copy

aider/website/_data/quant.yml

@@ -0,0 +1,93 @@
+- dirname: 2024-11-09-11-09-15--Qwen2.5-Coder-32B-Instruct
+  test_cases: 133
+  model: HuggingFace weights via glhf.chat
+  released: 2024-11-12
+  edit_format: diff
+  commit_hash: ec9982a
+  pass_rate_1: 59.4
+  pass_rate_2: 71.4
+  percent_cases_well_formed: 94.7
+  error_outputs: 17
+  num_malformed_responses: 17
+  num_with_malformed_responses: 7
+  user_asks: 1
+  lazy_comments: 0
+  syntax_errors: 0
+  indentation_errors: 0
+  exhausted_context_windows: 0
+  test_timeouts: 3
+  command: aider --model openai/Qwen2.5-Coder-32B-Instruct
+  date: 2024-11-09
+  versions: 0.59.2.dev
+  seconds_per_case: 22.5
+  total_cost: 0.0000
+
+- dirname: 2024-11-20-15-17-37--qwen25-32b-or-diff
+  test_cases: 133
+  model: openrouter/qwen/qwen-2.5-coder-32b-instruct
+  edit_format: diff
+  commit_hash: e917424
+  pass_rate_1: 49.6
+  pass_rate_2: 65.4
+  percent_cases_well_formed: 84.2
+  error_outputs: 43
+  num_malformed_responses: 31
+  num_with_malformed_responses: 21
+  user_asks: 43
+  lazy_comments: 0
+  syntax_errors: 2
+  indentation_errors: 2
+  exhausted_context_windows: 12
+  test_timeouts: 2
+  command: aider --model openrouter/qwen/qwen-2.5-coder-32b-instruct
+  date: 2024-11-20
+  versions: 0.63.3.dev
+  seconds_per_case: 40.7
+  total_cost: 0.1497
+
+- dirname: 2024-09-20-21-47-17--qwen2.5-32b-instruct-q8_0-whole
+  test_cases: 133
+  model: ollama/qwen2.5:32b-instruct-q8_0
+  edit_format: whole
+  commit_hash: 2753ac6
+  pass_rate_1: 46.6
+  pass_rate_2: 58.6
+  percent_cases_well_formed: 100.0
+  error_outputs: 0
+  num_malformed_responses: 0
+  num_with_malformed_responses: 0
+  user_asks: 1
+  lazy_comments: 0
+  syntax_errors: 0
+  indentation_errors: 0
+  exhausted_context_windows: 0
+  test_timeouts: 2
+  command: aider --model ollama/qwen2.5:32b-instruct-q8_0
+  date: 2024-09-20
+  versions: 0.56.1.dev
+  seconds_per_case: 1763.7
+  total_cost: 0.0000
+
+- dirname: 2024-09-30-14-09-43--qwen2.5-32b-whole-2
+  test_cases: 133
+  model: ollama/qwen2.5:32b
+  edit_format: whole
+  commit_hash: 765c4cb
+  pass_rate_1: 44.4
+  pass_rate_2: 54.1
+  percent_cases_well_formed: 100.0
+  error_outputs: 0
+  num_malformed_responses: 0
+  num_with_malformed_responses: 0
+  user_asks: 9
+  lazy_comments: 0
+  syntax_errors: 0
+  indentation_errors: 0
+  exhausted_context_windows: 0
+  test_timeouts: 3
+  command: aider --model ollama/qwen2.5:32b
+  date: 2024-09-30
+  versions: 0.58.1.dev
+  seconds_per_case: 134.9
+  total_cost: 0.0000
+  

aider/website/_includes/quant-chart.js

@@ -20,8 +20,7 @@ document.addEventListener('DOMContentLoaded', function () {
   {% endfor %}
 
   allData.forEach(function(row) {
-    // Split the model name on \n to create array of lines
-    chartData.labels.push(row.model.split('\\n'));
+    chartData.labels.push(row.model);
     chartData.datasets[0].data.push(row.pass_rate_2);
   });
 

aider/website/_posts/2024-11-21-quantization.md

@@ -16,13 +16,28 @@ aider's code editing benchmark, rivaling closed source frontier models.
 But pay attention to how your model is being quantized, as it
 can strongly impact code editing skill.
 Heavily quantized models are often used by cloud API providers
-and local model servers like ollama.
-
-The graph below compares 4 different versions of the Qwen 2.5 32B model,
-served both locally and from cloud providers:
+and local model servers like Ollama.
 
 <canvas id="quantChart" width="800" height="450" style="margin: 20px 0"></canvas>
 <script src="https://cdn.jsdelivr.net/npm/chart.js"></script>
 <script>
 {% include quant-chart.js %}
 </script>
+
+The graph above compares 4 different versions of the Qwen 2.5 32B model,
+served both locally and from cloud providers.
+
+- The [HuggingFace weights](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct) served via [glhf.chat](https://glhf.chat).
+- The results from [OpenRouter's mix of providers](https://openrouter.ai/qwen/qwen-2.5-coder-32b-instruct/providers).
+- Two Ollama models run locally.
+
+The best version of the model rivals GPT-4o, while the worst performer
+is more like GPT-3.5 Turbo.
+
+
+## Choosing providers with OpenRouter
+
+OpenRouter allows you to ignore specific providers in your
+[preferences](https://openrouter.ai/settings/preferences).
+This can be effective to exclude highly quantized or otherwise
+undesirable providers.