Update README (#42)

README.md

@@ -40,7 +40,7 @@ press = ExpectedAttentionPress(compression_ratio=0.5)
 answer = pipe(context, question=question, press=press)["answer"]
 ```
 
-In the snippet above, the compression is only applied on the context tokens so that you can evaluate the compression for different questions. Check the [Wikipedia notebook demo](notebooks/wikipedia_demo.ipynb) for a more detailed example (also available on Colab [here](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1JNvaTKuuAHrl49dYB9-mdEH_y52Ib-NP?usp=drive_link)).
+In the snippet above, the compression is only applied on the context tokens so that you can evaluate the compression for different questions. Check the [Wikipedia notebook demo](notebooks/wikipedia_demo.ipynb) for a more detailed example (also available on Colab [here](https://colab.research.google.com/drive/1JNvaTKuuAHrl49dYB9-mdEH_y52Ib-NP)).
 
 > [!IMPORTANT]  
 > We focus on compression during the pre-filling phase as the KV cache becomes a bottleneck for long-context sequence (100k - 1M tokens) which are essentially long context prompts. This would typically apply to improving prompt caching systems.
@@ -54,26 +54,28 @@ We welcome contributions! To add a new press, simply open an issue or submit a p
 
 ## Available presses
 
-All current presses are training free. Several of them inherit from `ScorerPress` and rely on a score to prune the KV pairs with lowest importance:
+All current presses are training free and inherit from `BasePress` ([source](kvpress/presses/base_press.py)). 
 
-- `RandomPress`: random score
-- `KnormPress`: inverse norm of the key ([paper](https://arxiv.org/abs/2406.11430))
-- `SnapKVPress`: average attention weight of the last queries ([paper](https://arxiv.org/abs/2404.14469))
-- `ExpectedAttentionPress` (ours): expected attention weight during the generation phase  (see [this notebook](notebooks/expected_attention.ipynb))
-- `StreamingLLMPress`: keep only the initial and recent tokens ([paper](https://arxiv.org/abs/2309.17453))
-- `TOVAPress`: attention weight of the last query averaged across heads ([paper](https://arxiv.org/abs/2401.06104))
-- `ObservedAttentionPress`: average attention weight observed during in pre-filling phase (similar to [H2O](https://arxiv.org/abs/2306.14048))
+Several presses inherit from `ScorerPress` ([source](kvpress/presses/scorer_press.py)) and rely on a score to prune the KV pairs with lowest importance:
+
+- `RandomPress` ([source](kvpress/presses/random_press.py)): random score
+- `KnormPress` ([source](kvpress/presses/knorm_press.py), [paper](https://arxiv.org/abs/2406.11430)): inverse norm of the key
+- `SnapKVPress` ([source](kvpress/presses/snapkv_press.py), [paper](https://arxiv.org/abs/2404.14469)): average attention weight of the last queries
+- `ExpectedAttentionPress` ([source](kvpress/presses/expected_attention_press.py), [notebook](notebooks/expected_attention.ipynb)): expected attention weight during the generation phase 
+- `StreamingLLMPress` ([source](kvpress/presses/streaming_llm_press.py), [paper](https://arxiv.org/abs/2309.17453)): keep only the initial and recent tokens 
+- `TOVAPress` ([source](kvpress/presses/tova_press.py), [paper](https://arxiv.org/abs/2401.06104)): attention weight of the last query averaged across heads 
+- `ObservedAttentionPress` ([source](kvpress/presses/observed_attention_press.py), [paper](https://arxiv.org/abs/2306.14048)): average attention weight observed during in pre-filling phase
 
 Some presses rely on a different logic:
-- `ThinKPress`: compress the dimensions of the keys based on the channel attention score on the last queries ([paper](https://arxiv.org/pdf/2407.21018))
-- `SimLayerKVPress`: identify "lazy" layers, and apply the StreamingLLM approach to them ([paper](https://arxiv.org/abs/2410.13846))
-
-Finally we provide special presses:
-- `AdaKVPress`: prune bottom scores of any `ScorerPress` but across all heads, achieving head-wise compressions ([paper](https://arxiv.org/abs/2407.11550))
-- `PerLayerCompressionPress`: compress each layer with a different compression ratio (experimental). This press can be used with any other press that allows to set a compression_ratio
-- `ComposedPress`: compose multiple presses together by chaining their forward hooks
-- `KeyRerotationPress`: rerotate pruned keys to have continuous RoPE embeddings. This press can be used with any other press that inherits from `ScorerPress`.
-- `ChunkPress`: compress the KV cache on each sequence chunk separately. This can yield to more uniform compression across long sequences. The press can be used with any other press that inherits from `ScorerPress`. The method was introduced [here](https://direct.mit.edu/tacl/article/doi/10.1162/tacl_a_00716/125280).
+- `ThinKPress` ([source](kvpress/presses/think_press.py), [paper](https://arxiv.org/pdf/2407.21018)): compress the dimensions of the keys based on the channel attention score on the last queries 
+- `SimLayerKVPress` ([source](kvpress/presses/simlayerkv_press.py), [paper](https://arxiv.org/abs/2410.13846)): identify "lazy" layers, and apply the StreamingLLM approach to them 
+
+Finally we provide wrapper presses that can be combined with other presses:
+- `AdaKVPress` ([source](kvpress/presses/adakv_press.py), [paper](https://arxiv.org/abs/2407.11550)): prune bottom scores of any `ScorerPress` but across all heads, achieving head-wise compressions 
+- `PerLayerCompressionPress` ([source](kvpress/presses/per_layer_compression_press.py)): compress each layer with a different compression ratio (experimental)
+- `ComposedPress` ([source](kvpress/presses/composed_press.py)): compose multiple presses together by chaining their forward hooks
+- `KeyRerotationPress` ([source](kvpress/presses/key_rerotation_press.py)): rerotate pruned keys to have continuous RoPE embeddings
+- `ChunkPress` ([source](kvpress/presses/chunk_press.py), [paper](https://direct.mit.edu/tacl/article/doi/10.1162/tacl_a_00716/125280)): compress the KV cache on each sequence chunk separately. This can yield to more uniform compression across long sequences
 
 For a detailed list of existing KV cache compression methods, check [Awesome-KV-Cache-Compression](https://github.com/October2001/Awesome-KV-Cache-Compression) or [Awesome-LLM-Compression](https://github.com/HuangOwen/Awesome-LLM-Compression?tab=readme-ov-file#kv-cache-compression)