7 changed files with 53800 additions and 17934 deletions
--- a/dev-0/out-hidden_size=128.tsv
+++ b/dev-0/out-hidden_size=128.tsv
--- a/dev-0/out-hidden_size=512.tsv
+++ b/dev-0/out-hidden_size=512.tsv
--- a/dev-0/out.tsv
+++ b/dev-0/out.tsv
--- a/run-fine-tuning.py
+++ b/run-fine-tuning.py
@ -201,7 +201,7 @@ def predict_words(dataset):
        src = tokenizer.encode(text, return_tensors="pt", truncation=True).to(device)
        output = model.generate(src, max_length=len(src[0]) + 1, do_sample=True, top_k=0, temperature=0.8,
                                num_return_sequences=1, no_repeat_ngram_size=2, output_scores=True)
-        probs, idxs = torch.softmax(output.scores[0][-1], dim=0).topk(30)
+        probs, idxs = torch.softmax(output.scores[0][-1], dim=0).topk(50)
        current_output = ''
        accumulated_probability = 0
        for prob, token_id in zip(probs, idxs):
--- a/test-A/out-hidden_size=128.tsv
+++ b/test-A/out-hidden_size=128.tsv
--- a/test-A/out-hidden_size=512.tsv
+++ b/test-A/out-hidden_size=512.tsv
--- a/test-A/out.tsv
+++ b/test-A/out.tsv