prithivMLmods commited on
Commit
d5b72b0
1 Parent(s): 33c7197

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +59 -1
README.md CHANGED
@@ -15,7 +15,9 @@ tags:
15
  - Extended-Stream
16
  - text-generation-inference
17
  ---
18
- ### Llama-Song-Stream-3B-Instruct
 
 
19
 
20
  | **File Name** | **Size** | **Description** | **Upload Status** |
21
  |----------------------------------------|--------------------|--------------------------------------------------|--------------------|
@@ -30,4 +32,60 @@ tags:
30
  | `tokenizer.json` | 17.2 MB | Tokenizer file (large LFS model tokenizer data). | Uploaded (LFS) |
31
  | `tokenizer_config.json` | 57.4 kB | Tokenizer configuration settings. | Uploaded |
32
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
33
  ---
 
15
  - Extended-Stream
16
  - text-generation-inference
17
  ---
18
+ ### **Llama-Song-Stream-3B-Instruct Model Card**
19
+
20
+ The **Llama-Song-Stream-3B-Instruct** is a fine-tuned language model built upon **meta-llama/Llama-3.2-3B-Instruct**. It is specifically trained on song lyrics generation tasks, utilizing chain-of-thought reasoning over lyrical datasets.
21
 
22
  | **File Name** | **Size** | **Description** | **Upload Status** |
23
  |----------------------------------------|--------------------|--------------------------------------------------|--------------------|
 
32
  | `tokenizer.json` | 17.2 MB | Tokenizer file (large LFS model tokenizer data). | Uploaded (LFS) |
33
  | `tokenizer_config.json` | 57.4 kB | Tokenizer configuration settings. | Uploaded |
34
 
35
+ ---
36
+
37
+ ## **Model Details**
38
+
39
+ ### **Key Metrics:**
40
+ - **Base Model:** `meta-llama/Llama-3.2-3B-Instruct`
41
+ - **Model Parameters:** 3B (billion parameters).
42
+ - **Fine-tuned dataset focus:** Song generation and lyric-based chain-of-thought reasoning.
43
+
44
+ ---
45
+
46
+ ### **Model Components**
47
+ 1. **Model Weights:**
48
+ - Split into two LFS shards:
49
+ - `pytorch_model-00001-of-00002.bin` - **4.97 GB**
50
+ - `pytorch_model-00002-of-00002.bin` - **1.46 GB**
51
+
52
+ 2. **Tokenizer Data:**
53
+ - Tokenizer includes LFS model configuration:
54
+ - `tokenizer.json` - **17.2 MB**
55
+ - `special_tokens_map.json` - **477 Bytes**
56
+ - `tokenizer_config.json` - **57.4 KB**
57
+
58
+ 3. **Configuration Files:**
59
+ - `config.json` - Model settings (**1.03 KB**).
60
+ - `generation_config.json` - Inference task parameters (**248 Bytes**).
61
+
62
+ ---
63
+
64
+ ### **Training Dataset**
65
+ - **Dataset Name:** [prithivMLmods/Song-Catalogue-Long-Thought](https://huggingface.co/datasets/prithivMLmods/Song-Catalogue-Long-Thought)
66
+ - **Total Examples:** 57,700+
67
+ - **Training Focus:** Chain-of-thought reasoning related to lyrical themes and patterns.
68
+
69
+ ---
70
+
71
+ ### **Intended Use Cases**
72
+ 1. **Song Lyrics Generation:**
73
+ Generate realistic, context-aware song lyrics from user prompts.
74
+
75
+ 2. **Creative Writing Tools:**
76
+ Aiding songwriters and lyricists by generating thematic drafts.
77
+
78
+ 3. **Text Manipulation via Prompts:**
79
+ Experiment with different styles, song structures, and lyrical themes.
80
+
81
+ ---
82
+
83
+ ### **Current Status:**
84
+ - **Inference API Status:**
85
+ The model lacks sufficient downloads or visibility for deployment to Hugging Face's Inference API.
86
+ - **Action Plan:** Increase visibility through applications and outreach.
87
+
88
+ - **Model Deployment Options:**
89
+ Use dedicated Inference Endpoints for direct access and deployment.
90
+
91
  ---