Upload folder using huggingface_hub

README.md

@@ -1,19 +1,17 @@
 ---
-license: llama3.2
+license: other
 language:
-- en
+  - en
 pipeline_tag: text-generation
 library_name: mlx
 tags:
-- mlx
-- svg
-- code-generation
-- graphics
-- vector-graphics
-- apple-silicon
-- llama
-base_model:
-- meta-llama/Llama-3.2-3B-Instruct
+  - mlx
+  - svg
+  - code-generation
+  - graphics
+  - vector-graphics
+  - apple-silicon
+  - llama
 ---
 
 # SVG-Master
@@ -38,9 +36,9 @@ Built on top of **Llama 3.2 3B Instruct**, this model has been instruction-tuned
 | **Base Model**         | Llama-3.2-3B-Instruct-4bit          |
 | **Model Size**         | 3.2 Billion Parameters              |
 | **Fine-tuning Method** | LoRA (QLoRA)                        |
-| **LoRA Config**        | 16 layers, Rank 8, Alpha/Scale 20   |
-| **Training Data**      | 2,000 High-quality SVG/Prompt pairs |
-| **Training Steps**     | 600 Iterations                      |
+| **LoRA Config**        | 16 layers, Rank 32, Alpha 64        |
+| **Training Data**      | 3,000 High-quality SVG/Prompt pairs |
+| **Training Steps**     | 2,000 Iterations                    |
 | **Optimizer**          | Adam (Learning Rate: 1e-4)          |
 | **Context Window**     | 128k Tokens (Llama 3.2 default)     |
 
@@ -52,10 +50,10 @@ Designed for Apple Silicon, use `mlx-lm` for the fastest experience.
 
 ```bash
 # Using uv (recommended)
-uv run mlx_lm.generate --model ./weights --adapter-path ./adapters --prompt "A minimalist sunset over a calm ocean"
+uv run mlx_lm.generate --model ./base_weights --adapter-path ./adapters --prompt "A minimalist sunset over a calm ocean"
 
 # Using standard python
-python -m mlx_lm.generate --model ./weights --adapter-path ./adapters --prompt "A minimalist sunset over a calm ocean"
+python -m mlx_lm.generate --model ./base_weights --adapter-path ./adapters --prompt "A minimalist sunset over a calm ocean"
 ```
 
 ### 2. Ollama Integration
@@ -77,6 +75,14 @@ response = generate(model, tokenizer, prompt="A golden trophy icon", max_tokens=
 print(response)
 ```
 
+## 📂 Project Structure
+
+- `base_weights/`: Original Llama-3.2-3B-Instruct-4bit weights.
+- `adapters/`: Fine-tuned LoRA adapter weights and configuration (Rank 32).
+- `data/`: Cleaned and balanced training/validation datasets.
+- `scripts/`: Automation scripts for training, verification, and export.
+- `weight_fused/`: Fully fused, dequantized GGUF/Safetensors models (for export).
+
 ## 📜 License
 
 This model inherits the [Llama 3.2 Community License Agreement](https://www.llama.com/llama3_2/license/).
@@ -85,4 +91,4 @@ This model inherits the [Llama 3.2 Community License Agreement](https://www.llam
 
 - **HuggingFace**: [fahidnasir/SVG-Master](https://huggingface.co/fahidnasir/SVG-Master)
 - **Kaggle**: [nfahid/svg-master](https://www.kaggle.com/models/nfahid/svg-master)
-- **Ollama**: [fahidnasir/svg-master](https://ollama.com/fahidnasir/svg-master)
+- **Ollama**: [fahidnasir/svg-master](https://ollama.com/fahidnasir/svg-master)

adapters/adapter_config.json

@@ -1,40 +1,44 @@
 {
-  "adapter_path": "svg-master/adapters",
-  "batch_size": 4,
-  "config": null,
-  "data": "svg-master/data",
-  "fine_tune_type": "lora",
-  "grad_accumulation_steps": 1,
-  "grad_checkpoint": false,
-  "iters": 600,
-  "learning_rate": 0.0001,
-  "lora_parameters": {
-    "rank": 8,
-    "dropout": 0.0,
-    "scale": 20.0
-  },
-  "lr_schedule": null,
-  "mask_prompt": false,
-  "max_seq_length": 2048,
-  "model": "mlx-community/Llama-3.2-3B-Instruct-4bit",
-  "num_layers": 16,
-  "optimizer": "adam",
-  "optimizer_config": {
-    "adam": {},
-    "adamw": {},
-    "muon": {},
-    "sgd": {},
-    "adafactor": {}
-  },
-  "project_name": null,
-  "report_to": null,
-  "resume_adapter_file": null,
-  "save_every": 100,
-  "seed": 0,
-  "steps_per_eval": 50,
-  "steps_per_report": 10,
-  "test": false,
-  "test_batches": 500,
-  "train": true,
-  "val_batches": 10
-}
+    "adapter_path": "./adapters",
+    "alpha": 64,
+    "batch_size": 4,
+    "config": "scripts/lora_config.yaml",
+    "data": "./data",
+    "dropout": 0.05,
+    "fine_tune_type": "lora",
+    "grad_accumulation_steps": 1,
+    "grad_checkpoint": true,
+    "iters": 2000,
+    "learning_rate": 0.0001,
+    "lora_layers": 16,
+    "lora_parameters": {
+        "rank": 8,
+        "dropout": 0.0,
+        "scale": 20.0
+    },
+    "lr_schedule": null,
+    "mask_prompt": false,
+    "max_seq_length": 2048,
+    "model": "./base_weights",
+    "num_layers": 16,
+    "optimizer": "adam",
+    "optimizer_config": {
+        "adam": {},
+        "adamw": {},
+        "muon": {},
+        "sgd": {},
+        "adafactor": {}
+    },
+    "project_name": null,
+    "rank": 32,
+    "report_to": null,
+    "resume_adapter_file": null,
+    "save_every": 200,
+    "seed": 0,
+    "steps_per_eval": 200,
+    "steps_per_report": 10,
+    "test": false,
+    "test_batches": 500,
+    "train": true,
+    "val_batches": 25
+}

adapters/adapters.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2eeea10a717d29f43aa3cf0c30b687d2ef3def8c41d846b693e39cbe1d7ab8a4
+oid sha256:641f49fc49c01f5cc5c55e17eb7c4976c9bcafb19dbee0d01b759df873ee9c73
 size 27811400

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2322764a26305efe30c88e111faf1c00a932bcac0b349a61e7050100e38099b1
+oid sha256:d75e1ee0ea653cc5b76191ec934c7c0d568e94d4e47846619f1f4bc715b7b265
 size 1807496278

model.safetensors.index.json

@@ -1,7 +1,6 @@
 {
     "metadata": {
-        "total_size": 1807423488,
-        "total_parameters": 3212749824
+        "total_size": 1807423488
     },
     "weight_map": {
         "model.embed_tokens.biases": "model.safetensors",

tokenizer_config.json

@@ -2050,6 +2050,7 @@
     }
   },
   "bos_token": "<|begin_of_text|>",
+  "chat_template": "{{- bos_token }}\n{%- if custom_tools is defined %}\n    {%- set tools = custom_tools %}\n{%- endif %}\n{%- if not tools_in_user_message is defined %}\n    {%- set tools_in_user_message = true %}\n{%- endif %}\n{%- if not date_string is defined %}\n    {%- if strftime_now is defined %}\n        {%- set date_string = strftime_now(\"%d %b %Y\") %}\n    {%- else %}\n        {%- set date_string = \"26 Jul 2024\" %}\n    {%- endif %}\n{%- endif %}\n{%- if not tools is defined %}\n    {%- set tools = none %}\n{%- endif %}\n\n{#- This block extracts the system message, so we can slot it into the right place. #}\n{%- if messages[0]['role'] == 'system' %}\n    {%- set system_message = messages[0]['content']|trim %}\n    {%- set messages = messages[1:] %}\n{%- else %}\n    {%- set system_message = \"\" %}\n{%- endif %}\n\n{#- System message #}\n{{- \"<|start_header_id|>system<|end_header_id|>\\n\\n\" }}\n{%- if tools is not none %}\n    {{- \"Environment: ipython\\n\" }}\n{%- endif %}\n{{- \"Cutting Knowledge Date: December 2023\\n\" }}\n{{- \"Today Date: \" + date_string + \"\\n\\n\" }}\n{%- if tools is not none and not tools_in_user_message %}\n    {{- \"You have access to the following functions. To call a function, please respond with JSON for a function call.\" }}\n    {{- 'Respond in the format {\"name\": function name, \"parameters\": dictionary of argument name and its value}.' }}\n    {{- \"Do not use variables.\\n\\n\" }}\n    {%- for t in tools %}\n        {{- t | tojson(indent=4) }}\n        {{- \"\\n\\n\" }}\n    {%- endfor %}\n{%- endif %}\n{{- system_message }}\n{{- \"<|eot_id|>\" }}\n\n{#- Custom tools are passed in a user message with some extra guidance #}\n{%- if tools_in_user_message and not tools is none %}\n    {#- Extract the first user message so we can plug it in here #}\n    {%- if messages | length != 0 %}\n        {%- set first_user_message = messages[0]['content']|trim %}\n        {%- set messages = messages[1:] %}\n    {%- else %}\n        {{- raise_exception(\"Cannot put tools in the first user message when there's no first user message!\") }}\n{%- endif %}\n    {{- '<|start_header_id|>user<|end_header_id|>\\n\\n' -}}\n    {{- \"Given the following functions, please respond with a JSON for a function call \" }}\n    {{- \"with its proper arguments that best answers the given prompt.\\n\\n\" }}\n    {{- 'Respond in the format {\"name\": function name, \"parameters\": dictionary of argument name and its value}.' }}\n    {{- \"Do not use variables.\\n\\n\" }}\n    {%- for t in tools %}\n        {{- t | tojson(indent=4) }}\n        {{- \"\\n\\n\" }}\n    {%- endfor %}\n    {{- first_user_message + \"<|eot_id|>\"}}\n{%- endif %}\n\n{%- for message in messages %}\n    {%- if not (message.role == 'ipython' or message.role == 'tool' or 'tool_calls' in message) %}\n        {{- '<|start_header_id|>' + message['role'] + '<|end_header_id|>\\n\\n'+ message['content'] | trim + '<|eot_id|>' }}\n    {%- elif 'tool_calls' in message %}\n        {%- if not message.tool_calls|length == 1 %}\n            {{- raise_exception(\"This model only supports single tool-calls at once!\") }}\n        {%- endif %}\n        {%- set tool_call = message.tool_calls[0].function %}\n        {{- '<|start_header_id|>assistant<|end_header_id|>\\n\\n' -}}\n        {{- '{\"name\": \"' + tool_call.name + '\", ' }}\n        {{- '\"parameters\": ' }}\n        {{- tool_call.arguments | tojson }}\n        {{- \"}\" }}\n        {{- \"<|eot_id|>\" }}\n    {%- elif message.role == \"tool\" or message.role == \"ipython\" %}\n        {{- \"<|start_header_id|>ipython<|end_header_id|>\\n\\n\" }}\n        {%- if message.content is mapping or message.content is iterable %}\n            {{- message.content | tojson }}\n        {%- else %}\n            {{- message.content }}\n        {%- endif %}\n        {{- \"<|eot_id|>\" }}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|start_header_id|>assistant<|end_header_id|>\\n\\n' }}\n{%- endif %}\n",
   "clean_up_tokenization_spaces": true,
   "eos_token": "<|eot_id|>",
   "extra_special_tokens": {},