; you get the exact model name on deployment.)

$ curl -sS https://api.axforge.ai/v1/embeddings \
  -H "Authorization: Bearer $AXFORGE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"clip-roberta-finetuned","input":"text to embed"}'

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

clip-roberta-finetuned

— · BF16

Weights ↗

Details

Trained / evaluated on

davanstrien/manuscript_noisy_labels_iiif