Update README.md

README.md

@@ -24,17 +24,16 @@ OpenAI's [Whisper large-v3](https://huggingface.co/openai/whisper-large-v3), the
 to date.
 
 Compared to previous Distil-Whisper models, the distillation procedure for distil-large-v3 has been adapted to give 
-**superior long-form transcription accuracy** with OpenAI's **sequential long-form algorithm**. The result is a distilled model 
-that performs to within 1% WER of large-v3 on long-form audio, independent of the transcription algorithm, and 
-outperforms distil-large-v2 by 4.8% and 0.7% using the sequential and chunked long-form algorithms respectively.
-Furthermore, the model is also faster than previous Distil-Whisper models: **5.7x faster than large-v3**, and 1.1x faster 
-than distil-large-v2.
+**superior long-form transcription accuracy** with OpenAI's **sequential long-form algorithm**. The result is a distilled 
+model that performs to within 1% WER of large-v3 on long-form audio using both the sequential and chunked algorithms, and 
+outperforms distil-large-v2 by 4.8% using the sequential algorithm. The model is also faster than previous Distil-Whisper 
+models: **6.3x faster than large-v3**, and 1.1x faster than distil-large-v2.
 
 | Model                                                                        | Params / M | Rel. Latency | Short-Form | Sequential Long-Form | Chunked Long-Form |
 |------------------------------------------------------------------------------|------------|--------------|------------|----------------------|-------------------|
 | [large-v3](https://huggingface.co/openai/whisper-large-v3)                   | 1550       | 1.0          | 8.4        | 10.0                 | 11.0              |
-| **[distil-large-v3](https://huggingface.co/distil-whisper/distil-large-v3)** | **756**    | **5.7**      | **9.8**    | **10.8**             | **10.9**          |
-| [distil-large-v2](https://huggingface.co/distil-whisper/distil-large-v2)     | 756        | 5.2          | 9.6        | 15.6                 | 11.6              |
+| **[distil-large-v3](https://huggingface.co/distil-whisper/distil-large-v3)** | **756**    | **6.3**      | **9.7**    | **10.8**             | **10.9**          |
+| [distil-large-v2](https://huggingface.co/distil-whisper/distil-large-v2)     | 756        | 5.8          | 9.6        | 15.6                 | 11.6              |
 
 Since the sequential algorithm is the "de-facto" transcription algorithm across the most popular Whisper libraries 
 (Whisper cpp, Faster-Whisper, OpenAI Whisper), this distilled model is designed to be compatible across all libraries types. 
@@ -49,15 +48,15 @@ with instructions for getting started [here](#library-integrations).
    * [Sequential Long-Form](#sequential-long-form)
    * [Chunked Long-Form](#chunked-long-form)
    * [Speculative Decoding](#speculative-decoding)
-2. [Additional Speed and Memory Improvements](#additional-speed--memory-improvements)
-3. [Library Integrations](#library-integrations)
-   * [OpenAI Whisper](#openai-whisper)
+   * [Additional Speed and Memory Improvements](#additional-speed--memory-improvements)
+2. [Library Integrations](#library-integrations)
    * [Whisper cpp](#whispercpp)
    * [Faster Whisper](#faster-whisper)
+   * [OpenAI Whisper](#openai-whisper)
    * [Transformers.js](#transformersjs)
    * [Candle](#candle)
-4. [Model Details](#model-details)
-5. [License](#license)
+3. [Model Details](#model-details)
+4. [License](#license)
 
 ## Transformers Usage
 
@@ -251,27 +250,27 @@ dataset = dataset.cast_column("audio", Audio(processor.feature_extractor.samplin
 sample = dataset[0]["audio"]
 
 inputs = processor(
-  sample["array"],
-  sampling_rate=sample["sampling_rate"],
-  return_tensors="pt",
-  truncation=False,
-  padding="longest",
-  return_attention_mask=True,
+    sample["array"],
+    sampling_rate=sample["sampling_rate"],
+    return_tensors="pt",
+    truncation=False,
+    padding="longest",
+    return_attention_mask=True,
 )
 inputs = inputs.to(device, dtype=torch_dtype)
 
 gen_kwargs = {
-  "max_new_tokens": 448,
-  "num_beams": 1,  # set > 1 for beam-search
-  "condition_on_prev_tokens": False,  # condition for previous context
-  "compression_ratio_threshold": 1.35,  # zlib compression ratio threshold (in token space)
-  "temperature": (0.0, 0.2, 0.4, 0.6, 0.8, 1.0),  # temperature fallback
-  "logprob_threshold": -1.0,
-  "no_speech_threshold": 0.6,
-  "return_timestamps": True,
+    "max_new_tokens": 448,
+    "num_beams": 1,  # set > 1 for beam-search
+    "condition_on_prev_tokens": False,  # condition for previous context
+    "compression_ratio_threshold": 1.35,  # zlib compression ratio threshold (in token space)
+    "temperature": (0.0, 0.2, 0.4, 0.6, 0.8, 1.0),  # temperature fallback
+    "logprob_threshold": -1.0,
+    "no_speech_threshold": 0.6,
+    "return_timestamps": True,
 }
 
-pred_ids = model.generate(**inputs, **gen_kwargs)
+pred_ids = model.generate(**i   nputs, **gen_kwargs)
 pred_text = processor.batch_decode(pred_ids, skip_special_tokens=True, decode_with_timestamps=False)
 
 print(pred_text)
@@ -379,13 +378,13 @@ print(result["text"])
 
 For more details on speculative decoding, refer to the blog post [Speculative Decoding for 2x Faster Whisper Inference](https://huggingface.co/blog/whisper-speculative-decoding).
 
-## Additional Speed & Memory Improvements
+### Additional Speed & Memory Improvements
 
 You can apply additional speed and memory improvements to Distil-Whisper to further reduce the inference speed and VRAM 
 requirements. These optimisations primarily target the attention kernel, swapping it from an eager implementation to a 
 more efficient flash attention version.
 
-### Flash Attention 2
+#### Flash Attention 2
 
 We recommend using [Flash-Attention 2](https://huggingface.co/docs/transformers/main/en/perf_infer_gpu_one#flashattention-2) 
 if your GPU allows for it. To do so, you first need to install [Flash Attention](https://github.com/Dao-AILab/flash-attention):
@@ -401,14 +400,16 @@ Then pass `attn_implementation="flash_attention_2"` to `from_pretrained`:
 + model = AutoModelForSpeechSeq2Seq.from_pretrained(model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True, attn_implementation="flash_attention_2")
 ```
 
-### Torch Scale-Product-Attention (SDPA)
+#### Torch Scale-Product-Attention (SDPA)
 
 If your GPU does not support Flash Attention, we recommend making use of PyTorch [scaled dot-product attention (SDPA)](https://pytorch.org/docs/stable/generated/torch.nn.functional.scaled_dot_product_attention.html). 
 This attention implementation is activated **by default** for PyTorch versions 2.1.1 or greater. To check 
-whether you have a compatible PyTorch version, run the following command:
+whether you have a compatible PyTorch version, run the following Python code snippet:
 
-```bash
-python -c "from transformers.utils import is_torch_sdpa_available; print(is_torch_sdpa_available())"
+```python
+from transformers.utils import is_torch_sdpa_available
+
+print(is_torch_sdpa_available())
 ```
 
 If the above returns `True`, you have a valid version of PyTorch installed and SDPA is activated by default. If it 
@@ -422,11 +423,11 @@ Once a valid PyTorch version is installed, SDPA is activated by default. It can
 + model = AutoModelForSpeechSeq2Seq.from_pretrained(model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True, attn_implementation="sdpa")
 ```
 
-### Torch compile
+#### Torch compile
 
 Coming soon...
 
-### 4-bit and 8-bit Inference
+#### 4-bit and 8-bit Inference
 
 Coming soon...
 
@@ -445,16 +446,18 @@ Steps for getting started:
 git clone https://github.com/ggerganov/whisper.cpp.git
 cd whisper.cpp
 ```
-2. Download the ggml weights for distil-large-v3 from the Hugging Face Hub:
+2. Download the GGML weights for distil-large-v3 from the Hugging Face Hub using the following Python snippet:
 
-```bash
-python -c "from huggingface_hub import hf_hub_download; hf_hub_download(repo_id='distil-whisper/distil-large-v3', filename='ggml-distil-large-v3.bin', local_dir='./models')"
+```python
+from huggingface_hub import hf_hub_download
+
+hf_hub_download(repo_id='distil-whisper/distil-large-v3-ggml', filename='ggml-distil-large-v3.bin', local_dir='./models')
 ```
 
 Note that if you do not have the `huggingface_hub` package installed, you can also download the weights with `wget`:
 
 ```bash
-wget https://huggingface.co/distil-whisper/distil-large-v3/resolve/main/ggml-distil-large-v3.bin -P ./models
+wget https://huggingface.co/distil-whisper/distil-large-v3-ggml/resolve/main/ggml-distil-large-v3.bin -P ./models
 ```
 
 3. Run inference using the provided sample audio:
@@ -834,8 +837,9 @@ If you use this model, please consider citing the [Distil-Whisper paper](https:/
 * OpenAI for the Whisper [model](https://huggingface.co/openai/whisper-large-v3), in particular Jong Wook Kim for the [original codebase](https://github.com/openai/whisper) and training discussions
 * Hugging Face 🤗 [Transformers](https://github.com/huggingface/transformers) for the model integration
 * [Georgi Gerganov](https://huggingface.co/ggerganov) for the Whisper cpp integration
+* [Systran team](https://github.com/SYSTRAN) for the Faster-Whisper integration
 * [Joshua Lochner](https://huggingface.co/xenova) for the Transformers.js integration
 * [Laurent Mazare](https://huggingface.co/lmz) for the Candle integration
 * [Vaibhav Srivastav](https://huggingface.co/reach-vb) for Distil-Whisper distribution
 * Google's [TPU Research Cloud (TRC)](https://sites.research.google/trc/about/) programme for Cloud TPU v4 compute resource
-* [Raghav Sonavane](https://huggingface.co/rsonavane/distil-whisper-large-v2-8-ls) for an early iteration of Distil-Whisper on the LibriSpeech dataset
+* [Raghav Sonavane](https://huggingface.co/rsonavane/distil-whisper-large-v2-8-ls) for an early iteration of Distil-Whisper on the LibriSpeech dataset