You can deploy this model and start earning money today!

cbpuschmann/
MiniLM-klimacoder_v0.6

Name: MiniLM-klimacoder_v0.6
Availability: InStock

setfit

safetensors

bert

sentence-transformers

text-classification

generated_from_setfit_trainer

2209.11055

sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2

finetune

model-index

Views

No views yet

SetFit with sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2

This is a SetFit model that can be used for Text Classification. This SetFit model uses sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2 as the Sentence Transformer embedding model. A LogisticRegression instance is used for classification.

The model has been trained using an efficient few-shot learning technique that involves:

Fine-tuning a Sentence Transformer with contrastive learning.
Training a classification head with features from the fine-tuned Sentence Transformer.

Model Details

Model Description

Model Type: SetFit
Sentence Transformer body: sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2
Classification head: a LogisticRegression instance
Maximum Sequence Length: 128 tokens
Number of Classes: 3 classes

Model Sources

Repository: SetFit on GitHub
Paper: Efficient Few-Shot Learning Without Prompts
Blogpost: SetFit: Efficient Few-Shot Learning Without Prompts

Model Labels

Label	Examples
neutral	'Die Aktionen von Klima-Aktivisten, die in mehreren Städten zu Verkehrsbehinderungen geführt haben, haben in der Öffentlichkeit sowohl Unterstützung als auch Kritik ausgelöst.' ' Die Diskussion über ein nationales Tempolimit auf Autobahnen spaltet weiterhin die Gemüter, während Experten die potenziellen Vorteile und Nachteile abwägen.' ' Der Bundestag wird in den kommenden Wochen über das geplante Heizungsgesetz debattieren.'
supportive	' "Die Aktionen von Gruppen wie Fridays for Future und der Letzten Generation zeigen, dass die junge Generation bereit ist, für eine lebenswerte Zukunft zu kämpfen."' ' Die Einführung eines nationalen Tempolimits auf Autobahnen könnte die Verkehrssicherheit erheblich verbessern und die Zahl der Verkehrstoten reduzieren.' '"Die jungen Aktivisten von Fridays for Future und die Letzte Generation haben mit ihren unkonventionellen Aktionen ein wichtiges Gespräch über die Dringlichkeit des Klimaschutzes angestoßen."'
opposed	'„Die Polizei musste am Freitag wiederholt mit harten Bandagen gegen die Klima-Rebellen vorgehen, die Straßen und Plätze in der Innenstadt blockierten, um für ihre Forderungen zu demonstrieren.“' ' "Ein Tempolimit auf deutschen Autobahnen würde den freiheitsliebenden Autofahrern das Herz brechen."' 'Die ständigen Straßenblockaden und Farbbeanspritzungen auf Kunstwerke haben viele Menschen in Deutschland mehr als nur gereizt - sie haben sie in ihrem täglichen Leben massiv behindert und zu einer wachsenden Ablehnung gegenüber den Klima-Aktivisten geführt.'

Evaluation

Metrics

Label	Accuracy
all	0.9570

Uses

Direct Use for Inference

First install the SetFit library:

pip install setfit

Then you can load this model and run inference.

python

from setfit import SetFitModel

# Download from the 🤗 Hub
model = SetFitModel.from_pretrained("cbpuschmann/MiniLM-klimacoder_v0.6")
# Run inference
preds = model(" \"Ein nationales Tempolimit auf Autobahnen wäre ein weiterer Schritt in Richtung eines überregulierten Staates, der den Bürgern ihre Freiheit stückweise entreißt.\"")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	10	25.7025	53

Label	Training Sample Count
neutral	318
opposed	388
supportive	410

Training Hyperparameters

batch_size: (32, 32)
num_epochs: (1, 1)
max_steps: -1
sampling_strategy: oversampling
body_learning_rate: (2e-05, 1e-05)
head_learning_rate: 0.01
loss: CosineSimilarityLoss
distance_metric: cosine_distance
margin: 0.25
end_to_end: False
use_amp: False
warmup_proportion: 0.1
l2_weight: 0.01
seed: 42
eval_max_steps: -1
load_best_model_at_end: False

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0000	1	0.2339	-
0.0019	50	0.2439	-
0.0039	100	0.2407	-
0.0058	150	0.2295	-
0.0078	200	0.2123	-
0.0097	250	0.1903	-
0.0116	300	0.153	-
0.0136	350	0.1322	-
0.0155	400	0.116	-
0.0174	450	0.0937	-
0.0194	500	0.0721	-
0.0213	550	0.0525	-
0.0233	600	0.0388	-
0.0252	650	0.0338	-
0.0271	700	0.026	-
0.0291	750	0.0224	-
0.0310	800	0.0122	-
0.0329	850	0.0088	-
0.0349	900	0.0079	-
0.0368	950	0.0055	-
0.0388	1000	0.004	-
0.0407	1050	0.0027	-
0.0426	1100	0.0025	-
0.0446	1150	0.0019	-
0.0465	1200	0.0014	-
0.0484	1250	0.0013	-
0.0504	1300	0.0006	-
0.0523	1350	0.0012	-
0.0543	1400	0.0006	-
0.0562	1450	0.0004	-
0.0581	1500	0.0003	-
0.0601	1550	0.0003	-
0.0620	1600	0.0003	-
0.0639	1650	0.0002	-
0.0659	1700	0.0007	-
0.0678	1750	0.0002	-
0.0698	1800	0.0002	-
0.0717	1850	0.0002	-
0.0736	1900	0.0003	-
0.0756	1950	0.0002	-
0.0775	2000	0.0001	-
0.0794	2050	0.0001	-
0.0814	2100	0.0001	-
0.0833	2150	0.0001	-
0.0853	2200	0.0008	-
0.0872	2250	0.0007	-
0.0891	2300	0.0007	-
0.0911	2350	0.0002	-
0.0930	2400	0.0001	-
0.0950	2450	0.0001	-
0.0969	2500	0.0014	-
0.0988	2550	0.0008	-
0.1008	2600	0.0009	-
0.1027	2650	0.0006	-
0.1046	2700	0.0008	-
0.1066	2750	0.0001	-
0.1085	2800	0.0	-
0.1105	2850	0.0	-
0.1124	2900	0.0	-
0.1143	2950	0.0	-
0.1163	3000	0.0	-
0.1182	3050	0.0	-
0.1201	3100	0.0	-
0.1221	3150	0.0	-
0.1240	3200	0.0	-
0.1260	3250	0.0	-
0.1279	3300	0.0	-
0.1298	3350	0.0	-
0.1318	3400	0.0	-
0.1337	3450	0.0	-
0.1356	3500	0.0	-
0.1376	3550	0.0	-
0.1395	3600	0.0	-
0.1415	3650	0.0	-
0.1434	3700	0.0	-
0.1453	3750	0.0	-
0.1473	3800	0.0	-
0.1492	3850	0.0	-
0.1511	3900	0.0	-
0.1531	3950	0.0	-
0.1550	4000	0.001	-
0.1570	4050	0.0012	-
0.1589	4100	0.0042	-
0.1608	4150	0.0023	-
0.1628	4200	0.001	-
0.1647	4250	0.001	-
0.1666	4300	0.0001	-
0.1686	4350	0.0	-
0.1705	4400	0.0	-
0.1725	4450	0.0	-
0.1744	4500	0.0	-
0.1763	4550	0.0003	-
0.1783	4600	0.0	-
0.1802	4650	0.0	-
0.1821	4700	0.0005	-
0.1841	4750	0.0009	-
0.1860	4800	0.0001	-
0.1880	4850	0.0	-
0.1899	4900	0.0	-
0.1918	4950	0.0	-
0.1938	5000	0.0	-
0.1957	5050	0.0	-
0.1977	5100	0.0	-
0.1996	5150	0.0	-
0.2015	5200	0.0	-
0.2035	5250	0.0	-
0.2054	5300	0.0	-
0.2073	5350	0.0	-
0.2093	5400	0.0	-
0.2112	5450	0.0	-
0.2132	5500	0.0	-
0.2151	5550	0.0	-
0.2170	5600	0.0	-
0.2190	5650	0.0	-
0.2209	5700	0.0	-
0.2228	5750	0.0	-
0.2248	5800	0.0	-
0.2267	5850	0.0	-
0.2287	5900	0.0	-
0.2306	5950	0.0	-
0.2325	6000	0.0	-
0.2345	6050	0.0	-
0.2364	6100	0.0	-
0.2383	6150	0.0	-
0.2403	6200	0.0	-
0.2422	6250	0.0	-
0.2442	6300	0.0	-
0.2461	6350	0.0	-
0.2480	6400	0.0	-
0.2500	6450	0.0	-
0.2519	6500	0.0	-
0.2538	6550	0.0	-
0.2558	6600	0.0	-
0.2577	6650	0.0	-
0.2597	6700	0.0	-
0.2616	6750	0.0	-
0.2635	6800	0.0	-
0.2655	6850	0.0	-
0.2674	6900	0.0	-
0.2693	6950	0.0	-
0.2713	7000	0.0	-
0.2732	7050	0.0	-
0.2752	7100	0.0	-
0.2771	7150	0.0	-
0.2790	7200	0.0	-
0.2810	7250	0.0	-
0.2829	7300	0.0	-
0.2849	7350	0.0	-
0.2868	7400	0.0	-
0.2887	7450	0.0	-
0.2907	7500	0.0	-
0.2926	7550	0.0	-
0.2945	7600	0.0	-
0.2965	7650	0.0	-
0.2984	7700	0.0	-
0.3004	7750	0.0	-
0.3023	7800	0.0	-
0.3042	7850	0.0	-
0.3062	7900	0.0	-
0.3081	7950	0.0	-
0.3100	8000	0.0	-
0.3120	8050	0.0	-
0.3139	8100	0.0	-
0.3159	8150	0.0	-
0.3178	8200	0.0	-
0.3197	8250	0.0	-
0.3217	8300	0.0	-
0.3236	8350	0.0	-
0.3255	8400	0.0	-
0.3275	8450	0.0	-
0.3294	8500	0.0	-
0.3314	8550	0.0	-
0.3333	8600	0.0	-
0.3352	8650	0.0	-
0.3372	8700	0.0	-
0.3391	8750	0.0	-
0.3410	8800	0.0	-
0.3430	8850	0.0	-
0.3449	8900	0.0	-
0.3469	8950	0.0	-
0.3488	9000	0.0	-
0.3507	9050	0.0	-
0.3527	9100	0.0	-
0.3546	9150	0.0	-
0.3565	9200	0.0042	-
0.3585	9250	0.0083	-
0.3604	9300	0.0071	-
0.3624	9350	0.0011	-
0.3643	9400	0.0008	-
0.3662	9450	0.001	-
0.3682	9500	0.0006	-
0.3701	9550	0.0	-
0.3720	9600	0.0	-
0.3740	9650	0.0004	-
0.3759	9700	0.0	-
0.3779	9750	0.0	-
0.3798	9800	0.0	-
0.3817	9850	0.0	-
0.3837	9900	0.0	-
0.3856	9950	0.0	-
0.3876	10000	0.0	-
0.3895	10050	0.0	-
0.3914	10100	0.0	-
0.3934	10150	0.0	-
0.3953	10200	0.0	-
0.3972	10250	0.0	-
0.3992	10300	0.0	-
0.4011	10350	0.0	-
0.4031	10400	0.0	-
0.4050	10450	0.0	-
0.4069	10500	0.0	-
0.4089	10550	0.0	-
0.4108	10600	0.0	-
0.4127	10650	0.0	-
0.4147	10700	0.0	-
0.4166	10750	0.0	-
0.4186	10800	0.0	-
0.4205	10850	0.0	-
0.4224	10900	0.0	-
0.4244	10950	0.0	-
0.4263	11000	0.0	-
0.4282	11050	0.0	-
0.4302	11100	0.0	-
0.4321	11150	0.0	-
0.4341	11200	0.0	-
0.4360	11250	0.0	-
0.4379	11300	0.0	-
0.4399	11350	0.0	-
0.4418	11400	0.0	-
0.4437	11450	0.0	-
0.4457	11500	0.0	-
0.4476	11550	0.0	-
0.4496	11600	0.0	-
0.4515	11650	0.0	-
0.4534	11700	0.0	-
0.4554	11750	0.0	-
0.4573	11800	0.0	-
0.4592	11850	0.0	-
0.4612	11900	0.0	-
0.4631	11950	0.0	-
0.4651	12000	0.0	-
0.4670	12050	0.0	-
0.4689	12100	0.0	-
0.4709	12150	0.0	-
0.4728	12200	0.0	-
0.4748	12250	0.0	-
0.4767	12300	0.0	-
0.4786	12350	0.0	-
0.4806	12400	0.0	-
0.4825	12450	0.0	-
0.4844	12500	0.0	-
0.4864	12550	0.0	-
0.4883	12600	0.0	-
0.4903	12650	0.0	-
0.4922	12700	0.0	-
0.4941	12750	0.0	-
0.4961	12800	0.0	-
0.4980	12850	0.0	-
0.4999	12900	0.0	-
0.5019	12950	0.0	-
0.5038	13000	0.0	-
0.5058	13050	0.0	-
0.5077	13100	0.0	-
0.5096	13150	0.0	-
0.5116	13200	0.0	-
0.5135	13250	0.0	-
0.5154	13300	0.0	-
0.5174	13350	0.0	-
0.5193	13400	0.0	-
0.5213	13450	0.0	-
0.5232	13500	0.0	-
0.5251	13550	0.0	-
0.5271	13600	0.0	-
0.5290	13650	0.0	-
0.5309	13700	0.0	-
0.5329	13750	0.0	-
0.5348	13800	0.0	-
0.5368	13850	0.0	-
0.5387	13900	0.0	-
0.5406	13950	0.0	-
0.5426	14000	0.0	-
0.5445	14050	0.0	-
0.5464	14100	0.0	-
0.5484	14150	0.0	-
0.5503	14200	0.0	-
0.5523	14250	0.0	-
0.5542	14300	0.0	-
0.5561	14350	0.0	-
0.5581	14400	0.0	-
0.5600	14450	0.0	-
0.5620	14500	0.0	-
0.5639	14550	0.0	-
0.5658	14600	0.0	-
0.5678	14650	0.0	-
0.5697	14700	0.0	-
0.5716	14750	0.0	-
0.5736	14800	0.0	-
0.5755	14850	0.0	-
0.5775	14900	0.0	-
0.5794	14950	0.0	-
0.5813	15000	0.0	-
0.5833	15050	0.0	-
0.5852	15100	0.0	-
0.5871	15150	0.0	-
0.5891	15200	0.0	-
0.5910	15250	0.0	-
0.5930	15300	0.0	-
0.5949	15350	0.0	-
0.5968	15400	0.0	-
0.5988	15450	0.0	-
0.6007	15500	0.0	-
0.6026	15550	0.0	-
0.6046	15600	0.0	-
0.6065	15650	0.0	-
0.6085	15700	0.0	-
0.6104	15750	0.0	-
0.6123	15800	0.0	-
0.6143	15850	0.0	-
0.6162	15900	0.0	-
0.6181	15950	0.0	-
0.6201	16000	0.0	-
0.6220	16050	0.0	-
0.6240	16100	0.0	-
0.6259	16150	0.0	-
0.6278	16200	0.0	-
0.6298	16250	0.0	-
0.6317	16300	0.0	-
0.6336	16350	0.0	-
0.6356	16400	0.0	-
0.6375	16450	0.0	-
0.6395	16500	0.0	-
0.6414	16550	0.0	-
0.6433	16600	0.0	-
0.6453	16650	0.0	-
0.6472	16700	0.0	-
0.6491	16750	0.0	-
0.6511	16800	0.0	-
0.6530	16850	0.0	-
0.6550	16900	0.0	-
0.6569	16950	0.0	-
0.6588	17000	0.0	-
0.6608	17050	0.0	-
0.6627	17100	0.0	-
0.6647	17150	0.0	-
0.6666	17200	0.0	-
0.6685	17250	0.0	-
0.6705	17300	0.0	-
0.6724	17350	0.0	-
0.6743	17400	0.0	-
0.6763	17450	0.0	-
0.6782	17500	0.0	-
0.6802	17550	0.0	-
0.6821	17600	0.0	-
0.6840	17650	0.0	-
0.6860	17700	0.0	-
0.6879	17750	0.0	-
0.6898	17800	0.0	-
0.6918	17850	0.0	-
0.6937	17900	0.0	-
0.6957	17950	0.0	-
0.6976	18000	0.0	-
0.6995	18050	0.0	-
0.7015	18100	0.0	-
0.7034	18150	0.0	-
0.7053	18200	0.0	-
0.7073	18250	0.0	-
0.7092	18300	0.0	-
0.7112	18350	0.0	-
0.7131	18400	0.0	-
0.7150	18450	0.0	-
0.7170	18500	0.0	-
0.7189	18550	0.0	-
0.7208	18600	0.0	-
0.7228	18650	0.0	-
0.7247	18700	0.0	-
0.7267	18750	0.0	-
0.7286	18800	0.0	-
0.7305	18850	0.0	-
0.7325	18900	0.0	-
0.7344	18950	0.0	-
0.7363	19000	0.0	-
0.7383	19050	0.0	-
0.7402	19100	0.0	-
0.7422	19150	0.0	-
0.7441	19200	0.0	-
0.7460	19250	0.0	-
0.7480	19300	0.0	-
0.7499	19350	0.0	-
0.7519	19400	0.0	-
0.7538	19450	0.0	-
0.7557	19500	0.0	-
0.7577	19550	0.0	-
0.7596	19600	0.0	-
0.7615	19650	0.0	-
0.7635	19700	0.0	-
0.7654	19750	0.0	-
0.7674	19800	0.0	-
0.7693	19850	0.0	-
0.7712	19900	0.0	-
0.7732	19950	0.0	-
0.7751	20000	0.0	-
0.7770	20050	0.0	-
0.7790	20100	0.0	-
0.7809	20150	0.0	-
0.7829	20200	0.0	-
0.7848	20250	0.0	-
0.7867	20300	0.0	-
0.7887	20350	0.0	-
0.7906	20400	0.0	-
0.7925	20450	0.0	-
0.7945	20500	0.0	-
0.7964	20550	0.0	-
0.7984	20600	0.0	-
0.8003	20650	0.0	-
0.8022	20700	0.0	-
0.8042	20750	0.0	-
0.8061	20800	0.0	-
0.8080	20850	0.0	-
0.8100	20900	0.0	-
0.8119	20950	0.0	-
0.8139	21000	0.0	-
0.8158	21050	0.0	-
0.8177	21100	0.0	-
0.8197	21150	0.0	-
0.8216	21200	0.0	-
0.8235	21250	0.0	-
0.8255	21300	0.0	-
0.8274	21350	0.0	-
0.8294	21400	0.0	-
0.8313	21450	0.0	-
0.8332	21500	0.0	-
0.8352	21550	0.0	-
0.8371	21600	0.0	-
0.8390	21650	0.0	-
0.8410	21700	0.0	-
0.8429	21750	0.0	-
0.8449	21800	0.0	-
0.8468	21850	0.0	-
0.8487	21900	0.0	-
0.8507	21950	0.0	-
0.8526	22000	0.0	-
0.8546	22050	0.0	-
0.8565	22100	0.0	-
0.8584	22150	0.0	-
0.8604	22200	0.0	-
0.8623	22250	0.0	-
0.8642	22300	0.0	-
0.8662	22350	0.0	-
0.8681	22400	0.0	-
0.8701	22450	0.0	-
0.8720	22500	0.0	-
0.8739	22550	0.0	-
0.8759	22600	0.0	-
0.8778	22650	0.0	-
0.8797	22700	0.0	-
0.8817	22750	0.0	-
0.8836	22800	0.0	-
0.8856	22850	0.0	-
0.8875	22900	0.0	-
0.8894	22950	0.0	-
0.8914	23000	0.0	-
0.8933	23050	0.0	-
0.8952	23100	0.0	-
0.8972	23150	0.0	-
0.8991	23200	0.0	-
0.9011	23250	0.0	-
0.9030	23300	0.0	-
0.9049	23350	0.0	-
0.9069	23400	0.0	-
0.9088	23450	0.0	-
0.9107	23500	0.0	-
0.9127	23550	0.0	-
0.9146	23600	0.0	-
0.9166	23650	0.0	-
0.9185	23700	0.0	-
0.9204	23750	0.0	-
0.9224	23800	0.0	-
0.9243	23850	0.0	-
0.9262	23900	0.0	-
0.9282	23950	0.0	-
0.9301	24000	0.0	-
0.9321	24050	0.0	-
0.9340	24100	0.0	-
0.9359	24150	0.0	-
0.9379	24200	0.0	-
0.9398	24250	0.0	-
0.9418	24300	0.0	-
0.9437	24350	0.0	-
0.9456	24400	0.0	-
0.9476	24450	0.0	-
0.9495	24500	0.0	-
0.9514	24550	0.0	-
0.9534	24600	0.0	-
0.9553	24650	0.0	-
0.9573	24700	0.0	-
0.9592	24750	0.0	-
0.9611	24800	0.0	-
0.9631	24850	0.0	-
0.9650	24900	0.0	-
0.9669	24950	0.0	-
0.9689	25000	0.0	-
0.9708	25050	0.0	-
0.9728	25100	0.0	-
0.9747	25150	0.0	-
0.9766	25200	0.0	-
0.9786	25250	0.0	-
0.9805	25300	0.0	-
0.9824	25350	0.0	-
0.9844	25400	0.0	-
0.9863	25450	0.0	-
0.9883	25500	0.0	-
0.9902	25550	0.0	-
0.9921	25600	0.0	-
0.9941	25650	0.0	-
0.9960	25700	0.0	-
0.9979	25750	0.0	-
0.9999	25800	0.0	-

Framework Versions

Python: 3.10.12
SetFit: 1.1.0
Sentence Transformers: 3.3.1
Transformers: 4.42.2
PyTorch: 2.5.1+cu121
Datasets: 3.2.0
Tokenizers: 0.19.1

Citation

BibTeX

bibtex

@article{https://doi.org/10.48550/arxiv.2209.11055,
    doi = {10.48550/ARXIV.2209.11055},
    url = {https://arxiv.org/abs/2209.11055},
    author = {Tunstall, Lewis and Reimers, Nils and Jo, Unso Eun Seo and Bates, Luke and Korat, Daniel and Wasserblat, Moshe and Pereg, Oren},
    keywords = {Computation and Language (cs.CL), FOS: Computer and information sciences, FOS: Computer and information sciences},
    title = {Efficient Few-Shot Learning Without Prompts},
    publisher = {arXiv},
    year = {2022},
    copyright = {Creative Commons Attribution 4.0 International}
}