George Saon

gsaon's activity

• All
• Models
• Datasets
• Spaces
• Buckets
• Papers
• Collections
• Community
• Posts
• Upvotes
• Likes
• Articles

New activity in ibm-granite/granite-speech-streaming-webgpu 26 days ago

Fix mic-start race causing orphaned dictation sessions on Safari

🔥 1

#1 opened 26 days ago by

kristunlee

New activity in ibm-granite/granite-speech-5.0-470m-turboctc-nc 28 days ago

Model systematically turn 's into is

👍 1

3

#8 opened 28 days ago by

nshmyrevgmail

New activity in ibm-granite/granite-speech-5.0-470m-turboctc-nc about 1 month ago

Demo for this model on Spaces

🔥 1

1

#7 opened about 1 month ago by

multimodalart

Update README with disclaimer

#6 opened about 1 month ago by

kristunlee

New activity in ibm-granite/granite-speech-5.0-470m-turboctc about 1 month ago

add pipeline_tag and library_name

#6 opened about 1 month ago by

eustlb

New activity in ibm-granite/granite-speech-5.0-470m-turboctc-nc about 1 month ago

convert to transformers format

👍 1

#4 opened about 1 month ago by

eustlb

New activity in ibm-granite/granite-speech-5.0-470m-turboctc about 1 month ago

bf16 as default dtype

#5 opened about 1 month ago by

eustlb

convert to transformers format

1

#4 opened about 1 month ago by

eustlb

New activity in ibm-granite/granite-speech-5.0-470m-turboctc-nc about 1 month ago

Update README.md

1

#3 opened about 1 month ago by

kristunlee

New activity in ibm-granite/granite-speech-5.0-470m-turboctc about 1 month ago

Unify tokenizer to 16384 with blank at id 0 (no +1 offset)

#3 opened about 1 month ago by

gsaon

New activity in ibm-granite/granite-speech-5.0-470m-turboctc-nc about 1 month ago

Simplify decode: remove num_special_tokens/token_id_offset (blank at id 0, 1:1 ids)

#2 opened about 1 month ago by

gsaon

Mirror model from gsaon/testCTC (masayuki 191a5h p25; bf16 weights)

#1 opened about 1 month ago by

gsaon

New activity in ibm-granite/granite-speech-5.0-470m-turboctc about 2 months ago

Update weights: robust T2623 p25_lr1e5_a0p4 (bf16, BN stats fp32)

#2 opened about 2 months ago by

gsaon

Add packaged CTC Conformer export (C-T26.23, roman_bpe tokenizer, bf16 weights)

#1 opened about 2 months ago by

gsaon

New activity in ibm-granite/granite-speech-4.1-2b 4 months ago

Is peak/loudness normalisation needed for quiet audio segments?

1

#9 opened 4 months ago by

GokseninYuksel

New activity in ibm-granite/granite-speech-4.1-2b 5 months ago

ONNX versions created:

🔥 2

1

#7 opened 5 months ago by

smcleod

What is the function of the weights in out_llm.safetensors?

1

#6 opened 5 months ago by

deathknight0

Update README

#4 opened 5 months ago by

kristunlee

Add punctuation and capitalization evaluation results

#2 opened 5 months ago by

gsaon

Upload folder using huggingface_hub

#1 opened 5 months ago by

gsaon