GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
abhishekgautamm
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
abhishekgautamm
Model Tree
openai/whisper-base
Input Modalities
Output Modalities
Supported Functionality
| Path | Format | Size | Use with |
|---|---|---|---|
/ (root) | Transformers, fp32 safetensors | 290 MB | transformers |
ct2/fp32/ | CTranslate2, float32 | 290 MB | faster-whisper |
ct2/int8/ | CTranslate2, int8 | 76 MB | faster-whisper (fast on CPU) |
The int8 build is within 0.2 WER points of fp32 on the held-out test sets below (same decoding, GPU).
Transformers
python
from transformers import pipeline asr = pipeline("automatic-speech-recognition", model="abhishekgautamm/whisper-base-hinglish", chunk_length_s=30)print(asr("clip.wav", generate_kwargs={"language": "en", "task": "transcribe"})["text"])faster-whisper (CTranslate2)
python
from huggingface_hub import snapshot_downloadfrom faster_whisper import WhisperModel path = snapshot_download("abhishekgautamm/whisper-base-hinglish-v2", allow_patterns=["ct2/int8/*"])model = WhisperModel(f"{path}/ct2/int8", device="cpu", compute_type="int8")segments, _ = model.transcribe("clip.wav", language="en", beam_size=1, without_timestamps=True)print("".join(s.text for s in segments).strip())Always decode with language="en" and without timestamps (without_timestamps=True in faster-whisper; transformers does this by default). The model was trained to emit Roman text under the English token,
for Hindi speech too. Adding no_repeat_ngram_size=3, repetition_penalty=1.1 (as in the evaluation) suppresses
rare repetition loops.
Word / character error rate (%), greedy decoding, fp32. References and outputs are lower-cased, punctuation-stripped and digits spelled out; Devanagari references are romanized with LIPI (the same tool used to build the training targets).
Public benchmarks, full test sets (none of these were used in training)
| Dataset | Domain | Clips | WER | CER |
|---|---|---|---|---|
| Svarah | Indian-accented English | 6,656 | 24.1 | 13.9 |
| Monsoon en-IN | Indian English | 2,102 | 16.4 | 8.7 |
| CoSHE-Eval | Long-form Hinglish (~60 s clips) | 1,985 | 36.6 | 19.8 |
| FLEURS hi_in | Hindi (read) | 418 | 34.5 | 16.0 |
Held-out test sets of the training corpora (official test splits; SPIRE/wSPIRE by held-out speaker, VoiceBench by held-out prompt)
| Dataset | Domain | Clips | WER | CER |
|---|---|---|---|---|
| IISc wSPIRE | Indian English, incl. whispered | 150 | 2.4 | 1.2 |
| IISc SPIRE-SIES | Spontaneous technical English | 150 | 23.9 | 14.8 |
| MUCS 2021 (Hi-En) | Hindi-English code-switch | 250 | 28.4 | 19.3 |
| MUCS 2021 (Hindi) | Hindi (read) | 250 | 29.2 | 12.6 |
| HiACC (adult) | Hinglish | 150 | 18.1 | 8.4 |
| HiACC (children) | Hinglish | 150 | 17.8 | 7.8 |
| VoiceBench | Spoken assistant prompts | 250 | 16.9 | 8.7 |
| SLURP | Assistant commands | 250 | 25.3 | 16.2 |
| Fluent Speech Commands | Short commands | 100 | 2.9 | 1.9 |
| Timers and Such | Numbers / times | 200 | 2.9 | 2.4 |
About 250 hours, with every target romanized by LIPI: MUCS 2021 Hindi-English (86 h), Common Voice English (48 h),
IISc wSPIRE (31 h), IISc SPIRE-SIES (26 h), MUCS 2021 Hindi (14 h), VoiceBench (13 h), SLURP (7 h),
Timers and Such (5 h), HiACC (4 h), Fluent Speech Commands (0.5 h). The language token is forced to en. Training ran for 3 epochs (released checkpoint),
LR 1e-4 with 300 warm-up steps, fp16 mixed precision.
hai / hain, ki / kii).from transformers import pipeline asr = pipeline("automatic-speech-recognition", model="abhishekgautamm/whisper-base-hinglish", chunk_length_s=30)print(asr("clip.wav", generate_kwargs={"language": "en", "task": "transcribe"})["text"])from huggingface_hub import snapshot_downloadfrom faster_whisper import WhisperModel path = snapshot_download("abhishekgautamm/whisper-base-hinglish-v2", allow_patterns=["ct2/int8/*"])model = WhisperModel(f"{path}/ct2/int8", device="cpu", compute_type="int8")segments, _ = model.transcribe("clip.wav", language="en", beam_size=1, without_timestamps=True)print("".join(s.text for s in segments).strip())