Add ZipNN stuff

README.md

@@ -5,13 +5,8 @@ license: apache-2.0
 datasets:
 - codeparrot/github-code-clean
 - bigcode/starcoderdata
-# - Stackexchange
-# - CommonCrawl
 - open-web-math/open-web-math
 - math-ai/StackMathQA
-# - Arxiv
-# - Wikipedia
-# - conceptofmind/FLAN_2022 # Original link is broken, we used IBM's filtered version
 metrics:
 - code_eval
 library_name: transformers
@@ -24,18 +19,18 @@ model-index:
   - task:
       type: text-generation
     dataset:
-        type: bigcode/humanevalpack 
-        name: HumanEvalSynthesis (Python)
+      type: bigcode/humanevalpack
+      name: HumanEvalSynthesis (Python)
     metrics:
     - name: pass@1
       type: pass@1
-      value: 36.0
+      value: 36
       verified: false
   - task:
       type: text-generation
     dataset:
-        type: bigcode/humanevalpack  
-        name: HumanEvalSynthesis (Average)
+      type: bigcode/humanevalpack
+      name: HumanEvalSynthesis (Average)
     metrics:
     - name: pass@1
       type: pass@1
@@ -44,8 +39,8 @@ model-index:
   - task:
       type: text-generation
     dataset:
-        type: bigcode/humanevalpack  
-        name:  HumanEvalExplain (Average)
+      type: bigcode/humanevalpack
+      name: HumanEvalExplain (Average)
     metrics:
     - name: pass@1
       type: pass@1
@@ -54,8 +49,8 @@ model-index:
   - task:
       type: text-generation
     dataset:
-        type: bigcode/humanevalpack  
-        name:  HumanEvalFix (Average)
+      type: bigcode/humanevalpack
+      name: HumanEvalFix (Average)
     metrics:
     - name: pass@1
       type: pass@1
@@ -64,135 +59,193 @@ model-index:
   - task:
       type: text-generation
     dataset:
-        type: repoqa  
-        name:  RepoQA (Python@16K)
+      type: repoqa
+      name: RepoQA (Python@16K)
     metrics:
     - name: pass@1 (thresh=0.5)
       type: pass@1 (thresh=0.5)
-      value: 40.0
+      value: 40
       verified: false
   - task:
       type: text-generation
     dataset:
-        type: repoqa  
-        name:  RepoQA (C++@16K)
+      type: repoqa
+      name: RepoQA (C++@16K)
     metrics:
     - name: pass@1 (thresh=0.5)
       type: pass@1 (thresh=0.5)
-      value: 36.0
+      value: 36
       verified: false
   - task:
       type: text-generation
     dataset:
-        type: repoqa  
-        name:  RepoQA (Java@16K)
+      type: repoqa
+      name: RepoQA (Java@16K)
     metrics:
     - name: pass@1 (thresh=0.5)
       type: pass@1 (thresh=0.5)
-      value: 37.0
+      value: 37
       verified: false
   - task:
       type: text-generation
     dataset:
-        type: repoqa  
-        name:  RepoQA (TypeScript@16K)
+      type: repoqa
+      name: RepoQA (TypeScript@16K)
     metrics:
     - name: pass@1 (thresh=0.5)
       type: pass@1 (thresh=0.5)
-      value: 27.0
+      value: 27
       verified: false
   - task:
       type: text-generation
     dataset:
-        type: repoqa  
-        name:  RepoQA (Rust@16K)
+      type: repoqa
+      name: RepoQA (Rust@16K)
     metrics:
     - name: pass@1 (thresh=0.5)
       type: pass@1 (thresh=0.5)
-      value: 29.0
+      value: 29
       verified: false
   - task:
       type: text-generation
     dataset:
-        type: lcc  
-        name:  LCC (Balanced)
+      type: lcc
+      name: LCC (Balanced)
     metrics:
-    - name: Exact Match@4K 
+    - name: Exact Match@4K
       type: Exact Match@4K
       value: 54.6
       verified: false
   - task:
       type: text-generation
     dataset:
-        type: lcc  
-        name:  LCC (Balanced)
+      type: lcc
+      name: LCC (Balanced)
     metrics:
-    - name: Exact Match@8K 
+    - name: Exact Match@8K
       type: Exact Match@8K
       value: 56.8
       verified: false
   - task:
       type: text-generation
     dataset:
-        type: lcc  
-        name:  LCC (Balanced)
+      type: lcc
+      name: LCC (Balanced)
     metrics:
-    - name: Exact Match@16K 
+    - name: Exact Match@16K
       type: Exact Match@16K
       value: 52.2
       verified: false
   - task:
       type: text-generation
     dataset:
-        type: lcc  
-        name:  LCC (Balanced)
+      type: lcc
+      name: LCC (Balanced)
     metrics:
-    - name: Exact Match@32K 
+    - name: Exact Match@32K
       type: Exact Match@32K
       value: 57.8
       verified: false
   - task:
       type: text-generation
     dataset:
-        type: repobench  
-        name: RepoBench-P (Balanced)
+      type: repobench
+      name: RepoBench-P (Balanced)
     metrics:
-    - name: Exact Match@4K 
+    - name: Exact Match@4K
       type: Exact Match@4K
       value: 39.8
       verified: false
   - task:
       type: text-generation
     dataset:
-        type: repobench  
-        name: RepoBench-P (Balanced)
+      type: repobench
+      name: RepoBench-P (Balanced)
     metrics:
-    - name: Exact Match@8K 
+    - name: Exact Match@8K
       type: Exact Match@8K
       value: 46.8
       verified: false
   - task:
       type: text-generation
     dataset:
-        type: repobench  
-        name: RepoBench-P (Balanced)
+      type: repobench
+      name: RepoBench-P (Balanced)
     metrics:
-    - name: Exact Match@16K 
+    - name: Exact Match@16K
       type: Exact Match@16K
       value: 43.1
       verified: false
   - task:
       type: text-generation
     dataset:
-        type: repobench  
-        name: RepoBench-Pn(Balanced)
+      type: repobench
+      name: RepoBench-Pn(Balanced)
     metrics:
-    - name: Exact Match@32K 
+    - name: Exact Match@32K
       type: Exact Match@32K
       value: 45.3
       verified: false
+base_model:
+- ibm-granite/granite-3b-code-base-128k
 ---
 
+# Disclaimer and Requirements
+
+This model is a clone of [**ibm-granite/granite-3b-code-base-128k**](https://huggingface.co/ibm-granite/granite-3b-code-base-128k) compressed using ZipNN. Compressed losslessly to 67% its original size, ZipNN saved ~3GB in storage and potentially ~2TB in data transfer **monthly**.
+
+### Requirement
+
+In order to use the model, ZipNN is necessary:
+```bash
+pip install zipnn
+```
+### Use This Model
+```python
+# Use a pipeline as a high-level helper
+from transformers import pipeline
+from zipnn import zipnn_hf
+
+zipnn_hf()
+
+messages = [
+    {"role": "user", "content": "Who are you?"},
+]
+pipe = pipeline("text-generation", model="royleibov/granite-3b-code-base-128k-ZipNN-Compressed")
+pipe(messages)
+```
+```python
+# Load model directly
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from zipnn import zipnn_hf
+
+zipnn_hf()
+
+model = AutoModelForCausalLM.from_pretrained(
+    "royleibov/granite-3b-code-base-128k-ZipNN-Compressed", 
+    device_map="cuda", 
+    torch_dtype="auto", 
+    trust_remote_code=True, 
+)
+tokenizer = AutoTokenizer.from_pretrained("royleibov/granite-3b-code-base-128k-ZipNN-Compressed")
+```
+### ZipNN
+ZipNN also allows you to seemlessly save local disk space in your cache after the model is downloaded.
+
+To compress the cached model, simply run:
+```bash
+python zipnn_compress_path.py safetensors --model royleibov/granite-3b-code-base-128k-ZipNN-Compressed --hf_cache
+```
+
+The model will be decompressed automatically and safely as long as `zipnn_hf()` is added at the top of the file like in the [example above](#use-this-model).
+
+To decompress manualy, simply run:
+```bash
+python zipnn_decompress_path.py --model royleibov/granite-3b-code-base-128k-ZipNN-Compressed --hf_cache
+```
+
+
 ![image/png](https://cdn-uploads.huggingface.co/production/uploads/62cd5057674cdb524450093d/1hzxoPwqkBJXshKVVe6_9.png)
 
 # Granite-3B-Code-Base-128K
@@ -217,11 +270,16 @@ This is a simple example of how to use **Granite-3B-Code-Base-128K** model.
 ```python
 import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer
+from zipnn import zipnn_hf
+
+zipnn_hf()
+
 device = "cuda" # or "cpu"
-model_path = "ibm-granite/granite-3b-code-base-128k"
+model_path = "royleibov/granite-3b-code-base-128k-ZipNN-Compressed"
 tokenizer = AutoTokenizer.from_pretrained(model_path)
 # drop device_map if running on CPU
 model = AutoModelForCausalLM.from_pretrained(model_path, device_map=device)
+
 model.eval()
 # change input text as desired
 input_text = "def generate():"
@@ -246,4 +304,4 @@ Starting from the base Granite model, this model was further pretrained on repos
 We train the Granite Code models using two of IBM's super computing clusters, namely Vela and Blue Vela, both outfitted with NVIDIA A100 and H100 GPUs respectively. These clusters provide a scalable and efficient infrastructure for training our models over thousands of GPUs.
 
 ## Ethical Considerations and Limitations
-The use of Large Language Models involves risks and ethical considerations people must be aware of. Regarding code generation, caution is urged against complete reliance on specific code models for crucial decisions or impactful information as the generated code is not guaranteed to work as intended. **Granite-3B-Code-Base-128K** model is not the exception in this regard. Even though this model is suited for multiple code-related tasks, it has not undergone any safety alignment, there it may produce problematic outputs. Additionally, it remains uncertain whether smaller models might exhibit increased susceptibility to hallucination in generation scenarios by copying source code verbatim from the training dataset due to their reduced sizes and memorization capacities. This aspect is currently an active area of research, and we anticipate more rigorous exploration, comprehension, and mitigations in this domain. Regarding ethics, a latent risk associated with all Large Language Models is their malicious utilization. We urge the community to use **Granite-3B-Code-Base-128K** model with ethical intentions and in a responsible way. 
+The use of Large Language Models involves risks and ethical considerations people must be aware of. Regarding code generation, caution is urged against complete reliance on specific code models for crucial decisions or impactful information as the generated code is not guaranteed to work as intended. **Granite-3B-Code-Base-128K** model is not the exception in this regard. Even though this model is suited for multiple code-related tasks, it has not undergone any safety alignment, there it may produce problematic outputs. Additionally, it remains uncertain whether smaller models might exhibit increased susceptibility to hallucination in generation scenarios by copying source code verbatim from the training dataset due to their reduced sizes and memorization capacities. This aspect is currently an active area of research, and we anticipate more rigorous exploration, comprehension, and mitigations in this domain. Regarding ethics, a latent risk associated with all Large Language Models is their malicious utilization. We urge the community to use **Granite-3B-Code-Base-128K** model with ethical intentions and in a responsible way.