Error when using text2vec-huggingface mpnet v2

Support
1

Hi,

I’m using WCS and have a Schema that uses text2vec-huggingface with the model paraphrase-multilingual-mpnet-base-v2. When i do a nearText search, i’ll get the following error:

explorer: get class: vectorize params: vectorize params: vectorize params: vectorize keywords: remote client vectorize: failed with status: 400 error: [Errno 2] No such file or directory: ‘/data/sentence-transformers_paraphrase-multilingual-mpnet-base-v2/sentence_xlnet_config.json’

I can’t figure out what goes wrong. Can anyone help me out?

The schema:

    {
      "class": "Bp2_scrapedPageChunks_v10",
      "invertedIndexConfig": {
        "bm25": {
          "b": 0.75,
          "k1": 1.2
        },
        "cleanupIntervalSeconds": 60,
        "stopwords": {
          "additions": null,
          "preset": "en",
          "removals": null
        }
      },
      "moduleConfig": {
        "text2vec-contextionary": {
          "vectorizeClassName": false
        },
        "text2vec-huggingface": {
          "model": "sentence-transformers/paraphrase-multilingual-mpnet-base-v2",
          "options": {
            "useCache": true,
            "useGPU": false,
            "waitForModel": true
          },
          "vectorizeClassName": false
        }
      },
      "properties": [
        {
          "dataType": [
            "text"
          ],
          "indexFilterable": true,
          "indexSearchable": true,
          "moduleConfig": {
            "text2vec-contextionary": {
              "vectorizePropertyName": false
            },
            "text2vec-huggingface": {
              "skip": false,
              "vectorizePropertyName": false
            }
          },
          "name": "content",
          "tokenization": "word"
        }
      ],
      "replicationConfig": {
        "factor": 1
      },
      "shardingConfig": {
        "virtualPerPhysical": 128,
        "desiredCount": 1,
        "actualCount": 1,
        "desiredVirtualCount": 128,
        "actualVirtualCount": 128,
        "key": "_id",
        "strategy": "hash",
        "function": "murmur3"
      },
      "vectorIndexConfig": {
        "skip": false,
        "cleanupIntervalSeconds": 300,
        "maxConnections": 64,
        "efConstruction": 128,
        "ef": -1,
        "dynamicEfMin": 100,
        "dynamicEfMax": 500,
        "dynamicEfFactor": 8,
        "vectorCacheMaxObjects": 1000000000000,
        "flatSearchCutoff": 40000,
        "distance": "cosine",
        "pq": {
          "enabled": false,
          "bitCompression": false,
          "segments": 0,
          "centroids": 256,
          "encoder": {
            "type": "kmeans",
            "distribution": "log-normal"
          }
        }
      },
      "vectorIndexType": "hnsw",
      "vectorizer": "text2vec-huggingface"
    }
2

Hi @joris - welcome!

That looks to me like it might be a problem with the HF inference API, but not sure what is exactly going on there. Have you tried other models by the way?

And @antas-marcin - have you seen this before?

3

Hi @jphwang ,

Same error for sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2

explorer: get class: vectorize params: vectorize params: vectorize params: vectorize keywords: remote client vectorize: failed with status: 400 error: [Errno 2] No such file or directory: ‘/data/sentence-transformers_paraphrase-multilingual-MiniLM-L12-v2/sentence_xlnet_config.json’

But, intfloat/multilingual-e5-base seems to work just fine! Thanks for your suggestion

1 Like
4

Hello @joris!

this error comes from HugggingFace API. I get the same error when I query HuggingFace API directly for embeddings:

{
    "error": "[Errno 2] No such file or directory: '/data/sentence-transformers_paraphrase-multilingual-mpnet-base-v2/sentence_xlnet_config.json'"
}

I think that you should switch the model or if you want to use this particular model then you would have to contact HuggingFace so that they would fix this.