Model save

README.md

@@ -1,73 +1,87 @@
 ---
-license: mit
-language:
-- en
-datasets:
-- speechbrain/LoquaciousSet
-base_model:
-- openai/whisper-large-v3-turbo
-- HuggingFaceTB/SmolLM3-3B
-pipeline_tag: automatic-speech-recognition
+library_name: transformers
 tags:
-- asr
-- speech-recognition
-- audio
-- smollm
-- whisper
-- mlp
+- generated_from_trainer
+model-index:
+- name: tiny-audio
+  results: []
 ---
 
-# Tiny Audio
-
-A speech recognition model trained in 24 hours on a single GPU for ~$12. Built with the [Tiny Audio](https://github.com/alexkroman/tiny-audio) codebase—a minimal, hackable framework for training ASR models.
-
-## Architecture
-
-```
-Audio (16kHz) → Whisper Encoder (frozen) → MLP Projector (trained) → SmolLM3-3B (frozen) → Text
-```
-
-**MLP Projector:**
-- Convolutional downsampling: 4x sequence compression via two stride-2 conv layers
-- Linear (1280 → 2048) → GELU → Linear (2048 → 2048)
-- Output normalization: RMSNorm
-
-## Training Details
-
-| | |
-|---|---|
-| **Dataset** | LoquaciousSet (25,000 hours) |
-| **Hardware** | Single NVIDIA A40 40GB |
-| **Training Time** | ~24 hours |
-| **Cost** | ~$12 |
-| **Trainable Parameters** | ~12M (projector only) |
-
-## Performance
-
-**Word Error Rate (WER): 12.14%** on LoquaciousSet test set.
-
-See the [community leaderboard](https://github.com/alexkroman/tiny-audio#leaderboard) for comparisons.
-
-## Usage
-
-```python
-from transformers import pipeline
-
-pipe = pipeline("automatic-speech-recognition", model="mazesmazes/tiny-audio", trust_remote_code=True)
-
-result = pipe("path/to/audio.wav")
-print(result["text"])
-```
-
-## Limitations
-
-- English only
-- Optimized for 16kHz audio; other sample rates are resampled automatically
-- Performance may degrade on heavily accented speech, noisy environments, or domain-specific jargon
-- Maximum audio length limited by context window
-
-## Learn More
-
-- **[Train your own model](https://github.com/alexkroman/tiny-audio)** — The full codebase with training scripts
-- **[Free 3-hour course](https://github.com/alexkroman/tiny-audio/blob/main/docs/course/0-course-overview.md)** — Build your own ASR system from scratch
-- **[Submit to leaderboard](https://github.com/alexkroman/tiny-audio#leaderboard)** — Share your trained model
+<!-- This model card has been generated automatically according to the information the Trainer had access to. You
+should probably proofread and complete it, then remove this comment. -->
+
+# tiny-audio
+
+This model is a fine-tuned version of [](https://huggingface.co/) on an unknown dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.2281
+
+## Model description
+
+More information needed
+
+## Intended uses & limitations
+
+More information needed
+
+## Training and evaluation data
+
+More information needed
+
+## Training procedure
+
+### Training hyperparameters
+
+The following hyperparameters were used during training:
+- learning_rate: 0.0003
+- train_batch_size: 6
+- eval_batch_size: 6
+- seed: 42
+- gradient_accumulation_steps: 3
+- total_train_batch_size: 18
+- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.95) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
+- lr_scheduler_type: cosine
+- lr_scheduler_warmup_steps: 500
+- num_epochs: 1
+
+### Training results
+
+| Training Loss | Epoch  | Step  | Validation Loss |
+|:-------------:|:------:|:-----:|:---------------:|
+| 0.4431        | 0.0298 | 2000  | 0.3491          |
+| 0.4701        | 0.0596 | 4000  | 0.3217          |
+| 0.4086        | 0.0894 | 6000  | 0.3092          |
+| 0.3937        | 0.1192 | 8000  | 0.2949          |
+| 0.336         | 0.1490 | 10000 | 0.2896          |
+| 0.3609        | 0.1788 | 12000 | 0.2827          |
+| 0.342         | 0.3128 | 14000 | 0.2654          |
+| 0.3576        | 0.3575 | 16000 | 0.2667          |
+| 0.3266        | 0.4022 | 18000 | 0.2550          |
+| 0.2951        | 0.3352 | 20000 | 0.2637          |
+| 0.3089        | 0.3687 | 22000 | 0.2646          |
+| 0.2892        | 0.4022 | 24000 | 0.2606          |
+| 0.3752        | 0.4357 | 26000 | 0.2547          |
+| 0.2865        | 0.4692 | 28000 | 0.2535          |
+| 0.327         | 0.5027 | 30000 | 0.2494          |
+| 0.3438        | 0.5363 | 32000 | 0.2453          |
+| 0.2843        | 0.5698 | 34000 | 0.2405          |
+| 0.3015        | 0.6033 | 36000 | 0.2374          |
+| 0.2904        | 0.6368 | 38000 | 0.2364          |
+| 0.2946        | 0.6703 | 40000 | 0.2340          |
+| 0.3428        | 0.7038 | 42000 | 0.2323          |
+| 0.3036        | 0.7374 | 44000 | 0.2299          |
+| 0.3381        | 0.7709 | 46000 | 0.2293          |
+| 0.2993        | 0.8044 | 48000 | 0.2291          |
+| 0.302         | 0.8379 | 50000 | 0.2282          |
+| 0.2779        | 0.8714 | 52000 | 0.2280          |
+| 0.2856        | 0.9049 | 54000 | 0.2281          |
+| 0.2904        | 0.9384 | 56000 | 0.2280          |
+| 0.3048        | 0.9720 | 58000 | 0.2281          |
+
+
+### Framework versions
+
+- Transformers 4.57.3
+- Pytorch 2.8.0+cu128
+- Datasets 3.6.0
+- Tokenizers 0.22.1