Samuel Fipps
Add evaluation results on the samsum config and train split of samsum (#4)
2775ced
|
raw
history blame
2.83 kB
metadata
language: en
tags:
  - summarization
model-index:
  - name: SamuelAllen123/t5-efficient-large-nl36_fine_tune_sum_V2
    results:
      - task:
          type: summarization
          name: Summarization
        dataset:
          name: samsum
          type: samsum
          config: samsum
          split: test
        metrics:
          - name: ROUGE-1
            type: rouge
            value: 50.4987
            verified: true
          - name: ROUGE-2
            type: rouge
            value: 25.6888
            verified: true
          - name: ROUGE-L
            type: rouge
            value: 41.7283
            verified: true
          - name: ROUGE-LSUM
            type: rouge
            value: 46.2626
            verified: true
          - name: loss
            type: loss
            value: 1.5158178806304932
            verified: true
          - name: gen_len
            type: gen_len
            value: 24.0342
            verified: true
      - task:
          type: summarization
          name: Summarization
        dataset:
          name: cnn_dailymail
          type: cnn_dailymail
          config: 3.0.0
          split: test
        metrics:
          - name: ROUGE-1
            type: rouge
            value: 34.4055
            verified: true
          - name: ROUGE-2
            type: rouge
            value: 14.127
            verified: true
          - name: ROUGE-L
            type: rouge
            value: 24.3353
            verified: true
          - name: ROUGE-LSUM
            type: rouge
            value: 31.6582
            verified: true
          - name: loss
            type: loss
            value: 2.4456119537353516
            verified: true
          - name: gen_len
            type: gen_len
            value: 45.928
            verified: true
      - task:
          type: summarization
          name: Summarization
        dataset:
          name: samsum
          type: samsum
          config: samsum
          split: train
        metrics:
          - name: ROUGE-1
            type: rouge
            value: 54.933
            verified: true
          - name: ROUGE-2
            type: rouge
            value: 31.7965
            verified: true
          - name: ROUGE-L
            type: rouge
            value: 47.0057
            verified: true
          - name: ROUGE-LSUM
            type: rouge
            value: 51.2027
            verified: true
          - name: loss
            type: loss
            value: 1.130684494972229
            verified: true
          - name: gen_len
            type: gen_len
            value: 23.7989
            verified: true

Trained on Samsum train split.

Parameters for training:

no_decay = ["bias", "LayerNorm.weight", "layer_norm.weight"] optimizer_grouped_parameters = [ { "params": [p for n, p in model.named_parameters() if not any(nd in n for nd in no_decay)], "weight_decay": 0.0, }, { "params": [p for n, p in model.named_parameters() if any(nd in n for nd in no_decay)], "weight_decay": 0.0, }, ]

lr = 0.00005 optimizer = torch.optim.RAdam(optimizer_grouped_parameters, lr=lr)

lr_scheduler = get_scheduler( name="linear", optimizer=optimizer, num_warmup_steps=0, num_training_steps=50005)

This was only for 10K steps

More details coming soon