Updates for Sep 25

Cleanup up some formatting and noticed missing Granite info.

Signed-off-by: JJ Asghar <[email protected]>

Author: jjasghar

docs/lab-2/README.md

@@ -160,7 +160,8 @@ take a look at the next screen shot:
 
 ![comment_code](../images/comment_code.png)
 
-It does some amazing work for you code, and really finally, take a look at [this video](https://www.youtube.com/watch?v=V3Yq6w9QaxI) if you want a quick video of other neat https://continue.dev functions we didn't go over.
+It does some amazing work for you code, and really finally, take a look at [this video](https://www.youtube.com/watch?v=V3Yq6w9QaxI)
+if you want a quick video of other neat <https://continue.dev> functions we didn't go over.
 
 On to the next lab!
 

docs/lab-3/getting-started-ilab.md renamed to docs/lab-3/README.md

@@ -16,6 +16,8 @@ it to have.
 
 !!! warning
     If you are running on Windows, you should use WSL2 for this, but it _is unsupported_ but we will do our best to get you success if you run into any problems.
+    **Note**: You will need to use something _other_ then `pip install 'instructlab[mps]'`, it depends on your GPU, and hardware. You'll need to do some homework
+    here.
 
 ## tl;dr
 
@@ -43,7 +45,8 @@ ilab model serve --model-path instructlab-granite-7b-lab-trained/instructlab-gra
 
 ### Install `ilab`
 
-1) Create a new directory called `instructlab` to store the files the `ilab` CLI needs when running and `cd` into the directory by running the following command:
+1) Open up `iTerm2` or `Terminal` (assuming you're on a Mac and we'll continue assuming this) and create a new directory called `instructlab` to store the files the `ilab`
+CLI needs when running and `cd` into the directory by running the following command:
 
 ```shell
 mkdir instructlab
@@ -249,10 +252,13 @@ After running `ilab config init` your directories will look like the following o
 ├─ ~/Library/Application\ Support/instructlab/checkpoints (4)
 ```
 
- 1) `/Users/USERNAME/Library/Caches/instructlab/models/`: Contains all downloaded large language models, including the saved output of ones you generate with ilab.
- 2) `~/Library/Application\ Support/instructlab/datasets/`: Contains data output from the SDG phase, built on modifications to the taxonomy repository.
- 3) `~/Library/Application\ Support/instructlab/taxonomy/`: Contains the skill and knowledge data.
- 4) `~/Users/USERNAME/Library/Caches/instructlab/checkpoints/`: Contains the output of the training process
+1) `/Users/USERNAME/Library/Caches/instructlab/models/`: Contains all downloaded large language models, including the saved output of ones you generate with ilab.
+
+2) `~/Library/Application\ Support/instructlab/datasets/`: Contains data output from the SDG phase, built on modifications to the taxonomy repository.
+
+3) `~/Library/Application\ Support/instructlab/taxonomy/`: Contains the skill and knowledge data.
+
+4) `~/Users/USERNAME/Library/Caches/instructlab/checkpoints/`: Contains the output of the training process
 
 ### 📥 Download the model
 
@@ -311,13 +317,13 @@ ilab model list
 ilab model serve
 ```
 
-erve a non-default model (e.g. Mixtral-8x7B-Instruct-v0.1):
+Serve a non-default model (e.g. Mixtral-8x7B-Instruct-v0.1):
 
 ```shell
 ilab model serve --model-path models/mixtral-8x7b-instruct-v0.1.Q4_K_M.gguf
 ```
 
-nce the model is served and ready, you'll see the following output:
+Once the model is served and ready, you'll see the following output:
 
 ```shell
 (venv) $ ilab model serve
@@ -330,7 +336,10 @@ Press CTRL+C to shut down the server.
 !!! note
     If multiple `ilab` clients try to connect to the same InstructLab server at the same time, the 1st will connect to the server while the others will start their own temporary server. This will require additional resources on the host machine.
 
-- Serve a non-default Safetensors model (e.g. granite-7b-lab). NOTE: this requires a GPU.
+- Serve a non-default Safetensors model (e.g. granite-7b-lab).
+
+!!! tip
+    `vllm` requires a GPU for the best success.
 
 Ensure vllm is installed:
 
@@ -384,3 +393,4 @@ Before you start adding new skills and knowledge to your model, you can check it
 ╰─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────── elapsed 12.008 seconds ─╯
 ```
 
+Nice! Now that we have InstructLab set up correctly, and we're talking to the model, now lets fine tune it with Lab 4!

docs/lab-4/README.md

@@ -14,6 +14,10 @@ ilab model chat
 /q
 ```
 
+
+TODO: WE NEED TO PUT IN A BLURB ABOUT GRANITE HERE
+
+
 ## Prepare to train your model
 
 !!! note
@@ -310,7 +314,7 @@ After you've built a good knowledge submission like above, the `qna.yaml`, the `
 finally the hosted `.md` file, you need to tell the teacher model to build questions around
 your seeded ones. Lets do that now.
 
-1. If you haven't yet, you'll need to pull down the default teacher model, this is done with this command:
+1) If you haven't yet, you'll need to pull down the default teacher model, this is done with this command:
 
 ```bash
 ilab model download
@@ -324,17 +328,18 @@ ilab serve --model-path models/merlinite-7b-lab-Q4_K_M.gguf
 # in another terminal
 ilab model chat
 ```
+
 After the model is working as expected, ie running "who is batman?" close out the `ilab serve` and `ilab model chat`.
 
-2. Next we need to generate the data, this is done with the following command:
+2) Next we need to generate the data, this is done with the following command:
 
 ```bash
 ilab data generate
 ```
 
 This can take some time, take note of the time in the right hand corner, this is building 1000 questions off of your initial 15.
 
-3. After this is complete, now we'll need to train the actual model. If this isn't a Mac M3, this will take **at least an hour**, so
+3) After this is complete, now we'll need to train the actual model. If this isn't a Mac M3, this will take **at least an hour**, so
 hopefully you can take a lunch break or something while this is running.
 
 ```bash
@@ -343,12 +348,13 @@ ilab model train --model-dir instructlab/granite-7b-lab --tokenizer-dir models/g
 
 This takes the granite model, leverages the tokenized version of it, and runs the SDG from the `generate` command against it.
 
-4. When this is completed, you'll need to test this model, which is the following command:
+4) When this is completed, you'll need to test this model, which is the following command:
+
 ```bash
 ilab model test --model-dir instructlab-granite-7b-lab-mlx-q
 ```
 
-5. Now to run the command on the Mac M3, or Apple hardware you'll need to convert it to a `gguf`, that is this next command.
+5) Now to run the command on the Mac M3, or Apple hardware you'll need to convert it to a `gguf`, that is this next command.
 
 !!! note
     You won't need to do this if you are running on Linux (or maybe Windows remember that's unsupported at the moment)
@@ -357,7 +363,7 @@ ilab model test --model-dir instructlab-granite-7b-lab-mlx-q
 ilab model convert --model-dir instructlab-granite-7b-lab-mlx-q
 ```
 
-6. Finally run the new model with `ilab model serve`.
+6) Finally run the new model with `ilab model serve`.
 
 ```bash
 ilab model serve --model-path instructlab-granite-7b-lab-trained/instructlab-granite-7b-lab-Q4_K_M.gguf
@@ -371,5 +377,4 @@ though your submission is great for the upstream model, and extremely valuable t
 When the full run from the upstream happens, the PR you submit with the new (or corrected) knowledge will be "baked in" better then the quantization
 method you use here, which will give much higher percentage of retrieval.
 
-
 <img src="https://count.asgharlabs.io/count?p=/lab4_opensource_ai_page>